Merge branch 'dev-stable-tc-scf' of https://github.com/AbdAmmar/qp2 into dev-stable-tc-scf

configure

@@ -215,7 +215,6 @@ EOF
               cd trexio-${VERSION}
               ./configure --prefix=\${QP_ROOT} --without-hdf5
               make -j 8 && make -j 8 check && make -j 8 install
-              cp ${QP_ROOT}/include/trexio_f.f90 ${QP_ROOT}/src/ezfio_files
               tar -zxvf "\${QP_ROOT}"/external/qp2-dependencies/${ARCHITECTURE}/ninja.tar.gz 
               mv ninja "\${QP_ROOT}"/bin/
 EOF
@@ -229,7 +228,6 @@ EOF
               cd trexio-${VERSION}
               ./configure --prefix=\${QP_ROOT}
               make -j 8 && make -j 8 check && make -j 8 install
-              cp ${QP_ROOT}/include/trexio_f.f90 ${QP_ROOT}/src/ezfio_files
 EOF
  
 

ocaml/Input_ao_basis.ml

@@ -44,8 +44,12 @@ end = struct
   let get_default = Qpackage.get_ezfio_default "ao_basis";;
 
   let read_ao_basis () =
-    Ezfio.get_ao_basis_ao_basis ()
+    let result =
-    |> AO_basis_name.of_string
+      Ezfio.get_ao_basis_ao_basis ()
+    in
+    if result <> "None" then
+      AO_basis_name.of_string result
+    else failwith "No basis"
   ;;
 
   let read_ao_num () =
@@ -267,7 +271,10 @@ end = struct
         |> Ezfio.set_ao_basis_ao_md5 ;
         Some result
       with
-      | _ -> (Ezfio.set_ao_basis_ao_md5 "None" ; None)
+      | _ -> ( "None"
+               |> Digest.string
+               |> Digest.to_hex
+               |> Ezfio.set_ao_basis_ao_md5 ; None)
   ;;
 
 

ocaml/Input_mo_basis.ml

@@ -56,7 +56,10 @@ end = struct
   let read_ao_md5 () =
     let ao_md5 =
       match (Input_ao_basis.Ao_basis.read ()) with
-      | None -> failwith "Unable to read AO basis"
+      | None -> ("None"
+                 |> Digest.string
+                 |> Digest.to_hex
+                 |> MD5.of_string)
       | Some result -> Input_ao_basis.Ao_basis.to_md5 result
     in
     let result =

ocaml/qp_run.ml

@@ -38,7 +38,8 @@ let run slave ?prefix exe ezfio_file =
       | Unix.Unix_error _ -> try_new_port (port_number+100)
     in
     let result =
-      try_new_port 41279
+      let port = 10*(Unix.getpid () mod 2823) + 32_769 in
+      try_new_port port
     in
     Zmq.Socket.close dummy_socket;
     Zmq.Context.terminate zmq_context;

scripts/compilation/cache_compile.py

@@ -1,7 +1,7 @@
 #!/usr/bin/env python3
 """
 Save the .o from a .f90
-and is the .o is asked a second time, retur it
+and is the .o is asked a second time, return it
 Take in argv command like:
      ifort -g  -openmp -I IRPF90_temp/Ezfio_files/ -c IRPF90_temp/Integrals_Monoelec/kin_ao_ints.irp.module.F90 -o IRPF90_temp/Integrals_Monoelec/kin_ao_ints.irp.module.o
 """

scripts/qp_import_trexio.py

@@ -13,11 +13,17 @@ Options:
 
 import sys
 import os
-import trexio
 import numpy as np
 from functools import reduce
 from ezfio import ezfio
 from docopt import docopt
+import qp_bitmasks
+
+try:
+  import trexio
+except ImportError:
+    print("Error: trexio python module is not found. Try python3 -m pip install trexio")
+    sys.exit(1)
 
 
 try:
@@ -90,14 +96,15 @@ def write_ezfio(trexio_filename, filename):
         p = re.compile(r'(\d*)$')
         label = [p.sub("", x).capitalize() for x in label]
         ezfio.set_nuclei_nucl_label(label)
+        print("OK")
 
     else:
         ezfio.set_nuclei_nucl_num(1)
         ezfio.set_nuclei_nucl_charge([0.])
         ezfio.set_nuclei_nucl_coord([0.,0.,0.])
         ezfio.set_nuclei_nucl_label(["X"])
+        print("None")
 
-    print("OK")
 
 
     print("Electrons\t...\t", end=' ')
@@ -105,12 +112,12 @@ def write_ezfio(trexio_filename, filename):
     try:
         num_beta = trexio.read_electron_dn_num(trexio_file)
     except:
-        num_beta = sum(charge)//2
+        num_beta = int(sum(charge))//2
 
     try:
         num_alpha = trexio.read_electron_up_num(trexio_file)
     except:
-        num_alpha = sum(charge) - num_beta
+        num_alpha = int(sum(charge)) - num_beta
 
     if num_alpha == 0:
         print("\n\nError: There are zero electrons in the TREXIO file.\n\n")
@@ -118,7 +125,7 @@ def write_ezfio(trexio_filename, filename):
     ezfio.set_electrons_elec_alpha_num(num_alpha)
     ezfio.set_electrons_elec_beta_num(num_beta)
 
-    print("OK")
+    print(f"{num_alpha} {num_beta}")
 
     print("Basis\t\t...\t", end=' ')
 
@@ -126,60 +133,113 @@ def write_ezfio(trexio_filename, filename):
     try:
         basis_type = trexio.read_basis_type(trexio_file)
 
-        if basis_type.lower() not in ["gaussian", "slater"]:
+        if basis_type.lower() in ["gaussian", "slater"]:
-            raise TypeError
+            shell_num   = trexio.read_basis_shell_num(trexio_file)
+            prim_num    = trexio.read_basis_prim_num(trexio_file)
+            ang_mom     = trexio.read_basis_shell_ang_mom(trexio_file)
+            nucl_index  = trexio.read_basis_nucleus_index(trexio_file)
+            exponent    = trexio.read_basis_exponent(trexio_file)
+            coefficient = trexio.read_basis_coefficient(trexio_file)
+            shell_index = trexio.read_basis_shell_index(trexio_file)
+            ao_shell    = trexio.read_ao_shell(trexio_file)
 
-        shell_num   = trexio.read_basis_shell_num(trexio_file)
+            ezfio.set_basis_basis("Read from TREXIO")
-        prim_num    = trexio.read_basis_prim_num(trexio_file)
+            ezfio.set_ao_basis_ao_basis("Read from TREXIO")
-        ang_mom     = trexio.read_basis_shell_ang_mom(trexio_file)
+            ezfio.set_basis_shell_num(shell_num)
-        nucl_index  = trexio.read_basis_nucleus_index(trexio_file)
+            ezfio.set_basis_prim_num(prim_num)
-        exponent    = trexio.read_basis_exponent(trexio_file)
+            ezfio.set_basis_shell_ang_mom(ang_mom)
-        coefficient = trexio.read_basis_coefficient(trexio_file)
+            ezfio.set_basis_basis_nucleus_index([ x+1 for x in nucl_index ])
-        shell_index = trexio.read_basis_shell_index(trexio_file)
+            ezfio.set_basis_prim_expo(exponent)
-        ao_shell    = trexio.read_ao_shell(trexio_file)
+            ezfio.set_basis_prim_coef(coefficient)
 
-        ezfio.set_basis_basis("Read from TREXIO")
+            nucl_shell_num = []
-        ezfio.set_basis_shell_num(shell_num)
+            prev = None
-        ezfio.set_basis_prim_num(prim_num)
+            m = 0
-        ezfio.set_basis_shell_ang_mom(ang_mom)
+            for i in ao_shell:
-        ezfio.set_basis_basis_nucleus_index([ x+1 for x in nucl_index ])
+                if i != prev:
-        ezfio.set_basis_prim_expo(exponent)
+                   m += 1
-        ezfio.set_basis_prim_coef(coefficient)
+                   if prev is None or nucl_index[i] != nucl_index[prev]:
+                        nucl_shell_num.append(m)
+                        m = 0
+                prev = i
+            assert (len(nucl_shell_num) == nucl_num)
 
-        nucl_shell_num = []
+            shell_prim_num = []
-        prev = None
+            prev = shell_index[0]
-        m = 0
+            count = 0
-        for i in ao_shell:
+            for i in shell_index:
-            if i != prev:
+                if i != prev:
-               m += 1
+                   shell_prim_num.append(count)
-               if prev is None or nucl_index[i] != nucl_index[prev]:
+                   count = 0
-                    nucl_shell_num.append(m)
+                count += 1
-                    m = 0
+                prev = i
-            prev = i
+            shell_prim_num.append(count)
-        assert (len(nucl_shell_num) == nucl_num)
 
-        shell_prim_num = []
+            assert (len(shell_prim_num) == shell_num)
-        prev = shell_index[0]
-        count = 0
-        for i in shell_index:
-            if i != prev:
-               shell_prim_num.append(count)
-               count = 0
-            count += 1
-            prev = i
-        shell_prim_num.append(count)
 
-        assert (len(shell_prim_num) == shell_num)
+            ezfio.set_basis_shell_prim_num(shell_prim_num)
-
+            ezfio.set_basis_shell_index([x+1 for x in shell_index])
-        ezfio.set_basis_shell_prim_num(shell_prim_num)
+            ezfio.set_basis_nucleus_shell_num(nucl_shell_num)
-        ezfio.set_basis_shell_index([x+1 for x in shell_index])
-        ezfio.set_basis_nucleus_shell_num(nucl_shell_num)
 
 
-        shell_factor = trexio.read_basis_shell_factor(trexio_file)
+            shell_factor = trexio.read_basis_shell_factor(trexio_file)
-        prim_factor  = trexio.read_basis_prim_factor(trexio_file)
+            prim_factor  = trexio.read_basis_prim_factor(trexio_file)
 
-        print("OK")
+        elif basis_type.lower() == "numerical":
+
+            shell_num   = trexio.read_basis_shell_num(trexio_file)
+            prim_num    = shell_num
+            ang_mom     = trexio.read_basis_shell_ang_mom(trexio_file)
+            nucl_index  = trexio.read_basis_nucleus_index(trexio_file)
+            exponent    = [1.]*prim_num
+            coefficient = [1.]*prim_num
+            shell_index = [i for i in range(shell_num)]
+            ao_shell    = trexio.read_ao_shell(trexio_file)
+
+            ezfio.set_basis_basis("None")
+            ezfio.set_ao_basis_ao_basis("None")
+            ezfio.set_basis_shell_num(shell_num)
+            ezfio.set_basis_prim_num(prim_num)
+            ezfio.set_basis_shell_ang_mom(ang_mom)
+            ezfio.set_basis_basis_nucleus_index([ x+1 for x in nucl_index ])
+            ezfio.set_basis_prim_expo(exponent)
+            ezfio.set_basis_prim_coef(coefficient)
+
+            nucl_shell_num = []
+            prev = None
+            m = 0
+            for i in ao_shell:
+                if i != prev:
+                   m += 1
+                   if prev is None or nucl_index[i] != nucl_index[prev]:
+                        nucl_shell_num.append(m)
+                        m = 0
+                prev = i
+            assert (len(nucl_shell_num) == nucl_num)
+
+            shell_prim_num = []
+            prev = shell_index[0]
+            count = 0
+            for i in shell_index:
+                if i != prev:
+                   shell_prim_num.append(count)
+                   count = 0
+                count += 1
+                prev = i
+            shell_prim_num.append(count)
+
+            assert (len(shell_prim_num) == shell_num)
+
+            ezfio.set_basis_shell_prim_num(shell_prim_num)
+            ezfio.set_basis_shell_index([x+1 for x in shell_index])
+            ezfio.set_basis_nucleus_shell_num(nucl_shell_num)
+
+            shell_factor = trexio.read_basis_shell_factor(trexio_file)
+            prim_factor  = [1.]*prim_num
+        else:
+           raise TypeError
+
+        print(basis_type)
     except:
         print("None")
         ezfio.set_ao_basis_ao_cartesian(True)
@@ -256,9 +316,11 @@ def write_ezfio(trexio_filename, filename):
 #        ezfio.set_ao_basis_ao_prim_num_max(prim_num_max)
         ezfio.set_ao_basis_ao_coef(coef)
         ezfio.set_ao_basis_ao_expo(expo)
-        ezfio.set_ao_basis_ao_basis("Read from TREXIO")
 
-    print("OK")
+        print("OK")
+
+    else:
+        print("None")
 
 
     #                _
@@ -279,6 +341,7 @@ def write_ezfio(trexio_filename, filename):
     except:
       label = "None"
     ezfio.set_mo_basis_mo_label(label)
+    ezfio.set_determinants_mo_label(label)
 
     try:
       clss = trexio.read_mo_class(trexio_file)
@@ -303,10 +366,10 @@ def write_ezfio(trexio_filename, filename):
       for i in range(num_beta):
          mo_occ[i] += 1.
       ezfio.set_mo_basis_mo_occ(mo_occ)
+      print("OK")
     except:
-      pass
+      print("None")
 
-    print("OK")
 
 
     print("Pseudos\t\t...\t", end=' ')
@@ -386,8 +449,23 @@ def write_ezfio(trexio_filename, filename):
         ezfio.set_pseudo_pseudo_n_kl(pseudo_n_kl)
         ezfio.set_pseudo_pseudo_v_kl(pseudo_v_kl)
         ezfio.set_pseudo_pseudo_dz_kl(pseudo_dz_kl)
+        print("OK")
 
+    else:
+        print("None")
 
+    print("Determinant\t\t...\t", end=' ')
+    alpha = [ i for i in range(num_alpha) ]
+    beta  = [ i for i in range(num_beta) ]
+    if trexio.has_mo_spin(trexio_file):
+       spin = trexio.read_mo_spin(trexio_file)
+       beta  = [ i for i in range(mo_num) if spin[i] == 1 ]
+       beta  = [ beta[i] for i in range(num_beta) ]
+
+    alpha = qp_bitmasks.BitMask(alpha)
+    beta  = qp_bitmasks.BitMask(beta )
+    print(alpha)
+    print(beta)
     print("OK")
 
 

scripts/utility/qp_bitmasks.py

@@ -22,7 +22,7 @@ def int_to_string(s):
         assert s>=0
     AssertionError
     """
-    assert type(s) in (int, long)
+    assert type(s) == int
     assert s>=0
     return '{s:0b}'.format(s=s)
 
@@ -62,7 +62,7 @@ def int_to_bitmask(s,bit_kind_size=BIT_KIND_SIZE):
     ['1111111111111111111111111111111111111111111111111111111111110110']
     >>>
     """
-    assert type(s) in (int, long)
+    assert type(s) == int
     if s < 0:
         s = s + (1 << bit_kind_size)
     return ['{s:0{width}b}'.format(s=s,width=bit_kind_size)]
@@ -104,7 +104,7 @@ class BitMask(object):
     return self._data_int[i]
 
   def __setitem__(self,i,value):
-    if type(value) in (int,long):
+    if type(value) == int :
         self._data_int[i] = value
     elif type(value) == str:
         s = string_to_bitmask(value,bit_kind_size=self.bit_kind_size)[0]

src/ao_basis/EZFIO.cfg

@@ -67,3 +67,15 @@ doc: Use normalized primitive functions
 interface: ezfio, provider
 default: true
 
+[ao_expoim_cosgtos]
+type: double precision
+doc: imag part for Exponents for each primitive of each cosGTOs |AO|
+size: (ao_basis.ao_num,ao_basis.ao_prim_num_max)
+interface: ezfio, provider
+
+[use_cosgtos]
+type: logical
+doc: If true, use cosgtos for AO integrals
+interface: ezfio
+default: False
+

src/ao_basis/cosgtos.irp.f

@@ -0,0 +1,34 @@
+BEGIN_PROVIDER [ logical, use_cosgtos  ]
+  implicit none
+  BEGIN_DOC
+! If true, use cosgtos for AO integrals
+  END_DOC
+
+  logical                        :: has
+  PROVIDE ezfio_filename
+  use_cosgtos = .False.
+  if (mpi_master) then
+    call ezfio_has_ao_basis_use_cosgtos(has)
+    if (has) then
+!      write(6,'(A)') '.. >>>>> [ IO READ: use_cosgtos ] <<<<< ..'
+      call ezfio_get_ao_basis_use_cosgtos(use_cosgtos)
+    else
+      call ezfio_set_ao_basis_use_cosgtos(use_cosgtos)
+    endif
+  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( use_cosgtos, 1, MPI_LOGICAL, 0, MPI_COMM_WORLD, ierr)
+    if (ierr /= MPI_SUCCESS) then
+      stop 'Unable to read use_cosgtos with MPI'
+    endif
+  IRP_ENDIF
+
+!  call write_time(6)
+
+END_PROVIDER

src/ao_many_one_e_ints/listj1b.irp.f

@@ -62,6 +62,7 @@ END_PROVIDER
   double precision :: tmp_cent_x, tmp_cent_y, tmp_cent_z
 
   provide j1b_pen
+  provide j1b_pen_coef
 
   List_all_comb_b2_coef = 0.d0
   List_all_comb_b2_expo = 0.d0
@@ -127,8 +128,8 @@ END_PROVIDER
     List_all_comb_b2_expo(    1) = 0.d0
     List_all_comb_b2_cent(1:3,1) = 0.d0
     do i = 1, nucl_num
-      List_all_comb_b2_coef(  i+1) = -1.d0
+      List_all_comb_b2_coef(  i+1) = -1.d0 * j1b_pen_coef(i)
-      List_all_comb_b2_expo(  i+1) = j1b_pen(   i)
+      List_all_comb_b2_expo(  i+1) = j1b_pen(i)
       List_all_comb_b2_cent(1,i+1) = nucl_coord(i,1)
       List_all_comb_b2_cent(2,i+1) = nucl_coord(i,2)
       List_all_comb_b2_cent(3,i+1) = nucl_coord(i,3)
@@ -225,6 +226,7 @@ END_PROVIDER
   double precision :: dx, dy, dz, r2
 
   provide j1b_pen
+  provide j1b_pen_coef
 
   List_all_comb_b3_coef = 0.d0
   List_all_comb_b3_expo = 0.d0
@@ -296,8 +298,8 @@ END_PROVIDER
 
     do i = 1, nucl_num
       ii = ii + 1
-      List_all_comb_b3_coef(  ii) = -2.d0
+      List_all_comb_b3_coef(  ii) = -2.d0 * j1b_pen_coef(i)
-      List_all_comb_b3_expo(  ii) = j1b_pen(   i)
+      List_all_comb_b3_expo(  ii) = j1b_pen(i)
       List_all_comb_b3_cent(1,ii) = nucl_coord(i,1)
       List_all_comb_b3_cent(2,ii) = nucl_coord(i,2)
       List_all_comb_b3_cent(3,ii) = nucl_coord(i,3)
@@ -305,7 +307,7 @@ END_PROVIDER
 
     do i = 1, nucl_num
       ii = ii + 1
-      List_all_comb_b3_coef(  ii) = 1.d0
+      List_all_comb_b3_coef(  ii) = 1.d0 * j1b_pen_coef(i) * j1b_pen_coef(i)
       List_all_comb_b3_expo(  ii) = 2.d0 * j1b_pen(i)
       List_all_comb_b3_cent(1,ii) = nucl_coord(i,1)
       List_all_comb_b3_cent(2,ii) = nucl_coord(i,2)
@@ -337,7 +339,7 @@ END_PROVIDER
         
         ii = ii + 1
         ! x 2 to avoid doing integrals twice
-        List_all_comb_b3_coef(  ii) = 2.d0 * dexp(-tmp1*tmp2*tmp4*r2)
+        List_all_comb_b3_coef(  ii) = 2.d0 * dexp(-tmp1*tmp2*tmp4*r2) * j1b_pen_coef(i) * j1b_pen_coef(j)
         List_all_comb_b3_expo(  ii) = tmp3
         List_all_comb_b3_cent(1,ii) = tmp4 * (tmp1 * xi + tmp2 * xj)
         List_all_comb_b3_cent(2,ii) = tmp4 * (tmp1 * yi + tmp2 * yj)

src/ao_one_e_ints/NEED

@@ -1,3 +1,2 @@
 ao_basis
 pseudo
-cosgtos_ao_int

src/ao_one_e_ints/pot_ao_ints.irp.f

@@ -104,6 +104,9 @@ BEGIN_PROVIDER [ double precision, ao_integrals_n_e, (ao_num,ao_num)]
     IF(do_pseudo) THEN
        ao_integrals_n_e += ao_pseudo_integrals
     ENDIF
+    IF(point_charges) THEN
+       ao_integrals_n_e += ao_integrals_pt_chrg
+    ENDIF
 
   endif
 
@@ -455,10 +458,12 @@ recursive subroutine I_x1_pol_mult_one_e(a,c,R1x,R1xp,R2x,d,nd,n_pt_in)
     do ix=0,nx
       X(ix) *= dble(c)
     enddo
-    call multiply_poly(X,nx,R2x,2,d,nd)
+!    call multiply_poly(X,nx,R2x,2,d,nd)
+    call multiply_poly_c2(X,nx,R2x,d,nd)
     ny=0
     call I_x2_pol_mult_one_e(c,R1x,R1xp,R2x,Y,ny,n_pt_in)
-    call multiply_poly(Y,ny,R1x,2,d,nd)
+!    call multiply_poly(Y,ny,R1x,2,d,nd)
+    call multiply_poly_c2(Y,ny,R1x,d,nd)
   else
     do ix=0,n_pt_in
       X(ix) = 0.d0
@@ -469,7 +474,8 @@ recursive subroutine I_x1_pol_mult_one_e(a,c,R1x,R1xp,R2x,d,nd,n_pt_in)
     do ix=0,nx
       X(ix) *= dble(a-1)
     enddo
-    call multiply_poly(X,nx,R2x,2,d,nd)
+!    call multiply_poly(X,nx,R2x,2,d,nd)
+    call multiply_poly_c2(X,nx,R2x,d,nd)
 
     nx = nd
     do ix=0,n_pt_in
@@ -479,10 +485,12 @@ recursive subroutine I_x1_pol_mult_one_e(a,c,R1x,R1xp,R2x,d,nd,n_pt_in)
     do ix=0,nx
       X(ix) *= dble(c)
     enddo
-    call multiply_poly(X,nx,R2x,2,d,nd)
+!    call multiply_poly(X,nx,R2x,2,d,nd)
+    call multiply_poly_c2(X,nx,R2x,d,nd)
     ny=0
     call I_x1_pol_mult_one_e(a-1,c,R1x,R1xp,R2x,Y,ny,n_pt_in)
-    call multiply_poly(Y,ny,R1x,2,d,nd)
+!    call multiply_poly(Y,ny,R1x,2,d,nd)
+    call multiply_poly_c2(Y,ny,R1x,d,nd)
   endif
 end
 
@@ -519,7 +527,8 @@ recursive subroutine I_x2_pol_mult_one_e(c,R1x,R1xp,R2x,d,nd,dim)
     do ix=0,nx
       X(ix) *= dble(c-1)
     enddo
-    call multiply_poly(X,nx,R2x,2,d,nd)
+!    call multiply_poly(X,nx,R2x,2,d,nd)
+    call multiply_poly_c2(X,nx,R2x,d,nd)
     ny = 0
     do ix=0,dim
       Y(ix) = 0.d0
@@ -527,7 +536,8 @@ recursive subroutine I_x2_pol_mult_one_e(c,R1x,R1xp,R2x,d,nd,dim)
 
     call I_x1_pol_mult_one_e(0,c-1,R1x,R1xp,R2x,Y,ny,dim)
     if(ny.ge.0)then
-      call multiply_poly(Y,ny,R1xp,2,d,nd)
+!      call multiply_poly(Y,ny,R1xp,2,d,nd)
+      call multiply_poly_c2(Y,ny,R1xp,d,nd)
     endif
   endif
 end

src/ao_two_e_ints/EZFIO.cfg

@@ -4,6 +4,19 @@ doc: Read/Write |AO| integrals from/to disk [ Write | Read | None ]
 interface: ezfio,provider,ocaml
 default: None
 
+[ao_integrals_threshold]
+type: Threshold
+doc: If | (pq|rs) | < `ao_integrals_threshold` then (pq|rs) is zero
+interface: ezfio,provider,ocaml
+default: 1.e-15
+ezfio_name: threshold_ao
+
+[ao_cholesky_threshold]
+type: Threshold
+doc: If | (ii|jj) | < `ao_cholesky_threshold` then (ii|jj) is zero
+interface: ezfio,provider,ocaml
+default: 1.e-12
+
 [do_direct_integrals]
 type: logical
 doc: Compute integrals on the fly (very slow, only for debugging)

src/ao_two_e_ints/cholesky.irp.f

@@ -4,29 +4,7 @@ BEGIN_PROVIDER [ integer, cholesky_ao_num_guess ]
  ! Number of Cholesky vectors in AO basis
  END_DOC
 
- integer :: i,j,k,l
+ cholesky_ao_num_guess = ao_num*ao_num 
- double precision :: xnorm0, x, integral
- double precision, external :: ao_two_e_integral
-
- cholesky_ao_num_guess = 0
- xnorm0 = 0.d0
- x = 0.d0
- do j=1,ao_num
-   do i=1,ao_num
-     integral = ao_two_e_integral(i,i,j,j)
-     if (integral > ao_integrals_threshold) then
-       cholesky_ao_num_guess += 1
-     else
-       x += integral
-     endif
-   enddo
- enddo
- print *, 'Cholesky decomposition of AO integrals'
- print *, '--------------------------------------'
- print *, ''
- print *, 'Estimated Error: ', x
- print *, 'Guess size: ', cholesky_ao_num_guess, '(', 100.d0*dble(cholesky_ao_num_guess)/dble(ao_num*ao_num), ' %)'
-
 END_PROVIDER
 
  BEGIN_PROVIDER [ integer, cholesky_ao_num ]
@@ -39,7 +17,7 @@ END_PROVIDER
  END_DOC
 
  type(c_ptr) :: ptr
- integer :: fd, i,j,k,l, rank
+ integer :: fd, i,j,k,l,m,rank
  double precision, pointer :: ao_integrals(:,:,:,:)
  double precision, external :: ao_two_e_integral
 
@@ -49,28 +27,83 @@ END_PROVIDER
    8, fd, .False., ptr)
  call c_f_pointer(ptr, ao_integrals, (/ao_num, ao_num, ao_num, ao_num/))
 
- double precision :: integral
+ print*, 'Providing the AO integrals (Cholesky)'
+ call wall_time(wall_1)
+ call cpu_time(cpu_1)
+
+ ao_integrals = 0.d0
+
+ double precision :: integral, cpu_1, cpu_2, wall_1, wall_2
  logical, external :: ao_two_e_integral_zero
- !$OMP PARALLEL DO DEFAULT(SHARED) PRIVATE(i,j,k,l, integral) SCHEDULE(dynamic)
+  double precision, external :: get_ao_two_e_integral
- do l=1,ao_num
+
-  do j=1,l
+ if (read_ao_two_e_integrals) then
-   do k=1,ao_num
+   PROVIDE ao_two_e_integrals_in_map
-    do i=1,k
+
-     if (ao_two_e_integral_zero(i,j,k,l)) cycle
+   !$OMP PARALLEL DEFAULT(SHARED) PRIVATE(i,j,k,l, integral, wall_2)
-     integral = ao_two_e_integral(i,k,j,l)
+   do m=0,9
-     ao_integrals(i,k,j,l) = integral
+     do l=1+m,ao_num,10
-     ao_integrals(k,i,j,l) = integral
+       !$OMP DO SCHEDULE(dynamic)
-     ao_integrals(i,k,l,j) = integral
+       do j=1,ao_num
-     ao_integrals(k,i,l,j) = integral
+         do k=1,ao_num
-    enddo
+           do i=1,ao_num
+             if (ao_two_e_integral_zero(i,j,k,l)) cycle
+             integral = get_ao_two_e_integral(i,j,k,l, ao_integrals_map)
+             ao_integrals(i,k,j,l) = integral
+           enddo
+         enddo
+       enddo
+       !$OMP END DO NOWAIT
+     enddo
+     !$OMP MASTER
+     call wall_time(wall_2)
+     print '(I10,'' %  in'', 4X, F10.2, '' s.'')', (m+1) * 10, wall_2-wall_1
+     !$OMP END MASTER
    enddo
-  enddo
+   !$OMP END PARALLEL
- enddo
+
- !$OMP END PARALLEL DO
+ else
+
+   !$OMP PARALLEL DEFAULT(SHARED) PRIVATE(i,j,k,l, integral, wall_2)
+   do m=0,9
+     do l=1+m,ao_num,10
+       !$OMP DO SCHEDULE(dynamic)
+       do j=1,l
+         do k=1,ao_num
+           do i=1,min(k,j)
+             if (ao_two_e_integral_zero(i,j,k,l)) cycle
+             integral = ao_two_e_integral(i,k,j,l)
+             ao_integrals(i,k,j,l) = integral
+             ao_integrals(k,i,j,l) = integral
+             ao_integrals(i,k,l,j) = integral
+             ao_integrals(k,i,l,j) = integral
+             ao_integrals(j,l,i,k) = integral
+             ao_integrals(j,l,k,i) = integral
+             ao_integrals(l,j,i,k) = integral
+             ao_integrals(l,j,k,i) = integral
+           enddo
+         enddo
+       enddo
+       !$OMP END DO NOWAIT
+     enddo
+     !$OMP MASTER
+     call wall_time(wall_2)
+     print '(I10,'' %  in'', 4X, F10.2, '' s.'')', (m+1) * 10, wall_2-wall_1
+     !$OMP END MASTER
+   enddo
+   !$OMP END PARALLEL
+
+   call wall_time(wall_2)
+   call cpu_time(cpu_2)
+   print*, 'AO integrals provided:'
+   print*, ' cpu  time :',cpu_2 - cpu_1, 's'
+   print*, ' wall time :',wall_2 - wall_1, 's  ( x ', (cpu_2-cpu_1)/(wall_2-wall_1+tiny(1.d0)), ' )'
+
+ endif
 
  ! Call Lapack
  cholesky_ao_num = cholesky_ao_num_guess
- call pivoted_cholesky(ao_integrals, cholesky_ao_num, ao_integrals_threshold, ao_num*ao_num, cholesky_ao)
+ call pivoted_cholesky(ao_integrals, cholesky_ao_num, ao_cholesky_threshold, ao_num*ao_num, cholesky_ao)
  print *, 'Rank: ', cholesky_ao_num, '(', 100.d0*dble(cholesky_ao_num)/dble(ao_num*ao_num), ' %)'
 
  ! Remove mmap

src/ao_two_e_ints/two_e_Coul_integrals_cosgtos.irp.f

@@ -29,14 +29,14 @@ double precision function ao_two_e_integral_cosgtos(i, j, k, l)
   complex*16          :: integral5, integral6, integral7, integral8
   complex*16          :: integral_tot
 
-  double precision    :: ao_two_e_integral_cosgtos_schwartz_accel
+  double precision    :: ao_2e_cosgtos_schwartz_accel
   complex*16          :: ERI_cosgtos
   complex*16          :: general_primitive_integral_cosgtos
 
   if(ao_prim_num(i) * ao_prim_num(j) * ao_prim_num(k) * ao_prim_num(l) > 1024) then
 
     !print *, ' with shwartz acc '
-    ao_two_e_integral_cosgtos = ao_two_e_integral_cosgtos_schwartz_accel(i, j, k, l)
+    ao_two_e_integral_cosgtos = ao_2e_cosgtos_schwartz_accel(i, j, k, l)
 
   else
     !print *, ' without shwartz acc '
@@ -294,7 +294,7 @@ end function ao_two_e_integral_cosgtos
 
 ! ---
 
-double precision function ao_two_e_integral_cosgtos_schwartz_accel(i, j, k, l)
+double precision function ao_2e_cosgtos_schwartz_accel(i, j, k, l)
 
   BEGIN_DOC
   !  integral of the AO basis <ik|jl> or (ij|kl)
@@ -329,7 +329,7 @@ double precision function ao_two_e_integral_cosgtos_schwartz_accel(i, j, k, l)
   complex*16                    :: ERI_cosgtos
   complex*16                    :: general_primitive_integral_cosgtos
 
-  ao_two_e_integral_cosgtos_schwartz_accel = 0.d0
+  ao_2e_cosgtos_schwartz_accel = 0.d0
 
   dim1 = n_pt_max_integrals
 
@@ -519,8 +519,7 @@ double precision function ao_two_e_integral_cosgtos_schwartz_accel(i, j, k, l)
 
             integral_tot = integral1 + integral2 + integral3 + integral4 + integral5 + integral6 + integral7 + integral8
 
-            ao_two_e_integral_cosgtos_schwartz_accel = ao_two_e_integral_cosgtos_schwartz_accel &
+            ao_2e_cosgtos_schwartz_accel = ao_2e_cosgtos_schwartz_accel + coef4 * 2.d0 * real(integral_tot)
-                                                     + coef4 * 2.d0 * real(integral_tot)
           enddo ! s
         enddo  ! r
       enddo   ! q
@@ -698,8 +697,7 @@ double precision function ao_two_e_integral_cosgtos_schwartz_accel(i, j, k, l)
 
             integral_tot = integral1 + integral2 + integral3 + integral4 + integral5 + integral6 + integral7 + integral8
 
-            ao_two_e_integral_cosgtos_schwartz_accel = ao_two_e_integral_cosgtos_schwartz_accel &
+            ao_2e_cosgtos_schwartz_accel = ao_2e_cosgtos_schwartz_accel + coef4 * 2.d0 * real(integral_tot)
-                                      + coef4 * 2.d0 * real(integral_tot)
           enddo ! s
         enddo  ! r
       enddo   ! q
@@ -709,11 +707,11 @@ double precision function ao_two_e_integral_cosgtos_schwartz_accel(i, j, k, l)
 
   deallocate(schwartz_kl)
 
-end function ao_two_e_integral_cosgtos_schwartz_accel
+end function ao_2e_cosgtos_schwartz_accel
 
 ! ---
 
-BEGIN_PROVIDER [ double precision, ao_two_e_integral_cosgtos_schwartz, (ao_num,ao_num)  ]
+BEGIN_PROVIDER [ double precision, ao_2e_cosgtos_schwartz, (ao_num,ao_num)]
 
   BEGIN_DOC
   !  Needed to compute Schwartz inequalities
@@ -723,16 +721,16 @@ BEGIN_PROVIDER [ double precision, ao_two_e_integral_cosgtos_schwartz, (ao_num,a
   integer          :: i, k
   double precision :: ao_two_e_integral_cosgtos
 
-  ao_two_e_integral_cosgtos_schwartz(1,1) = ao_two_e_integral_cosgtos(1, 1, 1, 1)
+  ao_2e_cosgtos_schwartz(1,1) = ao_two_e_integral_cosgtos(1, 1, 1, 1)
 
- !$OMP PARALLEL DO PRIVATE(i,k)                                       &
+ !$OMP PARALLEL DO PRIVATE(i,k)                           &
- !$OMP             DEFAULT(NONE)                                      &
+ !$OMP             DEFAULT(NONE)                          &
- !$OMP             SHARED(ao_num, ao_two_e_integral_cosgtos_schwartz) &
+ !$OMP             SHARED(ao_num, ao_2e_cosgtos_schwartz) &
  !$OMP             SCHEDULE(dynamic)
   do i = 1, ao_num
     do k = 1, i
-      ao_two_e_integral_cosgtos_schwartz(i,k) = dsqrt(ao_two_e_integral_cosgtos(i, i, k, k))
+      ao_2e_cosgtos_schwartz(i,k) = dsqrt(ao_two_e_integral_cosgtos(i, i, k, k))
-      ao_two_e_integral_cosgtos_schwartz(k,i) = ao_two_e_integral_cosgtos_schwartz(i,k)
+      ao_2e_cosgtos_schwartz(k,i) = ao_2e_cosgtos_schwartz(i,k)
     enddo
   enddo
  !$OMP END PARALLEL DO

src/ao_two_e_ints/two_e_integrals.irp.f

@@ -590,8 +590,20 @@ double precision function general_primitive_integral(dim,            &
     d_poly(i)=0.d0
   enddo
 
-  !DIR$ FORCEINLINE
+!  call multiply_poly(Ix_pol,n_Ix,Iy_pol,n_Iy,d_poly,n_pt_tmp)
-  call multiply_poly(Ix_pol,n_Ix,Iy_pol,n_Iy,d_poly,n_pt_tmp)
+  integer :: ib, ic
+  if (ior(n_Ix,n_Iy) >= 0) then
+    do ib=0,n_Ix
+      do ic = 0,n_Iy
+        d_poly(ib+ic) = d_poly(ib+ic) + Iy_pol(ic) * Ix_pol(ib)
+      enddo
+    enddo
+
+    do n_pt_tmp = n_Ix+n_Iy, 0, -1
+      if (d_poly(n_pt_tmp) /= 0.d0) exit
+    enddo
+  endif
+
   if (n_pt_tmp == -1) then
     return
   endif
@@ -600,8 +612,21 @@ double precision function general_primitive_integral(dim,            &
     d1(i)=0.d0
   enddo
 
-  !DIR$ FORCEINLINE
+!  call multiply_poly(d_poly ,n_pt_tmp ,Iz_pol,n_Iz,d1,n_pt_out)
-  call multiply_poly(d_poly ,n_pt_tmp ,Iz_pol,n_Iz,d1,n_pt_out)
+  if (ior(n_pt_tmp,n_Iz) >= 0) then
+    ! Bottleneck here
+    do ib=0,n_pt_tmp
+      do ic = 0,n_Iz
+        d1(ib+ic) = d1(ib+ic) + Iz_pol(ic) * d_poly(ib)
+      enddo
+    enddo
+
+    do n_pt_out = n_pt_tmp+n_Iz, 0, -1
+      if (d1(n_pt_out) /= 0.d0) exit
+    enddo
+  endif
+
+
   double precision               :: rint_sum
   accu = accu + rint_sum(n_pt_out,const,d1)
 
@@ -948,8 +973,9 @@ recursive subroutine I_x1_pol_mult_recurs(a,c,B_10,B_01,B_00,C_00,D_00,d,nd,n_pt
     X(ix) *= dble(a-1)
   enddo
 
-  !DIR$ FORCEINLINE
+!  !DIR$ FORCEINLINE
-  call multiply_poly(X,nx,B_10,2,d,nd)
+!  call multiply_poly(X,nx,B_10,2,d,nd)
+  call multiply_poly_c2(X,nx,B_10,d,nd)
 
   nx = nd
   !DIR$ LOOP COUNT(8)
@@ -970,8 +996,9 @@ recursive subroutine I_x1_pol_mult_recurs(a,c,B_10,B_01,B_00,C_00,D_00,d,nd,n_pt
         X(ix) *= c
       enddo
     endif
-    !DIR$ FORCEINLINE
+!    !DIR$ FORCEINLINE
-    call multiply_poly(X,nx,B_00,2,d,nd)
+!    call multiply_poly(X,nx,B_00,2,d,nd)
+    call multiply_poly_c2(X,nx,B_00,d,nd)
   endif
 
   ny=0
@@ -988,9 +1015,9 @@ recursive subroutine I_x1_pol_mult_recurs(a,c,B_10,B_01,B_00,C_00,D_00,d,nd,n_pt
     call I_x1_pol_mult_recurs(a-1,c,B_10,B_01,B_00,C_00,D_00,Y,ny,n_pt_in)
   endif
 
-  !DIR$ FORCEINLINE
+!  !DIR$ FORCEINLINE
-  call multiply_poly(Y,ny,C_00,2,d,nd)
+!  call multiply_poly(Y,ny,C_00,2,d,nd)
-
+  call multiply_poly_c2(Y,ny,C_00,d,nd)
 end
 
 recursive subroutine I_x1_pol_mult_a1(c,B_10,B_01,B_00,C_00,D_00,d,nd,n_pt_in)
@@ -1028,8 +1055,9 @@ recursive subroutine I_x1_pol_mult_a1(c,B_10,B_01,B_00,C_00,D_00,d,nd,n_pt_in)
     enddo
   endif
 
-  !DIR$ FORCEINLINE
+!  !DIR$ FORCEINLINE
-  call multiply_poly(X,nx,B_00,2,d,nd)
+!  call multiply_poly(X,nx,B_00,2,d,nd)
+  call multiply_poly_c2(X,nx,B_00,d,nd)
 
   ny=0
 
@@ -1039,8 +1067,9 @@ recursive subroutine I_x1_pol_mult_a1(c,B_10,B_01,B_00,C_00,D_00,d,nd,n_pt_in)
   enddo
   call I_x2_pol_mult(c,B_10,B_01,B_00,C_00,D_00,Y,ny,n_pt_in)
 
-  !DIR$ FORCEINLINE
+!  !DIR$ FORCEINLINE
-  call multiply_poly(Y,ny,C_00,2,d,nd)
+!  call multiply_poly(Y,ny,C_00,2,d,nd)
+  call multiply_poly_c2(Y,ny,C_00,d,nd)
 
 end
 
@@ -1067,8 +1096,9 @@ recursive subroutine I_x1_pol_mult_a2(c,B_10,B_01,B_00,C_00,D_00,d,nd,n_pt_in)
   nx = 0
   call I_x2_pol_mult(c,B_10,B_01,B_00,C_00,D_00,X,nx,n_pt_in)
 
-  !DIR$ FORCEINLINE
+!  !DIR$ FORCEINLINE
-  call multiply_poly(X,nx,B_10,2,d,nd)
+!  call multiply_poly(X,nx,B_10,2,d,nd)
+  call multiply_poly_c2(X,nx,B_10,d,nd)
 
   nx = nd
   !DIR$ LOOP COUNT(8)
@@ -1086,8 +1116,9 @@ recursive subroutine I_x1_pol_mult_a2(c,B_10,B_01,B_00,C_00,D_00,d,nd,n_pt_in)
     enddo
   endif
 
-  !DIR$ FORCEINLINE
+!  !DIR$ FORCEINLINE
-  call multiply_poly(X,nx,B_00,2,d,nd)
+!  call multiply_poly(X,nx,B_00,2,d,nd)
+  call multiply_poly_c2(X,nx,B_00,d,nd)
 
   ny=0
   !DIR$ LOOP COUNT(8)
@@ -1097,9 +1128,9 @@ recursive subroutine I_x1_pol_mult_a2(c,B_10,B_01,B_00,C_00,D_00,d,nd,n_pt_in)
   !DIR$ FORCEINLINE
   call I_x1_pol_mult_a1(c,B_10,B_01,B_00,C_00,D_00,Y,ny,n_pt_in)
 
-  !DIR$ FORCEINLINE
+!  !DIR$ FORCEINLINE
-  call multiply_poly(Y,ny,C_00,2,d,nd)
+!  call multiply_poly(Y,ny,C_00,2,d,nd)
-
+  call multiply_poly_c2(Y,ny,C_00,d,nd)
 end
 
 recursive subroutine I_x2_pol_mult(c,B_10,B_01,B_00,C_00,D_00,d,nd,dim)
@@ -1146,8 +1177,10 @@ recursive subroutine I_x2_pol_mult(c,B_10,B_01,B_00,C_00,D_00,d,nd,dim)
       Y(1) = D_00(1)
       Y(2) = D_00(2)
 
-      !DIR$ FORCEINLINE
+!      !DIR$ FORCEINLINE
-      call multiply_poly(Y,ny,D_00,2,d,nd)
+!      call multiply_poly(Y,ny,D_00,2,d,nd)
+      call multiply_poly_c2(Y,ny,D_00,d,nd)
+
       return
 
       case default
@@ -1164,8 +1197,9 @@ recursive subroutine I_x2_pol_mult(c,B_10,B_01,B_00,C_00,D_00,d,nd,dim)
         X(ix) *= dble(c-1)
       enddo
 
-      !DIR$ FORCEINLINE
+!      !DIR$ FORCEINLINE
-      call multiply_poly(X,nx,B_01,2,d,nd)
+!      call multiply_poly(X,nx,B_01,2,d,nd)
+      call multiply_poly_c2(X,nx,B_01,d,nd)
 
       ny = 0
       !DIR$ LOOP COUNT(6)
@@ -1174,8 +1208,9 @@ recursive subroutine I_x2_pol_mult(c,B_10,B_01,B_00,C_00,D_00,d,nd,dim)
       enddo
       call I_x2_pol_mult(c-1,B_10,B_01,B_00,C_00,D_00,Y,ny,dim)
 
-      !DIR$ FORCEINLINE
+!      !DIR$ FORCEINLINE
-      call multiply_poly(Y,ny,D_00,2,d,nd)
+!      call multiply_poly(Y,ny,D_00,2,d,nd)
+      call multiply_poly_c2(Y,ny,D_00,d,nd)
 
   end select
 end
@@ -1233,3 +1268,34 @@ subroutine compute_ao_integrals_jl(j,l,n_integrals,buffer_i,buffer_value)
   enddo
 
 end
+
+
+subroutine multiply_poly_local(b,nb,c,nc,d,nd)
+  implicit none
+  BEGIN_DOC
+  ! Multiply two polynomials
+  ! D(t) += B(t)*C(t)
+  END_DOC
+
+  integer, intent(in)            :: nb, nc
+  integer, intent(out)           :: nd
+  double precision, intent(in)   :: b(0:nb), c(0:nc)
+  double precision, intent(inout) :: d(0:nb+nc)
+
+  integer                        :: ndtmp
+  integer                        :: ib, ic, id, k
+  if(ior(nc,nb) < 0) return !False if nc>=0 and nb>=0
+
+  do ib=0,nb
+    do ic = 0,nc
+      d(ib+ic) = d(ib+ic) + c(ic) * b(ib)
+    enddo
+  enddo
+
+  do nd = nb+nc,0,-1
+    if (d(nd) /= 0.d0) exit
+  enddo
+
+end
+
+

src/becke_numerical_grid/grid_becke_vector.irp.f

@@ -1,10 +1,13 @@
 
 BEGIN_PROVIDER [integer, n_points_final_grid]
-  implicit none
+
   BEGIN_DOC
   ! Number of points which are non zero
   END_DOC
-  integer                        :: i,j,k,l
+
+  implicit none
+  integer :: i, j, k, l
+
   n_points_final_grid = 0
   do j = 1, nucl_num
     do i = 1, n_points_radial_grid -1
@@ -16,9 +19,11 @@ BEGIN_PROVIDER [integer, n_points_final_grid]
       enddo
     enddo
   enddo
-  print*,'n_points_final_grid = ',n_points_final_grid
+
-  print*,'n max point         = ',n_points_integration_angular*(n_points_radial_grid*nucl_num - 1)
+  print*,' n_points_final_grid = ', n_points_final_grid
+  print*,' n max point         = ', n_points_integration_angular*(n_points_radial_grid*nucl_num - 1)
   call ezfio_set_becke_numerical_grid_n_points_final_grid(n_points_final_grid)
+
 END_PROVIDER
 
 ! ---
@@ -41,6 +46,10 @@ END_PROVIDER
   implicit none
   integer          :: i, j, k, l, i_count
   double precision :: r(3)
+  double precision :: wall0, wall1
+
+  call wall_time(wall0)
+  print *, ' Providing final_grid_points ...'
 
   i_count = 0
   do j = 1, nucl_num
@@ -62,20 +71,34 @@ END_PROVIDER
     enddo
   enddo
 
+  FREE grid_points_per_atom
+  FREE final_weight_at_r
+
+  call wall_time(wall1)
+  print *, ' wall time for final_grid_points,', wall1 - wall0
+  call print_memory_usage()
+
 END_PROVIDER
 
 ! ---
 
 BEGIN_PROVIDER [double precision, final_grid_points_transp, (n_points_final_grid,3)]
-  implicit none
+
   BEGIN_DOC
-! Transposed final_grid_points
+  ! Transposed final_grid_points
   END_DOC
 
+  implicit none
   integer :: i,j
-  do j=1,3
+
-    do i=1,n_points_final_grid
+  do j = 1, 3
+    do i = 1, n_points_final_grid
       final_grid_points_transp(i,j) = final_grid_points(j,i)
     enddo
   enddo
+
 END_PROVIDER
+
+! ---
+
+

src/bi_ort_ints/bi_ort_ints.irp.f

@@ -1,13 +1,37 @@
+! ---
+
 program bi_ort_ints
-  implicit none
+
   BEGIN_DOC
-! TODO : Put the documentation of the program here
+  ! TODO : Put the documentation of the program here
   END_DOC
+
+  implicit none
+
   my_grid_becke = .True.
-  my_n_pt_r_grid = 10
+  !my_n_pt_r_grid = 10
-  my_n_pt_a_grid = 14
+  !my_n_pt_a_grid = 14
-  touch  my_grid_becke my_n_pt_r_grid my_n_pt_a_grid
+  my_n_pt_r_grid = 30
- call test_3e
+  my_n_pt_a_grid = 50
+  touch my_grid_becke my_n_pt_r_grid my_n_pt_a_grid
+
+! call test_3e
+! call test_5idx
+! call test_5idx2
+  call test_4idx()
+  call test_4idx_n4()
+  !call test_4idx2()
+  !call test_5idx2
+  !call test_5idx
+end
+
+subroutine test_5idx2
+  PROVIDE three_e_5_idx_cycle_2_bi_ort
+end
+
+subroutine test_4idx2()
+  !PROVIDE three_e_4_idx_direct_bi_ort 
+  PROVIDE three_e_4_idx_exch23_bi_ort
 end
 
 subroutine test_3e
@@ -16,6 +40,7 @@ subroutine test_3e
  double precision :: accu, contrib,new,ref
  i = 1
  k = 1
+ n = 0
  accu = 0.d0
  do i = 1, mo_num
   do k = 1, mo_num
@@ -31,6 +56,7 @@ subroutine test_3e
          print*,'pb !!'
          print*,i,k,j,l,m,n
          print*,ref,new,contrib
+         stop
         endif
       enddo
      enddo
@@ -42,3 +68,394 @@ subroutine test_3e
 
 
 end
+
+subroutine test_5idx
+ implicit none
+ integer :: i,k,j,l,m,n,ipoint
+ double precision :: accu, contrib,new,ref
+ double precision, external :: three_e_5_idx_exch12_bi_ort
+ i = 1
+ k = 1
+ n = 0
+ accu = 0.d0
+ PROVIDE three_e_5_idx_direct_bi_ort_old
+
+ do i = 1, mo_num
+  do k = 1, mo_num
+   do j = 1, mo_num
+    do l = 1, mo_num
+     do m = 1, mo_num
+!      if (dabs(three_e_5_idx_direct_bi_ort(m,l,j,k,i) - three_e_5_idx_exch12_bi_ort(m,l,i,k,j)) > 1.d-10) then 
+!         stop
+!      endif
+      new = three_e_5_idx_direct_bi_ort(m,l,j,k,i)
+      ref = three_e_5_idx_direct_bi_ort_old(m,l,j,k,i)
+      contrib = dabs(new - ref)
+      accu += contrib
+      if(contrib .gt. 1.d-10)then
+       print*,'direct'
+       print*,i,k,j,l,m
+       print*,ref,new,contrib
+       stop
+      endif
+!
+!      new = three_e_5_idx_exch12_bi_ort(m,l,j,k,i)
+!      ref = three_e_5_idx_exch12_bi_ort_old(m,l,j,k,i)
+!      contrib = dabs(new - ref)
+!      accu += contrib
+!      if(contrib .gt. 1.d-10)then
+!       print*,'exch12'
+!       print*,i,k,j,l,m
+!       print*,ref,new,contrib
+!       stop
+!      endif
+!
+!
+!      new = three_e_5_idx_cycle_1_bi_ort(m,l,j,k,i)
+!      ref = three_e_5_idx_cycle_1_bi_ort_old(m,l,j,k,i)
+!      contrib = dabs(new - ref)
+!      accu += contrib
+!      if(contrib .gt. 1.d-10)then
+!       print*,'cycle1'
+
+!       print*,i,k,j,l,m
+!       print*,ref,new,contrib
+!       stop
+!      endif
+!
+!      new = three_e_5_idx_cycle_2_bi_ort(m,l,j,k,i)
+!      ref = three_e_5_idx_cycle_2_bi_ort_old(m,l,j,k,i)
+!      contrib = dabs(new - ref)
+!      accu += contrib
+!      if(contrib .gt. 1.d-10)then
+!       print*,'cycle2'
+!       print*,i,k,j,l,m
+!       print*,ref,new,contrib
+!       stop
+!      endif
+!
+!      new = three_e_5_idx_exch23_bi_ort(m,l,j,k,i)
+!      ref = three_e_5_idx_exch23_bi_ort_old(m,l,j,k,i)
+!      contrib = dabs(new - ref)
+!      accu += contrib
+!      if(contrib .gt. 1.d-10)then
+!       print*,'exch23'
+!       print*,i,k,j,l,m
+!       print*,ref,new,contrib
+!       stop
+!      endif
+!
+!      new = three_e_5_idx_exch13_bi_ort(m,l,j,k,i)
+!      ref = three_e_5_idx_exch13_bi_ort_old(m,l,j,k,i)
+!      contrib = dabs(new - ref)
+!      accu += contrib
+!      if(contrib .gt. 1.d-10)then
+!       print*,'exch13'
+!       print*,i,k,j,l,m
+!       print*,ref,new,contrib
+!       stop
+!      endif
+!
+!      new = three_e_5_idx_cycle_1_bi_ort(m,l,j,k,i)
+!      ref = three_e_5_idx_cycle_1_bi_ort_old(m,l,j,k,i)
+!      contrib = dabs(new - ref)
+!      accu += contrib
+!      if(contrib .gt. 1.d-10)then
+!       print*,'cycle1'
+!       print*,i,k,j,l,m
+!       print*,ref,new,contrib
+!       stop
+!      endif
+!
+!      new = three_e_5_idx_cycle_2_bi_ort(m,l,j,k,i)
+!      ref = three_e_5_idx_cycle_2_bi_ort_old(m,l,j,k,i)
+!      contrib = dabs(new - ref)
+!      accu += contrib
+!      if(contrib .gt. 1.d-10)then
+!       print*,'cycle2'
+!       print*,i,k,j,l,m
+!       print*,ref,new,contrib
+!       stop
+!      endif
+!
+!      new = three_e_5_idx_exch23_bi_ort(m,l,j,k,i)
+!      ref = three_e_5_idx_exch23_bi_ort_old(m,l,j,k,i)
+!      contrib = dabs(new - ref)
+!      accu += contrib
+!      if(contrib .gt. 1.d-10)then
+!       print*,'exch23'
+!       print*,i,k,j,l,m
+!       print*,ref,new,contrib
+!       stop
+!      endif
+!
+!      new = three_e_5_idx_exch13_bi_ort(m,l,j,k,i)
+!      ref = three_e_5_idx_exch13_bi_ort_old(m,l,j,k,i)
+!      contrib = dabs(new - ref)
+!      accu += contrib
+!      if(contrib .gt. 1.d-10)then
+!       print*,'exch13'
+!       print*,i,k,j,l,m
+!       print*,ref,new,contrib
+!       stop
+!      endif
+!
+     enddo
+    enddo
+   enddo
+  enddo
+ enddo
+ print*,'accu = ',accu/dble(mo_num)**5
+
+
+end
+
+! ---
+
+subroutine test_4idx_n4()
+
+  implicit none
+  integer          :: i, j, k, l
+  double precision :: accu, contrib, new, ref, thr
+
+  thr = 1d-10
+
+  PROVIDE three_e_4_idx_direct_bi_ort_old
+  PROVIDE three_e_4_idx_direct_bi_ort_n4
+
+  accu = 0.d0
+  do i = 1, mo_num
+    do j = 1, mo_num
+      do k = 1, mo_num
+        do l = 1, mo_num
+
+          new = three_e_4_idx_direct_bi_ort_n4 (l,k,j,i)
+          ref = three_e_4_idx_direct_bi_ort_old(l,k,j,i)
+          contrib = dabs(new - ref)
+          accu += contrib
+          if(contrib .gt. thr) then
+            print*, ' problem in three_e_4_idx_direct_bi_ort_n4'
+            print*, l, k, j, i
+            print*, ref, new, contrib
+            stop
+          endif
+
+        enddo
+      enddo
+    enddo
+  enddo
+  print*, ' accu on three_e_4_idx_direct_bi_ort_n4 = ', accu / dble(mo_num)**4
+
+  ! ---
+
+  PROVIDE three_e_4_idx_exch13_bi_ort_old
+  PROVIDE three_e_4_idx_exch13_bi_ort_n4
+
+  accu = 0.d0
+  do i = 1, mo_num
+    do j = 1, mo_num
+      do k = 1, mo_num
+        do l = 1, mo_num
+
+          new = three_e_4_idx_exch13_bi_ort_n4 (l,k,j,i)
+          ref = three_e_4_idx_exch13_bi_ort_old(l,k,j,i)
+          contrib = dabs(new - ref)
+          accu += contrib
+          if(contrib .gt. thr) then
+            print*, ' problem in three_e_4_idx_exch13_bi_ort_n4'
+            print*, l, k, j, i
+            print*, ref, new, contrib
+            stop
+          endif
+
+        enddo
+      enddo
+    enddo
+  enddo
+  print*, ' accu on three_e_4_idx_exch13_bi_ort_n4 = ', accu / dble(mo_num)**4
+
+  ! ---
+
+  PROVIDE three_e_4_idx_cycle_1_bi_ort_old
+  PROVIDE three_e_4_idx_cycle_1_bi_ort_n4
+
+  accu = 0.d0
+  do i = 1, mo_num
+    do j = 1, mo_num
+      do k = 1, mo_num
+        do l = 1, mo_num
+
+          new = three_e_4_idx_cycle_1_bi_ort_n4 (l,k,j,i)
+          ref = three_e_4_idx_cycle_1_bi_ort_old(l,k,j,i)
+          contrib = dabs(new - ref)
+          accu += contrib
+          if(contrib .gt. thr) then
+            print*, ' problem in three_e_4_idx_cycle_1_bi_ort_n4'
+            print*, l, k, j, i
+            print*, ref, new, contrib
+            stop
+          endif
+
+        enddo
+      enddo
+    enddo
+  enddo
+  print*, ' accu on three_e_4_idx_cycle_1_bi_ort_n4 = ', accu / dble(mo_num)**4
+
+  ! ---
+
+  PROVIDE three_e_4_idx_exch23_bi_ort_old
+  PROVIDE three_e_4_idx_exch23_bi_ort_n4
+
+  accu = 0.d0
+  do i = 1, mo_num
+    do j = 1, mo_num
+      do k = 1, mo_num
+        do l = 1, mo_num
+
+          new = three_e_4_idx_exch23_bi_ort_n4  (l,k,j,i)
+          ref = three_e_4_idx_exch23_bi_ort_old(l,k,j,i)
+          contrib = dabs(new - ref)
+          accu += contrib
+          if(contrib .gt. thr) then
+            print*, ' problem in three_e_4_idx_exch23_bi_ort_n4'
+            print*, l, k, j, i
+            print*, ref, new, contrib
+            stop
+          endif
+
+        enddo
+      enddo
+    enddo
+  enddo
+  print*, ' accu on three_e_4_idx_exch23_bi_ort_n4 = ', accu / dble(mo_num)**4
+
+  ! ---
+
+  return
+end
+
+! ---
+
+subroutine test_4idx()
+
+  implicit none
+  integer          :: i, j, k, l
+  double precision :: accu, contrib, new, ref, thr
+
+  thr = 1d-10
+
+  PROVIDE three_e_4_idx_direct_bi_ort_old
+  PROVIDE three_e_4_idx_direct_bi_ort 
+
+  accu = 0.d0
+  do i = 1, mo_num
+    do j = 1, mo_num
+      do k = 1, mo_num
+        do l = 1, mo_num
+
+          new = three_e_4_idx_direct_bi_ort    (l,k,j,i)
+          ref = three_e_4_idx_direct_bi_ort_old(l,k,j,i)
+          contrib = dabs(new - ref)
+          accu += contrib
+          if(contrib .gt. thr) then
+            print*, ' problem in three_e_4_idx_direct_bi_ort'
+            print*, l, k, j, i
+            print*, ref, new, contrib
+            stop
+          endif
+
+        enddo
+      enddo
+    enddo
+  enddo
+  print*, ' accu on three_e_4_idx_direct_bi_ort = ', accu / dble(mo_num)**4
+
+  ! ---
+
+  PROVIDE three_e_4_idx_exch13_bi_ort_old
+  PROVIDE three_e_4_idx_exch13_bi_ort 
+
+  accu = 0.d0
+  do i = 1, mo_num
+    do j = 1, mo_num
+      do k = 1, mo_num
+        do l = 1, mo_num
+
+          new = three_e_4_idx_exch13_bi_ort   (l,k,j,i)
+          ref = three_e_4_idx_exch13_bi_ort_old(l,k,j,i)
+          contrib = dabs(new - ref)
+          accu += contrib
+          if(contrib .gt. thr) then
+            print*, ' problem in three_e_4_idx_exch13_bi_ort'
+            print*, l, k, j, i
+            print*, ref, new, contrib
+            stop
+          endif
+
+        enddo
+      enddo
+    enddo
+  enddo
+  print*, ' accu on three_e_4_idx_exch13_bi_ort = ', accu / dble(mo_num)**4
+
+  ! ---
+
+  PROVIDE three_e_4_idx_cycle_1_bi_ort_old
+  PROVIDE three_e_4_idx_cycle_1_bi_ort
+
+  accu = 0.d0
+  do i = 1, mo_num
+    do j = 1, mo_num
+      do k = 1, mo_num
+        do l = 1, mo_num
+
+          new = three_e_4_idx_cycle_1_bi_ort    (l,k,j,i)
+          ref = three_e_4_idx_cycle_1_bi_ort_old(l,k,j,i)
+          contrib = dabs(new - ref)
+          accu += contrib
+          if(contrib .gt. thr) then
+            print*, ' problem in three_e_4_idx_cycle_1_bi_ort'
+            print*, l, k, j, i
+            print*, ref, new, contrib
+            stop
+          endif
+
+        enddo
+      enddo
+    enddo
+  enddo
+  print*, ' accu on three_e_4_idx_cycle_1_bi_ort = ', accu / dble(mo_num)**4
+
+  ! ---
+
+  PROVIDE three_e_4_idx_exch23_bi_ort_old
+  PROVIDE three_e_4_idx_exch23_bi_ort
+
+  accu = 0.d0
+  do i = 1, mo_num
+    do j = 1, mo_num
+      do k = 1, mo_num
+        do l = 1, mo_num
+
+          new = three_e_4_idx_exch23_bi_ort    (l,k,j,i)
+          ref = three_e_4_idx_exch23_bi_ort_old(l,k,j,i)
+          contrib = dabs(new - ref)
+          accu += contrib
+          if(contrib .gt. thr) then
+            print*, ' problem in three_e_4_idx_exch23_bi_ort'
+            print*, l, k, j, i
+            print*, ref, new, contrib
+            stop
+          endif
+
+        enddo
+      enddo
+    enddo
+  enddo
+  print*, ' accu on three_e_4_idx_exch23_bi_ort = ', accu / dble(mo_num)**4
+
+  ! ---
+
+  return
+end

src/bi_ort_ints/semi_num_ints_mo.irp.f

@@ -54,7 +54,7 @@ BEGIN_PROVIDER [ double precision, mo_v_ki_bi_ortho_erf_rk_cst_mu_transp, (n_poi
     enddo
   enddo
 
-! FREE mo_v_ki_bi_ortho_erf_rk_cst_mu
+  !FREE mo_v_ki_bi_ortho_erf_rk_cst_mu
 
 END_PROVIDER 
 
@@ -124,6 +124,8 @@ BEGIN_PROVIDER [ double precision, int2_grad1_u12_ao_transp, (ao_num, ao_num, 3,
       enddo
     enddo
 
+    FREE int2_grad1_u12_ao_test
+
   else
 
     PROVIDE int2_grad1_u12_ao
@@ -138,10 +140,13 @@ BEGIN_PROVIDER [ double precision, int2_grad1_u12_ao_transp, (ao_num, ao_num, 3,
       enddo
     enddo
 
+    FREE int2_grad1_u12_ao
+
   endif
 
   call wall_time(wall1)
   print *, ' wall time for int2_grad1_u12_ao_transp ', wall1 - wall0
+  call print_memory_usage()
 
 END_PROVIDER 
 
@@ -150,7 +155,7 @@ END_PROVIDER
 BEGIN_PROVIDER [ double precision, int2_grad1_u12_bimo_transp, (mo_num, mo_num, 3, n_points_final_grid)]
 
   implicit none
-  integer :: ipoint
+  integer          :: ipoint
   double precision :: wall0, wall1
 
   PROVIDE mo_l_coef mo_r_coef
@@ -177,6 +182,7 @@ BEGIN_PROVIDER [ double precision, int2_grad1_u12_bimo_transp, (mo_num, mo_num,
 
   !call wall_time(wall1)
   !print *, ' Wall time for providing int2_grad1_u12_bimo_transp',wall1 - wall0
+  !call print_memory_usage()
 
 END_PROVIDER 
 
@@ -185,7 +191,11 @@ END_PROVIDER
 BEGIN_PROVIDER [ double precision, int2_grad1_u12_bimo_t, (n_points_final_grid, 3, mo_num, mo_num)]
 
   implicit none
-  integer :: i, j, ipoint
+  integer          :: i, j, ipoint
+  double precision :: wall0, wall1
+
+  !call wall_time(wall0)
+  !print *, ' Providing int2_grad1_u12_bimo_t ...'
 
   PROVIDE mo_l_coef mo_r_coef
   PROVIDE int2_grad1_u12_bimo_transp
@@ -200,6 +210,12 @@ BEGIN_PROVIDER [ double precision, int2_grad1_u12_bimo_t, (n_points_final_grid,
     enddo
   enddo
 
+  FREE int2_grad1_u12_bimo_transp
+
+  !call wall_time(wall1)
+  !print *, ' wall time for int2_grad1_u12_bimo_t,', wall1 - wall0
+  !call print_memory_usage()
+
 END_PROVIDER 
 
 ! ---

src/bi_ort_ints/three_body_ijm.irp.f

@@ -23,11 +23,11 @@ BEGIN_PROVIDER [ double precision, three_e_3_idx_direct_bi_ort, (mo_num, mo_num,
 
   provide mos_r_in_r_array_transp mos_l_in_r_array_transp
 
- !$OMP PARALLEL                 &
+  !$OMP PARALLEL                 &
- !$OMP DEFAULT (NONE)           &
+  !$OMP DEFAULT (NONE)           &
- !$OMP PRIVATE (i,j,m,integral) & 
+  !$OMP PRIVATE (i,j,m,integral) & 
- !$OMP SHARED (mo_num,three_e_3_idx_direct_bi_ort)
+  !$OMP SHARED (mo_num,three_e_3_idx_direct_bi_ort)
- !$OMP DO SCHEDULE (dynamic)
+  !$OMP DO SCHEDULE (dynamic)
   do i = 1, mo_num
     do j = 1, mo_num
       do m = j, mo_num
@@ -36,8 +36,8 @@ BEGIN_PROVIDER [ double precision, three_e_3_idx_direct_bi_ort, (mo_num, mo_num,
       enddo
     enddo
   enddo
- !$OMP END DO
+  !$OMP END DO
- !$OMP END PARALLEL
+  !$OMP END PARALLEL
 
   do i = 1, mo_num
     do j = 1, mo_num
@@ -49,6 +49,7 @@ BEGIN_PROVIDER [ double precision, three_e_3_idx_direct_bi_ort, (mo_num, mo_num,
 
   call wall_time(wall1)
   print *, ' wall time for three_e_3_idx_direct_bi_ort', wall1 - wall0
+  call print_memory_usage()
 
 END_PROVIDER 
 
@@ -102,6 +103,7 @@ BEGIN_PROVIDER [ double precision, three_e_3_idx_cycle_1_bi_ort, (mo_num, mo_num
 
   call wall_time(wall1)
   print *, ' wall time for three_e_3_idx_cycle_1_bi_ort', wall1 - wall0
+  call print_memory_usage()
 
 END_PROVIDER 
 
@@ -155,6 +157,7 @@ BEGIN_PROVIDER [ double precision, three_e_3_idx_cycle_2_bi_ort, (mo_num, mo_num
 
   call wall_time(wall1)
   print *, ' wall time for three_e_3_idx_cycle_2_bi_ort', wall1 - wall0
+  call print_memory_usage()
 
 END_PROVIDER 
 
@@ -208,6 +211,7 @@ BEGIN_PROVIDER [ double precision, three_e_3_idx_exch23_bi_ort, (mo_num, mo_num,
 
   call wall_time(wall1)
   print *, ' wall time for three_e_3_idx_exch23_bi_ort', wall1 - wall0
+  call print_memory_usage()
 
 END_PROVIDER 
 
@@ -261,6 +265,7 @@ BEGIN_PROVIDER [ double precision, three_e_3_idx_exch13_bi_ort, (mo_num, mo_num,
 
   call wall_time(wall1)
   print *, ' wall time for three_e_3_idx_exch13_bi_ort', wall1 - wall0
+  call print_memory_usage()
 
 END_PROVIDER 
 
@@ -306,6 +311,7 @@ BEGIN_PROVIDER [ double precision, three_e_3_idx_exch12_bi_ort, (mo_num, mo_num,
 
   call wall_time(wall1)
   print *, ' wall time for three_e_3_idx_exch12_bi_ort', wall1 - wall0
+  call print_memory_usage()
 
 END_PROVIDER 
 
@@ -359,6 +365,7 @@ BEGIN_PROVIDER [ double precision, three_e_3_idx_exch12_bi_ort_new, (mo_num, mo_
 
   call wall_time(wall1)
   print *, ' wall time for three_e_3_idx_exch12_bi_ort_new', wall1 - wall0
+  call print_memory_usage()
 
 END_PROVIDER 
 

src/bi_ort_ints/three_body_ijmk.irp.f

@@ -1,282 +1,230 @@
 
 ! ---
 
-BEGIN_PROVIDER [ double precision, three_e_4_idx_direct_bi_ort, (mo_num, mo_num, mo_num, mo_num)]
+ BEGIN_PROVIDER [ double precision, three_e_4_idx_direct_bi_ort , (mo_num, mo_num, mo_num, mo_num)]
+&BEGIN_PROVIDER [ double precision, three_e_4_idx_exch13_bi_ort , (mo_num, mo_num, mo_num, mo_num)]
+&BEGIN_PROVIDER [ double precision, three_e_4_idx_exch23_bi_ort , (mo_num, mo_num, mo_num, mo_num)]
+&BEGIN_PROVIDER [ double precision, three_e_4_idx_cycle_1_bi_ort, (mo_num, mo_num, mo_num, mo_num)]
 
   BEGIN_DOC
   !
   ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
   !
-  ! three_e_4_idx_direct_bi_ort(m,j,k,i) = <mjk|-L|mji> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  ! three_e_4_idx_direct_bi_ort (m,j,k,i) = < m j k | -L | m j i > ::: notice that i is the RIGHT MO and k is the LEFT MO
+  ! three_e_4_idx_exch13_bi_ort (m,j,k,i) = < m j k | -L | i j m > ::: notice that i is the RIGHT MO and k is the LEFT MO
+  ! three_e_4_idx_exch23_bi_ort (m,j,k,i) = < m j k | -L | j m i > ::: notice that i is the RIGHT MO and k is the LEFT MO
+  ! three_e_4_idx_cycle_1_bi_ort(m,j,k,i) = < m j k | -L | j i m > ::: notice that i is the RIGHT MO and k is the LEFT MO
   !
-  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  ! notice the -1 sign: in this way three_e_4_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
   !
-  END_DOC
+  ! three_e_4_idx_direct_bi_ort (m,j,k,i) : Lk Ri Imm Ijj + Lj Rj Imm Iki + Lm Rm Ijj Iki 
-
+  ! three_e_4_idx_exch13_bi_ort (m,j,k,i) : Lk Rm Imi Ijj + Lj Rj Imi Ikm + Lm Ri Ijj Ikm 
- implicit none
+  ! three_e_4_idx_exch23_bi_ort (m,j,k,i) : Lk Ri Imj Ijm + Lj Rm Imj Iki + Lm Rj Ijm Iki
- integer          :: i, j, k, m
+  ! three_e_4_idx_cycle_1_bi_ort(m,j,k,i) : Lk Rm Imj Iji + Lj Ri Imj Ikm + Lm Rj Iji Ikm 
- double precision :: integral, wall1, wall0
-
-  three_e_4_idx_direct_bi_ort = 0.d0
-  print *, ' Providing the three_e_4_idx_direct_bi_ort ...'
-  call wall_time(wall0)
-
-  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
-
- !$OMP PARALLEL                   &
- !$OMP DEFAULT (NONE)             &
- !$OMP PRIVATE (i,j,k,m,integral) & 
- !$OMP SHARED (mo_num,three_e_4_idx_direct_bi_ort)
- !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
-  do i = 1, mo_num
-    do k = 1, mo_num
-      do j = 1, mo_num
-        do m = 1, mo_num
-          call give_integrals_3_body_bi_ort(m, j, k, m, j, i, integral)
-          three_e_4_idx_direct_bi_ort(m,j,k,i) = -1.d0 * integral 
-        enddo
-      enddo
-    enddo
-  enddo
- !$OMP END DO
- !$OMP END PARALLEL
-
-  call wall_time(wall1)
-  print *, ' wall time for three_e_4_idx_direct_bi_ort', wall1 - wall0
-
-END_PROVIDER 
-
-! ---
-
-BEGIN_PROVIDER [ double precision, three_e_4_idx_cycle_1_bi_ort, (mo_num, mo_num, mo_num, mo_num)]
-
-  BEGIN_DOC
-  !
-  ! matrix element of the -L  three-body operator FOR THE FIRST CYCLIC PERMUTATION TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
-  !
-  ! three_e_4_idx_cycle_1_bi_ort(m,j,k,i) = <mjk|-L|jim> ::: notice that i is the RIGHT MO and k is the LEFT MO
-  !
-  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
   !
   END_DOC
 
   implicit none
-  integer          :: i, j, k, m
+  integer                       :: ipoint, i, j, k, m, n
-  double precision :: integral, wall1, wall0
+  double precision              :: wall1, wall0
+  double precision              :: tmp_loc_1, tmp_loc_2
+  double precision, allocatable :: tmp1(:,:,:), tmp2(:,:,:)
+  double precision, allocatable :: tmp_2d(:,:)
+  double precision, allocatable :: tmp_aux_1(:,:,:), tmp_aux_2(:,:)
 
-  three_e_4_idx_cycle_1_bi_ort = 0.d0
+  print *, ' Providing the three_e_4_idx_bi_ort ...'
-  print *, ' Providing the three_e_4_idx_cycle_1_bi_ort ...'
   call wall_time(wall0)
 
   provide mos_r_in_r_array_transp mos_l_in_r_array_transp
 
- !$OMP PARALLEL                   &
- !$OMP DEFAULT (NONE)             &
- !$OMP PRIVATE (i,j,k,m,integral) & 
- !$OMP SHARED (mo_num,three_e_4_idx_cycle_1_bi_ort)
- !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
-  do i = 1, mo_num
-    do k = 1, mo_num
-      do j = 1, mo_num
-        do m = 1, mo_num
-          call give_integrals_3_body_bi_ort(m, j, k, j, i, m, integral)
-          three_e_4_idx_cycle_1_bi_ort(m,j,k,i) = -1.d0 * integral 
-        enddo
-      enddo
-    enddo
-  enddo
- !$OMP END DO
- !$OMP END PARALLEL
 
-  call wall_time(wall1)
+  ! to reduce the number of operations
-  print *, ' wall time for three_e_4_idx_cycle_1_bi_ort', wall1 - wall0
+  allocate(tmp_aux_1(n_points_final_grid,4,mo_num))
+  allocate(tmp_aux_2(n_points_final_grid,mo_num))
 
-END_PROVIDER 
+  !$OMP PARALLEL                                                  &
+  !$OMP DEFAULT (NONE)                                            &
+  !$OMP PRIVATE (n, ipoint)                                       &
+  !$OMP SHARED (mo_num, n_points_final_grid,                      &
+  !$OMP         mos_l_in_r_array_transp, mos_r_in_r_array_transp, &
+  !$OMP         int2_grad1_u12_bimo_t, final_weight_at_r_vector,  &
+  !$OMP         tmp_aux_1, tmp_aux_2)
+  !$OMP DO
+  do n = 1, mo_num
+    do ipoint = 1, n_points_final_grid
 
-! --
+        tmp_aux_1(ipoint,1,n) = int2_grad1_u12_bimo_t(ipoint,1,n,n) * final_weight_at_r_vector(ipoint)
+        tmp_aux_1(ipoint,2,n) = int2_grad1_u12_bimo_t(ipoint,2,n,n) * final_weight_at_r_vector(ipoint)
+        tmp_aux_1(ipoint,3,n) = int2_grad1_u12_bimo_t(ipoint,3,n,n) * final_weight_at_r_vector(ipoint)
+        tmp_aux_1(ipoint,4,n) = mos_l_in_r_array_transp(ipoint,n) * mos_r_in_r_array_transp(ipoint,n) * final_weight_at_r_vector(ipoint)
 
-BEGIN_PROVIDER [ double precision, three_e_4_idx_cycle_2_bi_ort, (mo_num, mo_num, mo_num, mo_num)]
+        tmp_aux_2(ipoint,n) = mos_l_in_r_array_transp(ipoint,n) * mos_r_in_r_array_transp(ipoint,n)
-
-  BEGIN_DOC
-  !
-  ! matrix element of the -L  three-body operator FOR THE FIRST CYCLIC PERMUTATION TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
-  !
-  ! three_e_4_idx_cycle_2_bi_ort(m,j,k,i) = <mjk|-L|imj> ::: notice that i is the RIGHT MO and k is the LEFT MO
-  !
-  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
-  !
-  END_DOC
-
-  implicit none
-  integer          :: i, j, k, m
-  double precision :: integral, wall1, wall0
-
-  three_e_4_idx_cycle_2_bi_ort = 0.d0
-  print *, ' Providing the three_e_4_idx_cycle_2_bi_ort ...'
-  call wall_time(wall0)
-
-  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
-
- !$OMP PARALLEL                   &
- !$OMP DEFAULT (NONE)             &
- !$OMP PRIVATE (i,j,k,m,integral) & 
- !$OMP SHARED (mo_num,three_e_4_idx_cycle_2_bi_ort)
- !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
-  do i = 1, mo_num
-    do k = 1, mo_num
-      do j = 1, mo_num
-        do m = 1, mo_num
-          call give_integrals_3_body_bi_ort(m, j, k, i, m, j, integral)
-          three_e_4_idx_cycle_2_bi_ort(m,j,k,i) = -1.d0 * integral 
-        enddo
-      enddo
-    enddo
-  enddo
- !$OMP END DO
- !$OMP END PARALLEL
-
-  call wall_time(wall1)
-  print *, ' wall time for three_e_4_idx_cycle_2_bi_ort', wall1 - wall0
-
-END_PROVIDER 
-
-! ---
-
-BEGIN_PROVIDER [ double precision, three_e_4_idx_exch23_bi_ort, (mo_num, mo_num, mo_num, mo_num)]
-
-  BEGIN_DOC
-  !
-  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
-  !
-  ! three_e_4_idx_exch23_bi_ort(m,j,k,i) = <mjk|-L|jmi> ::: notice that i is the RIGHT MO and k is the LEFT MO
-  !
-  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
-  !
-  END_DOC
-
-  implicit none
-  integer          :: i, j, k, m
-  double precision :: integral, wall1, wall0
-
-  three_e_4_idx_exch23_bi_ort = 0.d0
-  print *, ' Providing the three_e_4_idx_exch23_bi_ort ...'
-  call wall_time(wall0)
-
-  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
-
- !$OMP PARALLEL                   &
- !$OMP DEFAULT (NONE)             &
- !$OMP PRIVATE (i,j,k,m,integral) & 
- !$OMP SHARED (mo_num,three_e_4_idx_exch23_bi_ort)
- !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
-  do i = 1, mo_num
-    do k = 1, mo_num
-      do j = 1, mo_num
-        do m = 1, mo_num
-          call give_integrals_3_body_bi_ort(m, j, k, j, m, i, integral)
-          three_e_4_idx_exch23_bi_ort(m,j,k,i) = -1.d0 * integral 
-        enddo
-      enddo
-    enddo
-  enddo
- !$OMP END DO
- !$OMP END PARALLEL
-
-  call wall_time(wall1)
-  print *, ' wall time for three_e_4_idx_exch23_bi_ort', wall1 - wall0
-
-END_PROVIDER
-
-! ---
-
-BEGIN_PROVIDER [ double precision, three_e_4_idx_exch13_bi_ort, (mo_num, mo_num, mo_num, mo_num)]
-
-  BEGIN_DOC
-  !
-  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
-  !
-  ! three_e_4_idx_exch13_bi_ort(m,j,k,i) = <mjk|-L|ijm> ::: notice that i is the RIGHT MO and k is the LEFT MO
-  !
-  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
-  END_DOC
-
-  implicit none
-  integer          :: i, j, k, m
-  double precision :: integral, wall1, wall0
-
-  three_e_4_idx_exch13_bi_ort = 0.d0
-  print *, ' Providing the three_e_4_idx_exch13_bi_ort ...'
-  call wall_time(wall0)
-
-  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
-
- !$OMP PARALLEL                   &
- !$OMP DEFAULT (NONE)             &
- !$OMP PRIVATE (i,j,k,m,integral) & 
- !$OMP SHARED (mo_num,three_e_4_idx_exch13_bi_ort)
- !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
-  do i = 1, mo_num
-    do k = 1, mo_num
-      do j = 1, mo_num
-        do m = 1, mo_num
-          call give_integrals_3_body_bi_ort(m, j, k, i, j, m, integral)
-          three_e_4_idx_exch13_bi_ort(m,j,k,i) = -1.d0 * integral 
-        enddo
-      enddo
-    enddo
-  enddo
- !$OMP END DO
- !$OMP END PARALLEL
-
-  call wall_time(wall1)
-  print *, ' wall time for three_e_4_idx_exch13_bi_ort', wall1 - wall0
-
-END_PROVIDER 
-
-! ---
-
-BEGIN_PROVIDER [ double precision, three_e_4_idx_exch12_bi_ort, (mo_num, mo_num, mo_num, mo_num)]
-
-  BEGIN_DOC
-  ! 
-  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
-  !
-  ! three_e_4_idx_exch12_bi_ort(m,j,k,i) = <mjk|-L|mij> ::: notice that i is the RIGHT MO and k is the LEFT MO
-  !
-  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
-  ! 
-  END_DOC
-
-  implicit none
-  integer          :: i, j, k, m
-  double precision :: integral, wall1, wall0
-
-  three_e_4_idx_exch12_bi_ort = 0.d0
-  print *, ' Providing the three_e_4_idx_exch12_bi_ort ...'
-  call wall_time(wall0)
-
-  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
-
- !$OMP PARALLEL                   &
- !$OMP DEFAULT (NONE)             &
- !$OMP PRIVATE (i,j,k,m,integral) & 
- !$OMP SHARED (mo_num,three_e_4_idx_exch12_bi_ort)
- !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
-  do i = 1, mo_num
-    do k = 1, mo_num
-      do j = 1, mo_num
-        do m = 1, mo_num
-          call give_integrals_3_body_bi_ort(m, j, k, m, i, j, integral)
-          three_e_4_idx_exch12_bi_ort(m,j,k,i) = -1.d0 * integral 
-        enddo
-      enddo
     enddo
   enddo
   !$OMP END DO
   !$OMP END PARALLEL
 
+  allocate(tmp_2d(mo_num,mo_num))
+  allocate(tmp1(n_points_final_grid,4,mo_num))
+  allocate(tmp2(n_points_final_grid,4,mo_num))
+
+  ! loops approach to break the O(N^4) scaling in memory
+  do k = 1, mo_num
+    do i = 1, mo_num
+
+    !$OMP PARALLEL                                                  &
+    !$OMP DEFAULT (NONE)                                            &
+    !$OMP PRIVATE (n, ipoint, tmp_loc_1, tmp_loc_2)                 &
+    !$OMP SHARED (mo_num, n_points_final_grid, i, k,                &
+    !$OMP         mos_l_in_r_array_transp, mos_r_in_r_array_transp, &
+    !$OMP         int2_grad1_u12_bimo_t, final_weight_at_r_vector,  &
+    !$OMP         tmp_aux_2, tmp1)
+    !$OMP DO
+    do n = 1, mo_num
+      do ipoint = 1, n_points_final_grid
+
+        tmp_loc_1 = mos_l_in_r_array_transp(ipoint,k) * mos_r_in_r_array_transp(ipoint,i)
+        tmp_loc_2 = tmp_aux_2(ipoint,n)
+
+        tmp1(ipoint,1,n) = int2_grad1_u12_bimo_t(ipoint,1,n,n) * tmp_loc_1 + int2_grad1_u12_bimo_t(ipoint,1,k,i) * tmp_loc_2
+        tmp1(ipoint,2,n) = int2_grad1_u12_bimo_t(ipoint,2,n,n) * tmp_loc_1 + int2_grad1_u12_bimo_t(ipoint,2,k,i) * tmp_loc_2
+        tmp1(ipoint,3,n) = int2_grad1_u12_bimo_t(ipoint,3,n,n) * tmp_loc_1 + int2_grad1_u12_bimo_t(ipoint,3,k,i) * tmp_loc_2
+        tmp1(ipoint,4,n) = int2_grad1_u12_bimo_t(ipoint,1,n,n) * int2_grad1_u12_bimo_t(ipoint,1,k,i) &
+                         + int2_grad1_u12_bimo_t(ipoint,2,n,n) * int2_grad1_u12_bimo_t(ipoint,2,k,i) &
+                         + int2_grad1_u12_bimo_t(ipoint,3,n,n) * int2_grad1_u12_bimo_t(ipoint,3,k,i)
+
+      enddo
+    enddo
+    !$OMP END DO
+    !$OMP END PARALLEL
+
+    call dgemm( 'T', 'N', mo_num, mo_num, 4*n_points_final_grid, 1.d0                       &
+              , tmp_aux_1(1,1,1), 4*n_points_final_grid, tmp1(1,1,1), 4*n_points_final_grid &
+              , 0.d0, tmp_2d(1,1), mo_num)
+
+    !$OMP PARALLEL DO PRIVATE(j,m)
+    do j = 1, mo_num
+      do m = 1, mo_num
+        three_e_4_idx_direct_bi_ort(m,j,k,i) = -tmp_2d(m,j)
+      enddo
+    enddo
+    !$OMP END PARALLEL DO
+
+
+
+    !$OMP PARALLEL                                                  &
+    !$OMP DEFAULT (NONE)                                            &
+    !$OMP PRIVATE (n, ipoint, tmp_loc_1, tmp_loc_2)                 &
+    !$OMP SHARED (mo_num, n_points_final_grid, i, k,                &
+    !$OMP         mos_l_in_r_array_transp, mos_r_in_r_array_transp, &
+    !$OMP         int2_grad1_u12_bimo_t, final_weight_at_r_vector,  &
+    !$OMP         tmp1, tmp2)
+    !$OMP DO
+    do n = 1, mo_num
+      do ipoint = 1, n_points_final_grid
+
+        tmp_loc_1 = mos_l_in_r_array_transp(ipoint,k) * mos_r_in_r_array_transp(ipoint,n)
+        tmp_loc_2 = mos_l_in_r_array_transp(ipoint,n) * mos_r_in_r_array_transp(ipoint,i)
+
+        tmp1(ipoint,1,n) = int2_grad1_u12_bimo_t(ipoint,1,n,i) * tmp_loc_1 + int2_grad1_u12_bimo_t(ipoint,1,k,n) * tmp_loc_2
+        tmp1(ipoint,2,n) = int2_grad1_u12_bimo_t(ipoint,2,n,i) * tmp_loc_1 + int2_grad1_u12_bimo_t(ipoint,2,k,n) * tmp_loc_2
+        tmp1(ipoint,3,n) = int2_grad1_u12_bimo_t(ipoint,3,n,i) * tmp_loc_1 + int2_grad1_u12_bimo_t(ipoint,3,k,n) * tmp_loc_2
+        tmp1(ipoint,4,n) = int2_grad1_u12_bimo_t(ipoint,1,n,i) * int2_grad1_u12_bimo_t(ipoint,1,k,n) &
+                         + int2_grad1_u12_bimo_t(ipoint,2,n,i) * int2_grad1_u12_bimo_t(ipoint,2,k,n) &
+                         + int2_grad1_u12_bimo_t(ipoint,3,n,i) * int2_grad1_u12_bimo_t(ipoint,3,k,n)
+
+        tmp2(ipoint,1,n) = final_weight_at_r_vector(ipoint) * int2_grad1_u12_bimo_t(ipoint,1,i,n)
+        tmp2(ipoint,2,n) = final_weight_at_r_vector(ipoint) * int2_grad1_u12_bimo_t(ipoint,2,i,n)
+        tmp2(ipoint,3,n) = final_weight_at_r_vector(ipoint) * int2_grad1_u12_bimo_t(ipoint,3,i,n)
+        tmp2(ipoint,4,n) = final_weight_at_r_vector(ipoint) * mos_l_in_r_array_transp(ipoint,i) * mos_r_in_r_array_transp(ipoint,n)
+      enddo
+    enddo
+    !$OMP END DO
+    !$OMP END PARALLEL
+
+    call dgemm( 'T', 'N', mo_num, mo_num, 4*n_points_final_grid, 1.d0                       &
+              , tmp1(1,1,1), 4*n_points_final_grid, tmp_aux_1(1,1,1), 4*n_points_final_grid &
+              , 0.d0, tmp_2d(1,1), mo_num)
+
+    !$OMP PARALLEL DO PRIVATE(j,m)
+    do j = 1, mo_num
+      do m = 1, mo_num
+        three_e_4_idx_exch13_bi_ort(m,j,k,i) = -tmp_2d(m,j)
+      enddo
+    enddo
+    !$OMP END PARALLEL DO
+
+    call dgemm( 'T', 'N', mo_num, mo_num, 4*n_points_final_grid, 1.d0                  &
+              , tmp1(1,1,1), 4*n_points_final_grid, tmp2(1,1,1), 4*n_points_final_grid &
+              , 0.d0, tmp_2d(1,1), mo_num)
+
+    !$OMP PARALLEL DO PRIVATE(j,m)
+    do j = 1, mo_num
+      do m = 1, mo_num
+        three_e_4_idx_cycle_1_bi_ort(m,i,k,j) = -tmp_2d(m,j)
+      enddo
+    enddo
+    !$OMP END PARALLEL DO
+
+  enddo ! i
+
+    do j = 1, mo_num
+
+      !$OMP PARALLEL                                                  &
+      !$OMP DEFAULT (NONE)                                            &
+      !$OMP PRIVATE (n, ipoint, tmp_loc_1, tmp_loc_2)                 &
+      !$OMP SHARED (mo_num, n_points_final_grid, j, k,                &
+      !$OMP         mos_l_in_r_array_transp, mos_r_in_r_array_transp, &
+      !$OMP         int2_grad1_u12_bimo_t, final_weight_at_r_vector,  &
+      !$OMP         tmp1, tmp2)
+      !$OMP DO
+      do n = 1, mo_num
+        do ipoint = 1, n_points_final_grid
+
+          tmp_loc_1 = final_weight_at_r_vector(ipoint) * mos_l_in_r_array_transp(ipoint,j) * mos_r_in_r_array_transp(ipoint,n)
+          tmp_loc_2 = final_weight_at_r_vector(ipoint) * mos_l_in_r_array_transp(ipoint,n) * mos_r_in_r_array_transp(ipoint,j)
+
+          tmp1(ipoint,1,n) = int2_grad1_u12_bimo_t(ipoint,1,n,j) * tmp_loc_1 + int2_grad1_u12_bimo_t(ipoint,1,j,n) * tmp_loc_2
+          tmp1(ipoint,2,n) = int2_grad1_u12_bimo_t(ipoint,2,n,j) * tmp_loc_1 + int2_grad1_u12_bimo_t(ipoint,2,j,n) * tmp_loc_2
+          tmp1(ipoint,3,n) = int2_grad1_u12_bimo_t(ipoint,3,n,j) * tmp_loc_1 + int2_grad1_u12_bimo_t(ipoint,3,j,n) * tmp_loc_2
+          tmp1(ipoint,4,n) = int2_grad1_u12_bimo_t(ipoint,1,n,j) * int2_grad1_u12_bimo_t(ipoint,1,j,n) &
+                           + int2_grad1_u12_bimo_t(ipoint,2,n,j) * int2_grad1_u12_bimo_t(ipoint,2,j,n) &
+                           + int2_grad1_u12_bimo_t(ipoint,3,n,j) * int2_grad1_u12_bimo_t(ipoint,3,j,n)
+
+          tmp2(ipoint,1,n) = int2_grad1_u12_bimo_t(ipoint,1,k,n)
+          tmp2(ipoint,2,n) = int2_grad1_u12_bimo_t(ipoint,2,k,n)
+          tmp2(ipoint,3,n) = int2_grad1_u12_bimo_t(ipoint,3,k,n)
+          tmp2(ipoint,4,n) = final_weight_at_r_vector(ipoint) * mos_l_in_r_array_transp(ipoint,k) * mos_r_in_r_array_transp(ipoint,n)
+        enddo
+      enddo
+      !$OMP END DO
+      !$OMP END PARALLEL
+
+      call dgemm( 'T', 'N', mo_num, mo_num, 4*n_points_final_grid, 1.d0                  &
+                , tmp1(1,1,1), 4*n_points_final_grid, tmp2(1,1,1), 4*n_points_final_grid &
+                , 0.d0, tmp_2d(1,1), mo_num)
+
+      !$OMP PARALLEL DO PRIVATE(i,m)
+      do i = 1, mo_num
+        do m = 1, mo_num
+          three_e_4_idx_exch23_bi_ort(m,j,k,i) = -tmp_2d(m,i)
+        enddo
+      enddo
+      !$OMP END PARALLEL DO
+
+    enddo ! j
+  enddo !k
+
+  deallocate(tmp_2d)
+  deallocate(tmp1)
+  deallocate(tmp2)
+  deallocate(tmp_aux_1)
+  deallocate(tmp_aux_2)
+
+
   call wall_time(wall1)
-  print *, ' wall time for three_e_4_idx_exch12_bi_ort', wall1 - wall0
+  print *, ' wall time for three_e_4_idx_bi_ort', wall1 - wall0
+  call print_memory_usage()
 
 END_PROVIDER 
 

src/bi_ort_ints/three_body_ijmk_n4.irp.f

@@ -0,0 +1,486 @@
+
+! ---
+
+ BEGIN_PROVIDER [ double precision, three_e_4_idx_direct_bi_ort_n4 ,  (mo_num, mo_num, mo_num, mo_num)]
+&BEGIN_PROVIDER [ double precision, three_e_4_idx_exch13_bi_ort_n4 ,  (mo_num, mo_num, mo_num, mo_num)]
+&BEGIN_PROVIDER [ double precision, three_e_4_idx_cycle_1_bi_ort_n4,  (mo_num, mo_num, mo_num, mo_num)]
+!&BEGIN_PROVIDER [ double precision, three_e_4_idx_exch12_bi_ort_n4,  (mo_num, mo_num, mo_num, mo_num)]
+!&BEGIN_PROVIDER [ double precision, three_e_4_idx_cycle_2_bi_ort_n4, (mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_4_idx_direct_bi_ort_n4 (m,j,k,i) = < m j k | -L | m j i > ::: notice that i is the RIGHT MO and k is the LEFT MO
+  ! three_e_4_idx_exch13_bi_ort_n4 (m,j,k,i) = < m j k | -L | i j m > ::: notice that i is the RIGHT MO and k is the LEFT MO
+  ! three_e_4_idx_exch12_bi_ort_n4 (m,j,k,i) = < m j k | -L | m i j > ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !                                          = three_e_4_idx_exch13_bi_ort_n4 (j,m,k,i) 
+  ! three_e_4_idx_cycle_1_bi_ort_n4(m,j,k,i) = < m j k | -L | j i m > ::: notice that i is the RIGHT MO and k is the LEFT MO
+  ! three_e_4_idx_cycle_2_bi_ort_n4(m,j,k,i) = < m j k | -L | i m j > ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !                                          = three_e_4_idx_cycle_1_bi_ort_n4(j,m,k,i)
+  !
+  ! notice the -1 sign: in this way three_e_4_idx_direct_bi_ort_n4 can be directly used to compute Slater rules with a + sign
+  !
+  ! three_e_4_idx_direct_bi_ort_n4 (m,j,k,i) : Lk Ri Imm Ijj + Lj Rj Imm Iki + Lm Rm Ijj Iki 
+  ! three_e_4_idx_exch13_bi_ort_n4 (m,j,k,i) : Lk Rm Imi Ijj + Lj Rj Imi Ikm + Lm Ri Ijj Ikm 
+  ! three_e_4_idx_cycle_1_bi_ort_n4(m,j,k,i) : Lk Rm Imj Iji + Lj Ri Imj Ikm + Lm Rj Iji Ikm 
+  !
+  END_DOC
+
+  implicit none
+  integer                       :: ipoint, i, j, k, l, m
+  double precision              :: wall1, wall0
+  double precision, allocatable :: tmp1(:,:,:,:), tmp2(:,:,:,:), tmp3(:,:,:,:)
+  double precision, allocatable :: tmp_4d(:,:,:,:)
+  double precision, allocatable :: tmp4(:,:,:)
+  double precision, allocatable :: tmp5(:,:)
+  double precision, allocatable :: tmp_3d(:,:,:)
+
+  print *, ' Providing the O(N^4) three_e_4_idx_bi_ort ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+
+  allocate(tmp_4d(mo_num,mo_num,mo_num,mo_num))
+
+  allocate(tmp1(n_points_final_grid,3,mo_num,mo_num))
+  allocate(tmp2(n_points_final_grid,3,mo_num,mo_num))
+  allocate(tmp3(n_points_final_grid,3,mo_num,mo_num))
+
+  !$OMP PARALLEL                                                  &
+  !$OMP DEFAULT (NONE)                                            &
+  !$OMP PRIVATE (i, l, ipoint)                                    &
+  !$OMP SHARED (mo_num, n_points_final_grid,                      &
+  !$OMP         mos_l_in_r_array_transp, mos_r_in_r_array_transp, &
+  !$OMP         int2_grad1_u12_bimo_t, final_weight_at_r_vector,  &
+  !$OMP         tmp1, tmp2, tmp3)
+  !$OMP DO COLLAPSE(2)
+  do i = 1, mo_num
+    do l = 1, mo_num
+      do ipoint = 1, n_points_final_grid
+
+        tmp1(ipoint,1,l,i) = int2_grad1_u12_bimo_t(ipoint,1,l,l) * mos_l_in_r_array_transp(ipoint,i) * final_weight_at_r_vector(ipoint)
+        tmp1(ipoint,2,l,i) = int2_grad1_u12_bimo_t(ipoint,2,l,l) * mos_l_in_r_array_transp(ipoint,i) * final_weight_at_r_vector(ipoint)
+        tmp1(ipoint,3,l,i) = int2_grad1_u12_bimo_t(ipoint,3,l,l) * mos_l_in_r_array_transp(ipoint,i) * final_weight_at_r_vector(ipoint)
+
+        tmp2(ipoint,1,l,i) = int2_grad1_u12_bimo_t(ipoint,1,l,l) * mos_r_in_r_array_transp(ipoint,i)
+        tmp2(ipoint,2,l,i) = int2_grad1_u12_bimo_t(ipoint,2,l,l) * mos_r_in_r_array_transp(ipoint,i)
+        tmp2(ipoint,3,l,i) = int2_grad1_u12_bimo_t(ipoint,3,l,l) * mos_r_in_r_array_transp(ipoint,i)
+
+        tmp3(ipoint,1,l,i) = int2_grad1_u12_bimo_t(ipoint,1,l,i) * mos_r_in_r_array_transp(ipoint,l)
+        tmp3(ipoint,2,l,i) = int2_grad1_u12_bimo_t(ipoint,2,l,i) * mos_r_in_r_array_transp(ipoint,l)
+        tmp3(ipoint,3,l,i) = int2_grad1_u12_bimo_t(ipoint,3,l,i) * mos_r_in_r_array_transp(ipoint,l)
+      enddo
+    enddo
+  enddo
+  !$OMP END DO
+  !$OMP END PARALLEL
+
+
+  call dgemm( 'T', 'N', mo_num*mo_num, mo_num*mo_num, 3*n_points_final_grid, 1.d0        &
+            , tmp1(1,1,1,1), 3*n_points_final_grid, tmp2(1,1,1,1), 3*n_points_final_grid &
+            , 0.d0, tmp_4d(1,1,1,1), mo_num*mo_num)
+
+  !$OMP PARALLEL DO PRIVATE(i,j,k,m)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          three_e_4_idx_direct_bi_ort_n4(m,j,k,i) = -tmp_4d(m,k,j,i)
+        enddo
+      enddo
+    enddo
+  enddo
+  !$OMP END PARALLEL DO
+
+  call dgemm( 'T', 'N', mo_num*mo_num, mo_num*mo_num, 3*n_points_final_grid, 1.d0        &
+            , tmp3(1,1,1,1), 3*n_points_final_grid, tmp1(1,1,1,1), 3*n_points_final_grid &
+            , 0.d0, tmp_4d(1,1,1,1), mo_num*mo_num)
+
+
+  !$OMP PARALLEL DO PRIVATE(i,j,k,m)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          three_e_4_idx_exch13_bi_ort_n4(m,j,k,i) = -tmp_4d(m,i,j,k)
+        enddo
+      enddo
+    enddo
+  enddo
+  !$OMP END PARALLEL DO
+
+
+
+  !$OMP PARALLEL                                                  &
+  !$OMP DEFAULT (NONE)                                            &
+  !$OMP PRIVATE (i, l, ipoint)                                    &
+  !$OMP SHARED (mo_num, n_points_final_grid,                      &
+  !$OMP         mos_l_in_r_array_transp, mos_r_in_r_array_transp, &
+  !$OMP         int2_grad1_u12_bimo_t, final_weight_at_r_vector,  &
+  !$OMP         tmp1)
+  !$OMP DO COLLAPSE(2)
+  do i = 1, mo_num
+    do l = 1, mo_num
+      do ipoint = 1, n_points_final_grid
+        tmp1(ipoint,1,l,i) = int2_grad1_u12_bimo_t(ipoint,1,i,l) * mos_l_in_r_array_transp(ipoint,l) * final_weight_at_r_vector(ipoint)
+        tmp1(ipoint,2,l,i) = int2_grad1_u12_bimo_t(ipoint,2,i,l) * mos_l_in_r_array_transp(ipoint,l) * final_weight_at_r_vector(ipoint)
+        tmp1(ipoint,3,l,i) = int2_grad1_u12_bimo_t(ipoint,3,i,l) * mos_l_in_r_array_transp(ipoint,l) * final_weight_at_r_vector(ipoint)
+      enddo
+    enddo
+  enddo
+  !$OMP END DO
+  !$OMP END PARALLEL
+
+
+  call dgemm( 'T', 'N', mo_num*mo_num, mo_num*mo_num, 3*n_points_final_grid, 1.d0        &
+            , tmp1(1,1,1,1), 3*n_points_final_grid, tmp2(1,1,1,1), 3*n_points_final_grid &
+            , 0.d0, tmp_4d(1,1,1,1), mo_num*mo_num)
+
+
+  deallocate(tmp2)
+
+  !$OMP PARALLEL DO PRIVATE(i,j,k,m)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          three_e_4_idx_exch13_bi_ort_n4(m,j,k,i) = three_e_4_idx_exch13_bi_ort_n4(m,j,k,i) - tmp_4d(m,k,j,i)
+        enddo
+      enddo
+    enddo
+  enddo
+  !$OMP END PARALLEL DO
+
+  call dgemm( 'T', 'N', mo_num*mo_num, mo_num*mo_num, 3*n_points_final_grid, 1.d0        &
+            , tmp1(1,1,1,1), 3*n_points_final_grid, tmp3(1,1,1,1), 3*n_points_final_grid &
+            , 0.d0, tmp_4d(1,1,1,1), mo_num*mo_num)
+
+  deallocate(tmp3)
+
+  !$OMP PARALLEL DO PRIVATE(i,j,k,m)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          three_e_4_idx_cycle_1_bi_ort_n4(m,j,k,i) = -tmp_4d(m,k,j,i)
+        enddo
+      enddo
+    enddo
+  enddo
+  !$OMP END PARALLEL DO
+
+
+
+  !$OMP PARALLEL                                                  &
+  !$OMP DEFAULT (NONE)                                            &
+  !$OMP PRIVATE (i, l, ipoint)                                    &
+  !$OMP SHARED (mo_num, n_points_final_grid,                      &
+  !$OMP         mos_l_in_r_array_transp, mos_r_in_r_array_transp, &
+  !$OMP         int2_grad1_u12_bimo_t, final_weight_at_r_vector,  &
+  !$OMP         tmp1)
+  !$OMP DO COLLAPSE(2)
+  do i = 1, mo_num
+    do l = 1, mo_num
+      do ipoint = 1, n_points_final_grid
+        tmp1(ipoint,1,l,i) = final_weight_at_r_vector(ipoint) * int2_grad1_u12_bimo_t(ipoint,1,l,l) * mos_l_in_r_array_transp(ipoint,i) * mos_r_in_r_array_transp(ipoint,i)
+        tmp1(ipoint,2,l,i) = final_weight_at_r_vector(ipoint) * int2_grad1_u12_bimo_t(ipoint,2,l,l) * mos_l_in_r_array_transp(ipoint,i) * mos_r_in_r_array_transp(ipoint,i)
+        tmp1(ipoint,3,l,i) = final_weight_at_r_vector(ipoint) * int2_grad1_u12_bimo_t(ipoint,3,l,l) * mos_l_in_r_array_transp(ipoint,i) * mos_r_in_r_array_transp(ipoint,i)
+      enddo
+    enddo
+  enddo
+  !$OMP END DO
+  !$OMP END PARALLEL
+
+  call dgemm( 'T', 'N', mo_num*mo_num, mo_num*mo_num, 3*n_points_final_grid, 1.d0                         &
+            , tmp1(1,1,1,1), 3*n_points_final_grid, int2_grad1_u12_bimo_t(1,1,1,1), 3*n_points_final_grid &
+            , 0.d0, tmp_4d(1,1,1,1), mo_num*mo_num)
+
+  deallocate(tmp1)
+
+  !$OMP PARALLEL DO PRIVATE(i,j,k,m)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          three_e_4_idx_direct_bi_ort_n4(m,j,k,i) = three_e_4_idx_direct_bi_ort_n4(m,j,k,i) - tmp_4d(m,j,k,i) - tmp_4d(j,m,k,i)
+        enddo
+      enddo
+    enddo
+  enddo
+  !$OMP END PARALLEL DO
+
+  deallocate(tmp_4d)
+
+
+  allocate(tmp_3d(mo_num,mo_num,mo_num))
+  allocate(tmp5(n_points_final_grid,mo_num))
+
+  !$OMP PARALLEL                                                  &
+  !$OMP DEFAULT (NONE)                                            &
+  !$OMP PRIVATE (i, ipoint)                                       &
+  !$OMP SHARED (mo_num, n_points_final_grid,                      &
+  !$OMP         mos_l_in_r_array_transp, mos_r_in_r_array_transp, &
+  !$OMP         final_weight_at_r_vector,                         &
+  !$OMP         tmp5)
+  !$OMP DO
+  do i = 1, mo_num
+    do ipoint = 1, n_points_final_grid
+      tmp5(ipoint,i) = final_weight_at_r_vector(ipoint) * mos_l_in_r_array_transp(ipoint,i) * mos_r_in_r_array_transp(ipoint,i)
+    enddo
+  enddo
+  !$OMP END DO
+  !$OMP END PARALLEL
+
+
+  allocate(tmp4(n_points_final_grid,mo_num,mo_num))
+
+  do m = 1, mo_num
+
+    !$OMP PARALLEL                                                 &
+    !$OMP DEFAULT (NONE)                                           &
+    !$OMP PRIVATE (i, k, ipoint)                                   &
+    !$OMP SHARED (mo_num, n_points_final_grid, m,                  &
+    !$OMP         int2_grad1_u12_bimo_t,                           &
+    !$OMP         tmp4)
+    !$OMP DO COLLAPSE(2)
+    do i = 1, mo_num
+      do k = 1, mo_num
+        do ipoint = 1, n_points_final_grid
+
+          tmp4(ipoint,k,i) = int2_grad1_u12_bimo_t(ipoint,1,k,m) * int2_grad1_u12_bimo_t(ipoint,1,m,i) &
+                           + int2_grad1_u12_bimo_t(ipoint,2,k,m) * int2_grad1_u12_bimo_t(ipoint,2,m,i) &
+                           + int2_grad1_u12_bimo_t(ipoint,3,k,m) * int2_grad1_u12_bimo_t(ipoint,3,m,i)
+        enddo
+      enddo
+    enddo
+    !$OMP END DO
+    !$OMP END PARALLEL
+
+    call dgemm( 'T', 'N', mo_num, mo_num*mo_num, n_points_final_grid, 1.d0       &
+              , tmp5(1,1), n_points_final_grid, tmp4(1,1,1), n_points_final_grid &
+              , 0.d0, tmp_3d(1,1,1), mo_num)
+
+    !$OMP PARALLEL DO PRIVATE(i,j,k)
+    do i = 1, mo_num
+      do k = 1, mo_num
+        do j = 1, mo_num
+          three_e_4_idx_exch13_bi_ort_n4(m,j,k,i) = three_e_4_idx_exch13_bi_ort_n4(m,j,k,i) - tmp_3d(j,k,i)
+        enddo
+      enddo
+    enddo
+    !$OMP END PARALLEL DO
+
+
+
+    !$OMP PARALLEL                                                 &
+    !$OMP DEFAULT (NONE)                                           &
+    !$OMP PRIVATE (j, k, ipoint)                                   &
+    !$OMP SHARED (mo_num, n_points_final_grid, m,                  &
+    !$OMP         mos_l_in_r_array_transp,                         &
+    !$OMP         int2_grad1_u12_bimo_t, final_weight_at_r_vector, &
+    !$OMP         tmp4)
+    !$OMP DO COLLAPSE(2)
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do ipoint = 1, n_points_final_grid
+
+          tmp4(ipoint,j,k) = final_weight_at_r_vector(ipoint) * mos_l_in_r_array_transp(ipoint,j)        &
+                           * ( int2_grad1_u12_bimo_t(ipoint,1,m,j) * int2_grad1_u12_bimo_t(ipoint,1,k,m) &
+                             + int2_grad1_u12_bimo_t(ipoint,2,m,j) * int2_grad1_u12_bimo_t(ipoint,2,k,m) &
+                             + int2_grad1_u12_bimo_t(ipoint,3,m,j) * int2_grad1_u12_bimo_t(ipoint,3,k,m) )
+        enddo
+      enddo
+    enddo
+    !$OMP END DO
+    !$OMP END PARALLEL
+
+    call dgemm( 'T', 'N', mo_num*mo_num, mo_num, n_points_final_grid, 1.d0                          &
+              , tmp4(1,1,1), n_points_final_grid, mos_r_in_r_array_transp(1,1), n_points_final_grid &
+              , 0.d0, tmp_3d(1,1,1), mo_num*mo_num)
+
+    !$OMP PARALLEL DO PRIVATE(i,j,k)
+    do i = 1, mo_num
+      do k = 1, mo_num
+        do j = 1, mo_num
+          three_e_4_idx_cycle_1_bi_ort_n4(m,j,k,i) = three_e_4_idx_cycle_1_bi_ort_n4(m,j,k,i) - tmp_3d(j,k,i)
+        enddo
+      enddo
+    enddo
+    !$OMP END PARALLEL DO
+
+  enddo
+
+  deallocate(tmp5)
+  deallocate(tmp_3d)
+
+
+
+  do i = 1, mo_num
+
+    !$OMP PARALLEL                                                 &
+    !$OMP DEFAULT (NONE)                                           &
+    !$OMP PRIVATE (m, j, ipoint)                                   &
+    !$OMP SHARED (mo_num, n_points_final_grid, i,                  &
+    !$OMP         mos_r_in_r_array_transp,                         &
+    !$OMP         int2_grad1_u12_bimo_t, final_weight_at_r_vector, &
+    !$OMP         tmp4)
+    !$OMP DO COLLAPSE(2)
+    do j = 1, mo_num
+      do m = 1, mo_num
+        do ipoint = 1, n_points_final_grid
+
+          tmp4(ipoint,m,j) = final_weight_at_r_vector(ipoint) * mos_r_in_r_array_transp(ipoint,m)        &
+                           * ( int2_grad1_u12_bimo_t(ipoint,1,m,j) * int2_grad1_u12_bimo_t(ipoint,1,j,i) &
+                             + int2_grad1_u12_bimo_t(ipoint,2,m,j) * int2_grad1_u12_bimo_t(ipoint,2,j,i) &
+                             + int2_grad1_u12_bimo_t(ipoint,3,m,j) * int2_grad1_u12_bimo_t(ipoint,3,j,i) )
+        enddo
+      enddo
+    enddo
+    !$OMP END DO
+    !$OMP END PARALLEL
+
+    call dgemm( 'T', 'N', mo_num*mo_num, mo_num, n_points_final_grid, -1.d0                         &
+              , tmp4(1,1,1), n_points_final_grid, mos_l_in_r_array_transp(1,1), n_points_final_grid &
+              , 1.d0, three_e_4_idx_cycle_1_bi_ort_n4(1,1,1,i), mo_num*mo_num)
+
+  enddo
+
+  deallocate(tmp4)
+
+
+!  !$OMP PARALLEL DO PRIVATE(i,j,k,m)
+!  do i = 1, mo_num
+!    do k = 1, mo_num
+!      do j = 1, mo_num
+!        do m = 1, mo_num
+!          three_e_4_idx_exch12_bi_ort_n4 (m,j,k,i) = three_e_4_idx_exch13_bi_ort_n4 (j,m,k,i)
+!          three_e_4_idx_cycle_2_bi_ort_n4(m,j,k,i) = three_e_4_idx_cycle_1_bi_ort_n4(j,m,k,i)
+!        enddo
+!      enddo
+!    enddo
+!  enddo
+!  !$OMP END PARALLEL DO
+
+
+  call wall_time(wall1)
+  print *, ' wall time for O(N^4) three_e_4_idx_bi_ort', wall1 - wall0
+  call print_memory_usage()
+
+END_PROVIDER 
+
+! ---
+
+BEGIN_PROVIDER [ double precision, three_e_4_idx_exch23_bi_ort_n4 , (mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_4_idx_exch23_bi_ort_n4 (m,j,k,i) = < m j k | -L | j m i > ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_4_idx_direct_bi_ort_n4 can be directly used to compute Slater rules with a + sign
+  !
+  ! three_e_4_idx_exch23_bi_ort_n4 (m,j,k,i) : Lk Ri Imj Ijm + Lj Rm Imj Iki + Lm Rj Ijm Iki
+  !
+  END_DOC
+
+  implicit none
+  integer                       :: i, j, k, l, m, ipoint
+  double precision              :: wall1, wall0
+  double precision, allocatable :: tmp1(:,:,:,:), tmp_4d(:,:,:,:)
+  double precision, allocatable :: tmp5(:,:,:), tmp6(:,:,:)
+
+  print *, ' Providing the O(N^4) three_e_4_idx_exch23_bi_ort_n4 ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+
+  allocate(tmp5(n_points_final_grid,mo_num,mo_num))
+  allocate(tmp6(n_points_final_grid,mo_num,mo_num))
+
+  !$OMP PARALLEL                                                  &
+  !$OMP DEFAULT (NONE)                                            &
+  !$OMP PRIVATE (i, l, ipoint)                                    &
+  !$OMP SHARED (mo_num, n_points_final_grid,                      &
+  !$OMP         mos_l_in_r_array_transp, mos_r_in_r_array_transp, &
+  !$OMP         int2_grad1_u12_bimo_t, final_weight_at_r_vector,  &
+  !$OMP         tmp5, tmp6)
+  !$OMP DO COLLAPSE(2)
+  do i = 1, mo_num
+    do l = 1, mo_num
+      do ipoint = 1, n_points_final_grid
+
+        tmp5(ipoint,l,i) = int2_grad1_u12_bimo_t(ipoint,1,l,i) * int2_grad1_u12_bimo_t(ipoint,1,i,l) &
+                         + int2_grad1_u12_bimo_t(ipoint,2,l,i) * int2_grad1_u12_bimo_t(ipoint,2,i,l) &
+                         + int2_grad1_u12_bimo_t(ipoint,3,l,i) * int2_grad1_u12_bimo_t(ipoint,3,i,l) 
+
+        tmp6(ipoint,l,i) = final_weight_at_r_vector(ipoint) * mos_l_in_r_array_transp(ipoint,l) * mos_r_in_r_array_transp(ipoint,i)
+      enddo
+    enddo
+  enddo
+  !$OMP END DO
+  !$OMP END PARALLEL
+
+  call dgemm( 'T', 'N', mo_num*mo_num, mo_num*mo_num, n_points_final_grid, -1.d0 &
+            , tmp5(1,1,1), n_points_final_grid, tmp6(1,1,1), n_points_final_grid &
+            , 0.d0, three_e_4_idx_exch23_bi_ort_n4(1,1,1,1), mo_num*mo_num)
+
+  deallocate(tmp5)
+  deallocate(tmp6)
+
+
+  allocate(tmp_4d(mo_num,mo_num,mo_num,mo_num))
+  allocate(tmp1(n_points_final_grid,3,mo_num,mo_num))
+
+  !$OMP PARALLEL                                                  &
+  !$OMP DEFAULT (NONE)                                            &
+  !$OMP PRIVATE (i, l, ipoint)                                    &
+  !$OMP SHARED (mo_num, n_points_final_grid,                      &
+  !$OMP         mos_l_in_r_array_transp, mos_r_in_r_array_transp, &
+  !$OMP         int2_grad1_u12_bimo_t, final_weight_at_r_vector,  &
+  !$OMP         tmp1)
+  !$OMP DO COLLAPSE(2)
+  do i = 1, mo_num
+    do l = 1, mo_num
+      do ipoint = 1, n_points_final_grid
+        tmp1(ipoint,1,l,i) = final_weight_at_r_vector(ipoint) * int2_grad1_u12_bimo_t(ipoint,1,l,i) * mos_l_in_r_array_transp(ipoint,i) * mos_r_in_r_array_transp(ipoint,l)
+        tmp1(ipoint,2,l,i) = final_weight_at_r_vector(ipoint) * int2_grad1_u12_bimo_t(ipoint,2,l,i) * mos_l_in_r_array_transp(ipoint,i) * mos_r_in_r_array_transp(ipoint,l)
+        tmp1(ipoint,3,l,i) = final_weight_at_r_vector(ipoint) * int2_grad1_u12_bimo_t(ipoint,3,l,i) * mos_l_in_r_array_transp(ipoint,i) * mos_r_in_r_array_transp(ipoint,l)
+      enddo
+    enddo
+  enddo
+  !$OMP END DO
+  !$OMP END PARALLEL
+
+  call dgemm( 'T', 'N', mo_num*mo_num, mo_num*mo_num, 3*n_points_final_grid, 1.d0                         &
+            , tmp1(1,1,1,1), 3*n_points_final_grid, int2_grad1_u12_bimo_t(1,1,1,1), 3*n_points_final_grid &
+            , 0.d0, tmp_4d(1,1,1,1), mo_num*mo_num)
+
+  deallocate(tmp1)
+
+  !$OMP PARALLEL DO PRIVATE(i,j,k,m)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          three_e_4_idx_exch23_bi_ort_n4(m,j,k,i) = three_e_4_idx_exch23_bi_ort_n4(m,j,k,i) - tmp_4d(m,j,k,i) - tmp_4d(j,m,k,i)
+        enddo
+      enddo
+    enddo
+  enddo
+  !$OMP END PARALLEL DO
+
+  deallocate(tmp_4d)
+
+
+  call wall_time(wall1)
+  print *, ' wall time for O(N^4) three_e_4_idx_exch23_bi_ort_n4', wall1 - wall0
+  call print_memory_usage()
+
+END_PROVIDER 
+
+! ---
+

src/bi_ort_ints/three_body_ijmk_old.irp.f

@@ -0,0 +1,290 @@
+
+! ---
+
+BEGIN_PROVIDER [ double precision, three_e_4_idx_direct_bi_ort_old, (mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_4_idx_direct_bi_ort_old(m,j,k,i) = <mjk|-L|mji> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  !
+  END_DOC
+
+ implicit none
+ integer          :: i, j, k, m
+ double precision :: integral, wall1, wall0
+
+  three_e_4_idx_direct_bi_ort_old = 0.d0
+  print *, ' Providing the three_e_4_idx_direct_bi_ort_old ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+ !$OMP PARALLEL                   &
+ !$OMP DEFAULT (NONE)             &
+ !$OMP PRIVATE (i,j,k,m,integral) & 
+ !$OMP SHARED (mo_num,three_e_4_idx_direct_bi_ort_old)
+ !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          call give_integrals_3_body_bi_ort(m, j, k, m, j, i, integral)
+          three_e_4_idx_direct_bi_ort_old(m,j,k,i) = -1.d0 * integral 
+        enddo
+      enddo
+    enddo
+  enddo
+ !$OMP END DO
+ !$OMP END PARALLEL
+
+  call wall_time(wall1)
+  print *, ' wall time for three_e_4_idx_direct_bi_ort_old', wall1 - wall0
+  call print_memory_usage()
+
+END_PROVIDER 
+
+! ---
+
+BEGIN_PROVIDER [ double precision, three_e_4_idx_cycle_1_bi_ort_old, (mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE FIRST CYCLIC PERMUTATION TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_4_idx_cycle_1_bi_ort_old(m,j,k,i) = <mjk|-L|jim> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  !
+  END_DOC
+
+  implicit none
+  integer          :: i, j, k, m
+  double precision :: integral, wall1, wall0
+
+  three_e_4_idx_cycle_1_bi_ort_old = 0.d0
+  print *, ' Providing the three_e_4_idx_cycle_1_bi_ort_old ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+ !$OMP PARALLEL                   &
+ !$OMP DEFAULT (NONE)             &
+ !$OMP PRIVATE (i,j,k,m,integral) & 
+ !$OMP SHARED (mo_num,three_e_4_idx_cycle_1_bi_ort_old)
+ !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          call give_integrals_3_body_bi_ort(m, j, k, j, i, m, integral)
+          three_e_4_idx_cycle_1_bi_ort_old(m,j,k,i) = -1.d0 * integral 
+        enddo
+      enddo
+    enddo
+  enddo
+ !$OMP END DO
+ !$OMP END PARALLEL
+
+  call wall_time(wall1)
+  print *, ' wall time for three_e_4_idx_cycle_1_bi_ort_old', wall1 - wall0
+  call print_memory_usage()
+
+END_PROVIDER 
+
+! --
+
+BEGIN_PROVIDER [ double precision, three_e_4_idx_cycle_2_bi_ort_old, (mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE FIRST CYCLIC PERMUTATION TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_4_idx_cycle_2_bi_ort_old(m,j,k,i) = <mjk|-L|imj> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  !
+  END_DOC
+
+  implicit none
+  integer          :: i, j, k, m
+  double precision :: integral, wall1, wall0
+
+  three_e_4_idx_cycle_2_bi_ort_old = 0.d0
+  print *, ' Providing the three_e_4_idx_cycle_2_bi_ort_old ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+ !$OMP PARALLEL                   &
+ !$OMP DEFAULT (NONE)             &
+ !$OMP PRIVATE (i,j,k,m,integral) & 
+ !$OMP SHARED (mo_num,three_e_4_idx_cycle_2_bi_ort_old)
+ !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          call give_integrals_3_body_bi_ort(m, j, k, i, m, j, integral)
+          three_e_4_idx_cycle_2_bi_ort_old(m,j,k,i) = -1.d0 * integral 
+        enddo
+      enddo
+    enddo
+  enddo
+ !$OMP END DO
+ !$OMP END PARALLEL
+
+  call wall_time(wall1)
+  print *, ' wall time for three_e_4_idx_cycle_2_bi_ort_old', wall1 - wall0
+  call print_memory_usage()
+
+END_PROVIDER 
+
+! ---
+
+BEGIN_PROVIDER [ double precision, three_e_4_idx_exch23_bi_ort_old, (mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_4_idx_exch23_bi_ort_old(m,j,k,i) = <mjk|-L|jmi> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  !
+  END_DOC
+
+  implicit none
+  integer          :: i, j, k, m
+  double precision :: integral, wall1, wall0
+
+  three_e_4_idx_exch23_bi_ort_old = 0.d0
+  print *, ' Providing the three_e_4_idx_exch23_bi_ort_old ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+ !$OMP PARALLEL                   &
+ !$OMP DEFAULT (NONE)             &
+ !$OMP PRIVATE (i,j,k,m,integral) & 
+ !$OMP SHARED (mo_num,three_e_4_idx_exch23_bi_ort_old)
+ !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          call give_integrals_3_body_bi_ort(m, j, k, j, m, i, integral)
+          three_e_4_idx_exch23_bi_ort_old(m,j,k,i) = -1.d0 * integral 
+        enddo
+      enddo
+    enddo
+  enddo
+ !$OMP END DO
+ !$OMP END PARALLEL
+
+  call wall_time(wall1)
+  print *, ' wall time for three_e_4_idx_exch23_bi_ort_old', wall1 - wall0
+  call print_memory_usage()
+
+END_PROVIDER
+
+! ---
+
+BEGIN_PROVIDER [ double precision, three_e_4_idx_exch13_bi_ort_old, (mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_4_idx_exch13_bi_ort_old(m,j,k,i) = <mjk|-L|ijm> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  END_DOC
+
+  implicit none
+  integer          :: i, j, k, m
+  double precision :: integral, wall1, wall0
+
+  three_e_4_idx_exch13_bi_ort_old = 0.d0
+  print *, ' Providing the three_e_4_idx_exch13_bi_ort_old ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+ !$OMP PARALLEL                   &
+ !$OMP DEFAULT (NONE)             &
+ !$OMP PRIVATE (i,j,k,m,integral) & 
+ !$OMP SHARED (mo_num,three_e_4_idx_exch13_bi_ort_old)
+ !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          call give_integrals_3_body_bi_ort(m, j, k, i, j, m, integral)
+          three_e_4_idx_exch13_bi_ort_old(m,j,k,i) = -1.d0 * integral 
+        enddo
+      enddo
+    enddo
+  enddo
+ !$OMP END DO
+ !$OMP END PARALLEL
+
+  call wall_time(wall1)
+  print *, ' wall time for three_e_4_idx_exch13_bi_ort_old', wall1 - wall0
+  call print_memory_usage()
+
+END_PROVIDER 
+
+! ---
+
+BEGIN_PROVIDER [ double precision, three_e_4_idx_exch12_bi_ort_old, (mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  ! 
+  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF SINGLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_4_idx_exch12_bi_ort_old(m,j,k,i) = <mjk|-L|mij> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  ! 
+  END_DOC
+
+  implicit none
+  integer          :: i, j, k, m
+  double precision :: integral, wall1, wall0
+
+  three_e_4_idx_exch12_bi_ort_old = 0.d0
+  print *, ' Providing the three_e_4_idx_exch12_bi_ort_old ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+ !$OMP PARALLEL                   &
+ !$OMP DEFAULT (NONE)             &
+ !$OMP PRIVATE (i,j,k,m,integral) & 
+ !$OMP SHARED (mo_num,three_e_4_idx_exch12_bi_ort_old)
+ !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          call give_integrals_3_body_bi_ort(m, j, k, m, i, j, integral)
+          three_e_4_idx_exch12_bi_ort_old(m,j,k,i) = -1.d0 * integral 
+        enddo
+      enddo
+    enddo
+  enddo
+  !$OMP END DO
+  !$OMP END PARALLEL
+
+  call wall_time(wall1)
+  print *, ' wall time for three_e_4_idx_exch12_bi_ort_old', wall1 - wall0
+  call print_memory_usage()
+
+END_PROVIDER 
+
+! ---
+

src/bi_ort_ints/three_body_ijmkl.irp.f

@@ -1,296 +1,245 @@
-
 ! ---
+double precision function three_e_5_idx_exch12_bi_ort(m,l,i,k,j) result(integral)
+  implicit none
+  integer, intent(in) :: m,l,j,k,i
+  integral = three_e_5_idx_direct_bi_ort(m,l,j,k,i)
+end
 
-BEGIN_PROVIDER [ double precision, three_e_5_idx_direct_bi_ort, (mo_num, mo_num, mo_num, mo_num, mo_num)]
+ BEGIN_PROVIDER [ double precision, three_e_5_idx_direct_bi_ort , (mo_num, mo_num, mo_num, mo_num, mo_num)]
+&BEGIN_PROVIDER [ double precision, three_e_5_idx_exch23_bi_ort , (mo_num, mo_num, mo_num, mo_num, mo_num)]
+&BEGIN_PROVIDER [ double precision, three_e_5_idx_exch13_bi_ort , (mo_num, mo_num, mo_num, mo_num, mo_num)]
+&BEGIN_PROVIDER [ double precision, three_e_5_idx_cycle_1_bi_ort, (mo_num, mo_num, mo_num, mo_num, mo_num)]
+&BEGIN_PROVIDER [ double precision, three_e_5_idx_cycle_2_bi_ort, (mo_num, mo_num, mo_num, mo_num, mo_num)]
 
   BEGIN_DOC
   !
   ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF DOUBLE EXCITATIONS AND BI ORTHO MOs
   !
-  ! three_e_5_idx_direct_bi_ort(m,l,j,k,i) = <mlk|-L|mji> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  ! three_e_5_idx_direct_bi_ort(m,l,j,k,i) = <mlk|-L|mji> :: : notice that i is the RIGHT MO and k is the LEFT MO
   !
   ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  !
   END_DOC
 
   implicit none
-  integer          :: i, j, k, m, l
-  double precision :: integral, wall1, wall0
 
-  three_e_5_idx_direct_bi_ort = 0.d0
+  integer                        :: i, j, k, m, l
-  print *, ' Providing the three_e_5_idx_direct_bi_ort ...'
+  double precision               :: wall1, wall0
-  call wall_time(wall0)
+  integer                        :: ipoint
+  double precision, allocatable  :: grad_mli(:,:), orb_mat(:,:,:)
+  double precision, allocatable  :: lk_grad_mi(:,:,:,:), rk_grad_im(:,:,:)
+  double precision, allocatable  :: lm_grad_ik(:,:,:,:), rm_grad_ik(:,:,:)
+  double precision, allocatable  :: tmp_mat(:,:,:)
 
   provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+  PROVIDE mo_l_coef mo_r_coef int2_grad1_u12_bimo_t
+
+  call print_memory_usage
+  print *, ' Providing the three_e_5_idx_bi_ort ...'
+  call wall_time(wall0)
+
+  three_e_5_idx_direct_bi_ort (:,:,:,:,:) = 0.d0
+  three_e_5_idx_cycle_1_bi_ort(:,:,:,:,:) = 0.d0
+  three_e_5_idx_cycle_2_bi_ort(:,:,:,:,:) = 0.d0
+  three_e_5_idx_exch23_bi_ort (:,:,:,:,:) = 0.d0
+  three_e_5_idx_exch13_bi_ort (:,:,:,:,:) = 0.d0
+
+  call print_memory_usage
+
+  allocate(tmp_mat(mo_num,mo_num,mo_num))
+  allocate(orb_mat(n_points_final_grid,mo_num,mo_num))
+
+  !$OMP PARALLEL DO PRIVATE (i,l,ipoint)
+  do i=1,mo_num
+    do l=1,mo_num
+      do ipoint=1, n_points_final_grid
+
+        orb_mat(ipoint,l,i) = final_weight_at_r_vector(ipoint)       &
+            * mos_l_in_r_array_transp(ipoint,l)                      &
+            * mos_r_in_r_array_transp(ipoint,i)
 
- !$OMP PARALLEL                     &
- !$OMP DEFAULT (NONE)               &
- !$OMP PRIVATE (i,j,k,m,l,integral) & 
- !$OMP SHARED (mo_num,three_e_5_idx_direct_bi_ort)
- !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
-  do i = 1, mo_num
-    do k = 1, mo_num
-      do j = 1, mo_num
-        do l = 1, mo_num
-          do m = 1, mo_num
-            call give_integrals_3_body_bi_ort(m, l, k, m, j, i, integral)
-            three_e_5_idx_direct_bi_ort(m,l,j,k,i) = -1.d0 * integral 
-          enddo
-        enddo
       enddo
     enddo
   enddo
- !$OMP END DO
+  !$OMP END PARALLEL DO
- !$OMP END PARALLEL
 
-  call wall_time(wall1)
+  tmp_mat = 0.d0
-  print *, ' wall time for three_e_5_idx_direct_bi_ort', wall1 - wall0
+  call print_memory_usage
 
-END_PROVIDER 
+  do m = 1, mo_num
 
-! ---
+    allocate(grad_mli(n_points_final_grid,mo_num))
 
-BEGIN_PROVIDER [ double precision, three_e_5_idx_cycle_1_bi_ort, (mo_num, mo_num, mo_num, mo_num, mo_num)]
+    do i=1,mo_num
+      !$OMP PARALLEL DO PRIVATE (l,ipoint)
+      do l=1,mo_num
+        do ipoint=1, n_points_final_grid
 
-  BEGIN_DOC
+          grad_mli(ipoint,l) =                                       &
-  !
+              int2_grad1_u12_bimo_t(ipoint,1,m,m) * int2_grad1_u12_bimo_t(ipoint,1,l,i) +&
-  ! matrix element of the -L  three-body operator FOR THE FIRST CYCLIC PERMUTATION TERMS OF DOUBLE EXCITATIONS AND BI ORTHO MOs
+              int2_grad1_u12_bimo_t(ipoint,2,m,m) * int2_grad1_u12_bimo_t(ipoint,2,l,i) +&
-  !
+              int2_grad1_u12_bimo_t(ipoint,3,m,m) * int2_grad1_u12_bimo_t(ipoint,3,l,i)
-  ! three_e_5_idx_cycle_1_bi_ort(m,l,j,k,i) = <mlk|-L|jim> ::: notice that i is the RIGHT MO and k is the LEFT MO
-  !
-  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
-  !
-  END_DOC
 
-  implicit none
-  integer          :: i, j, k, m, l
-  double precision :: integral, wall1, wall0
-
-  three_e_5_idx_cycle_1_bi_ort = 0.d0
-  print *, ' Providing the three_e_5_idx_cycle_1_bi_ort ...'
-  call wall_time(wall0)
-
-  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
-
- !$OMP PARALLEL                     &
- !$OMP DEFAULT (NONE)               &
- !$OMP PRIVATE (i,j,k,m,l,integral) & 
- !$OMP SHARED (mo_num,three_e_5_idx_cycle_1_bi_ort)
- !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
-  do i = 1, mo_num
-    do k = 1, mo_num
-      do j = 1, mo_num
-        do l = 1, mo_num
-          do m = 1, mo_num
-            call give_integrals_3_body_bi_ort(m, l, k, j, i, m, integral)
-            three_e_5_idx_cycle_1_bi_ort(m,l,j,k,i) = -1.d0 * integral 
-          enddo
         enddo
       enddo
-    enddo
+      !$OMP END PARALLEL DO
-  enddo
- !$OMP END DO
- !$OMP END PARALLEL
 
-  call wall_time(wall1)
+      call dgemm('T','N', mo_num*mo_num, mo_num, n_points_final_grid, 1.d0,&
-  print *, ' wall time for three_e_5_idx_cycle_1_bi_ort', wall1 - wall0
+          orb_mat, n_points_final_grid,                              &
+          grad_mli, n_points_final_grid,  0.d0,                      &
+          tmp_mat, mo_num*mo_num)
 
-END_PROVIDER 
+      !$OMP PARALLEL PRIVATE(j,k,l)
-
+      !$OMP DO
-! ---
+      do k = 1, mo_num
-
+        do j = 1, mo_num
-BEGIN_PROVIDER [ double precision, three_e_5_idx_cycle_2_bi_ort, (mo_num, mo_num, mo_num, mo_num, mo_num)]
-
-  BEGIN_DOC
-  !
-  ! matrix element of the -L  three-body operator FOR THE FIRST CYCLIC PERMUTATION TERMS OF DOUBLE EXCITATIONS AND BI ORTHO MOs
-  !
-  ! three_e_5_idx_cycle_2_bi_ort(m,l,j,k,i) = <mlk|-L|imj> ::: notice that i is the RIGHT MO and k is the LEFT MO
-  !
-  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
-  !
-  END_DOC
-
-  implicit none
-  integer          :: i, j, k, m, l
-  double precision :: integral, wall1, wall0
-
-  three_e_5_idx_cycle_2_bi_ort = 0.d0
-  print *, ' Providing the three_e_5_idx_cycle_2_bi_ort ...'
-  call wall_time(wall0)
-
-  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
-
- !$OMP PARALLEL                     &
- !$OMP DEFAULT (NONE)               &
- !$OMP PRIVATE (i,j,k,m,l,integral) & 
- !$OMP SHARED (mo_num,three_e_5_idx_cycle_2_bi_ort)
- !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
-  do i = 1, mo_num
-    do k = 1, mo_num
-      do j = 1, mo_num
-        do m = 1, mo_num
           do l = 1, mo_num
-            call give_integrals_3_body_bi_ort(m, l, k, i, m, j, integral)
+            three_e_5_idx_direct_bi_ort(m,l,j,k,i) = three_e_5_idx_direct_bi_ort(m,l,j,k,i) - tmp_mat(l,j,k)
-            three_e_5_idx_cycle_2_bi_ort(m,l,j,k,i) = -1.d0 * integral 
           enddo
         enddo
       enddo
-    enddo
+      !$OMP END DO
-  enddo
+      !$OMP DO
- !$OMP END DO
- !$OMP END PARALLEL
-
-  call wall_time(wall1)
-  print *, ' wall time for three_e_5_idx_cycle_2_bi_ort', wall1 - wall0
-
-END_PROVIDER 
-
-! ---
-
-BEGIN_PROVIDER [ double precision, three_e_5_idx_exch23_bi_ort, (mo_num, mo_num, mo_num, mo_num, mo_num)]
-
-  BEGIN_DOC
-  !
-  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF DOUBLE EXCITATIONS AND BI ORTHO MOs
-  !
-  ! three_e_5_idx_exch23_bi_ort(m,l,j,k,i) = <mlk|-L|jmi> ::: notice that i is the RIGHT MO and k is the LEFT MO
-  !
-  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
-  !
-  END_DOC
-
-  implicit none
-  integer          :: i, j, k, m, l
-  double precision :: integral, wall1, wall0
-
-  three_e_5_idx_exch23_bi_ort = 0.d0
-  print *, ' Providing the three_e_5_idx_exch23_bi_ort ...'
-  call wall_time(wall0)
-
-  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
-
- !$OMP PARALLEL                     &
- !$OMP DEFAULT (NONE)               &
- !$OMP PRIVATE (i,j,k,m,l,integral) & 
- !$OMP SHARED (mo_num,three_e_5_idx_exch23_bi_ort)
- !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
-  do i = 1, mo_num
-    do k = 1, mo_num
       do j = 1, mo_num
         do l = 1, mo_num
-          do m = 1, mo_num
+          do k = 1, mo_num
-            call give_integrals_3_body_bi_ort(m, l, k, j, m, i, integral)
+            three_e_5_idx_direct_bi_ort(m,k,i,l,j) = three_e_5_idx_direct_bi_ort(m,k,i,l,j) - tmp_mat(l,j,k)
-            three_e_5_idx_exch23_bi_ort(m,l,j,k,i) = -1.d0 * integral 
           enddo
         enddo
       enddo
+      !$OMP END DO
+      !$OMP END PARALLEL
     enddo
-  enddo
- !$OMP END DO
- !$OMP END PARALLEL
 
-  call wall_time(wall1)
+    deallocate(grad_mli)
-  print *, ' wall time for three_e_5_idx_exch23_bi_ort', wall1 - wall0
 
-END_PROVIDER 
+    allocate(lm_grad_ik(n_points_final_grid,3,mo_num,mo_num))
+    allocate(lk_grad_mi(n_points_final_grid,3,mo_num,mo_num))
 
-! ---
+    !$OMP PARALLEL DO PRIVATE (i,l,ipoint)
+    do i=1,mo_num
+      do l=1,mo_num
+        do ipoint=1, n_points_final_grid
 
-BEGIN_PROVIDER [ double precision, three_e_5_idx_exch13_bi_ort, (mo_num, mo_num, mo_num, mo_num, mo_num)]
+          lm_grad_ik(ipoint,1,l,i) = mos_l_in_r_array_transp(ipoint,m) * int2_grad1_u12_bimo_t(ipoint,1,l,i) * final_weight_at_r_vector(ipoint)
+          lm_grad_ik(ipoint,2,l,i) = mos_l_in_r_array_transp(ipoint,m) * int2_grad1_u12_bimo_t(ipoint,2,l,i) * final_weight_at_r_vector(ipoint)
+          lm_grad_ik(ipoint,3,l,i) = mos_l_in_r_array_transp(ipoint,m) * int2_grad1_u12_bimo_t(ipoint,3,l,i) * final_weight_at_r_vector(ipoint)
 
-  BEGIN_DOC
+          lk_grad_mi(ipoint,1,l,i) = mos_l_in_r_array_transp(ipoint,l) * int2_grad1_u12_bimo_t(ipoint,1,m,i) * final_weight_at_r_vector(ipoint)
-  !
+          lk_grad_mi(ipoint,2,l,i) = mos_l_in_r_array_transp(ipoint,l) * int2_grad1_u12_bimo_t(ipoint,2,m,i) * final_weight_at_r_vector(ipoint)
-  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF DOUBLE EXCITATIONS AND BI ORTHO MOs
+          lk_grad_mi(ipoint,3,l,i) = mos_l_in_r_array_transp(ipoint,l) * int2_grad1_u12_bimo_t(ipoint,3,m,i) * final_weight_at_r_vector(ipoint)
-  !
-  ! three_e_5_idx_exch13_bi_ort(m,l,j,k,i) = <mlk|-L|ijm> ::: notice that i is the RIGHT MO and k is the LEFT MO
-  !
-  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
-  !
-  END_DOC
 
-  implicit none
-  integer          :: i, j, k, m, l
-  double precision :: integral, wall1, wall0
-
-  three_e_5_idx_exch13_bi_ort = 0.d0
-  print *, ' Providing the three_e_5_idx_exch13_bi_ort ...'
-  call wall_time(wall0)
-
-  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
-
- !$OMP PARALLEL                     &
- !$OMP DEFAULT (NONE)               &
- !$OMP PRIVATE (i,j,k,m,l,integral) & 
- !$OMP SHARED (mo_num,three_e_5_idx_exch13_bi_ort)
- !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
-  do i = 1, mo_num
-    do k = 1, mo_num
-      do j = 1, mo_num
-        do l = 1, mo_num
-          do m = 1, mo_num
-            call give_integrals_3_body_bi_ort(m, l, k, i, j, m, integral)
-            three_e_5_idx_exch13_bi_ort(m,l,j,k,i) = -1.d0 * integral 
-          enddo
         enddo
       enddo
     enddo
-  enddo
+    !$OMP END PARALLEL DO
- !$OMP END DO
- !$OMP END PARALLEL
 
-  call wall_time(wall1)
+    allocate(rm_grad_ik(n_points_final_grid,3,mo_num))
-  print *, ' wall time for three_e_5_idx_exch13_bi_ort', wall1 - wall0
+    allocate(rk_grad_im(n_points_final_grid,3,mo_num))
 
-END_PROVIDER 
+    do i=1,mo_num
+      !$OMP PARALLEL DO PRIVATE (l,ipoint)
+      do l=1,mo_num
+        do ipoint=1, n_points_final_grid
 
-! ---
+          rm_grad_ik(ipoint,1,l) = mos_r_in_r_array_transp(ipoint,m) * int2_grad1_u12_bimo_t(ipoint,1,l,i)
+          rm_grad_ik(ipoint,2,l) = mos_r_in_r_array_transp(ipoint,m) * int2_grad1_u12_bimo_t(ipoint,2,l,i)
+          rm_grad_ik(ipoint,3,l) = mos_r_in_r_array_transp(ipoint,m) * int2_grad1_u12_bimo_t(ipoint,3,l,i)
 
-BEGIN_PROVIDER [ double precision, three_e_5_idx_exch12_bi_ort, (mo_num, mo_num, mo_num, mo_num, mo_num)]
+          rk_grad_im(ipoint,1,l) = mos_r_in_r_array_transp(ipoint,l) * int2_grad1_u12_bimo_t(ipoint,1,i,m)
+          rk_grad_im(ipoint,2,l) = mos_r_in_r_array_transp(ipoint,l) * int2_grad1_u12_bimo_t(ipoint,2,i,m)
+          rk_grad_im(ipoint,3,l) = mos_r_in_r_array_transp(ipoint,l) * int2_grad1_u12_bimo_t(ipoint,3,i,m)
 
-  BEGIN_DOC
+        enddo
-  !
+      enddo
-  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF DOUBLE EXCITATIONS AND BI ORTHO MOs
+      !$OMP END PARALLEL DO
-  !
-  ! three_e_5_idx_exch12_bi_ort(m,l,j,k,i) = <mlk|-L|mij> ::: notice that i is the RIGHT MO and k is the LEFT MO
-  !
-  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
-  !
-  END_DOC
 
-  implicit none
+      call dgemm('T','N', mo_num*mo_num, mo_num, 3*n_points_final_grid, 1.d0,&
-  integer          :: i, j, k, m, l
+          lm_grad_ik, 3*n_points_final_grid,                         &
-  double precision :: integral, wall1, wall0
+          rm_grad_ik, 3*n_points_final_grid,  0.d0,                  &
+          tmp_mat, mo_num*mo_num)
 
-  three_e_5_idx_exch12_bi_ort = 0.d0
+      !$OMP PARALLEL DO PRIVATE(j,k,l)
-  print *, ' Providing the three_e_5_idx_exch12_bi_ort ...'
+      do k = 1, mo_num
-  call wall_time(wall0)
+        do j = 1, mo_num
-
+          do l = 1, mo_num
-  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+            three_e_5_idx_direct_bi_ort(m,l,j,k,i) = three_e_5_idx_direct_bi_ort(m,l,j,k,i) - tmp_mat(l,j,k)
-
- !$OMP PARALLEL                     &
- !$OMP DEFAULT (NONE)               &
- !$OMP PRIVATE (i,j,k,m,l,integral) & 
- !$OMP SHARED (mo_num,three_e_5_idx_exch12_bi_ort)
- !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
-  do i = 1, mo_num
-    do k = 1, mo_num
-      do j = 1, mo_num
-        do l = 1, mo_num
-          do m = 1, mo_num
-            call give_integrals_3_body_bi_ort(m, l, k, m, i, j, integral)
-            three_e_5_idx_exch12_bi_ort(m,l,j,k,i) = -1.d0 * integral 
           enddo
         enddo
       enddo
+      !$OMP END PARALLEL DO
+
+      call dgemm('T','N', mo_num*mo_num, mo_num, 3*n_points_final_grid, 1.d0,&
+          lm_grad_ik, 3*n_points_final_grid,                         &
+          rk_grad_im, 3*n_points_final_grid,  0.d0,                  &
+          tmp_mat, mo_num*mo_num)
+
+      !$OMP PARALLEL DO PRIVATE(j,k,l)
+      do k = 1, mo_num
+        do j = 1, mo_num
+          do l = 1, mo_num
+            three_e_5_idx_cycle_1_bi_ort(m,l,j,i,k) = three_e_5_idx_cycle_1_bi_ort(m,l,j,i,k) - tmp_mat(l,k,j)
+            three_e_5_idx_cycle_2_bi_ort(m,i,j,k,l) = three_e_5_idx_cycle_2_bi_ort(m,i,j,k,l) - tmp_mat(k,j,l)
+            three_e_5_idx_exch23_bi_ort (m,i,j,k,l) = three_e_5_idx_exch23_bi_ort (m,i,j,k,l) - tmp_mat(k,l,j)
+            three_e_5_idx_exch13_bi_ort (m,l,j,i,k) = three_e_5_idx_exch13_bi_ort (m,l,j,i,k) - tmp_mat(l,j,k)
+          enddo
+        enddo
+      enddo
+      !$OMP END PARALLEL DO
+
+
+      call dgemm('T','N', mo_num*mo_num, mo_num, 3*n_points_final_grid, 1.d0,&
+          lk_grad_mi, 3*n_points_final_grid,                         &
+          rm_grad_ik, 3*n_points_final_grid,  0.d0,                  &
+          tmp_mat, mo_num*mo_num)
+
+      !$OMP PARALLEL DO PRIVATE(j,k,l)
+      do k = 1, mo_num
+        do j = 1, mo_num
+          do l = 1, mo_num
+            three_e_5_idx_cycle_1_bi_ort(m,l,j,k,i) = three_e_5_idx_cycle_1_bi_ort(m,l,j,k,i) - tmp_mat(k,j,l)
+            three_e_5_idx_cycle_2_bi_ort(m,l,i,k,j) = three_e_5_idx_cycle_2_bi_ort(m,l,i,k,j) - tmp_mat(l,j,k)
+            three_e_5_idx_exch23_bi_ort (m,l,j,k,i) = three_e_5_idx_exch23_bi_ort (m,l,j,k,i) - tmp_mat(l,j,k)
+            three_e_5_idx_exch13_bi_ort (m,l,i,k,j) = three_e_5_idx_exch13_bi_ort (m,l,i,k,j) - tmp_mat(k,j,l)
+          enddo
+        enddo
+      enddo
+      !$OMP END PARALLEL DO
+
+      call dgemm('T','N', mo_num*mo_num, mo_num, 3*n_points_final_grid, 1.d0,&
+          lk_grad_mi, 3*n_points_final_grid,                         &
+          rk_grad_im, 3*n_points_final_grid,  0.d0,                  &
+          tmp_mat, mo_num*mo_num)
+
+      !$OMP PARALLEL DO PRIVATE(j,k,l)
+      do k = 1, mo_num
+        do j = 1, mo_num
+          do l = 1, mo_num
+            three_e_5_idx_cycle_1_bi_ort(m,l,j,i,k) = three_e_5_idx_cycle_1_bi_ort(m,l,j,i,k) - tmp_mat(l,j,k)
+            three_e_5_idx_cycle_2_bi_ort(m,i,j,k,l) = three_e_5_idx_cycle_2_bi_ort(m,i,j,k,l) - tmp_mat(k,l,j)
+            three_e_5_idx_exch23_bi_ort (m,i,j,k,l) = three_e_5_idx_exch23_bi_ort (m,i,j,k,l) - tmp_mat(k,j,l)
+            three_e_5_idx_exch13_bi_ort (m,l,j,i,k) = three_e_5_idx_exch13_bi_ort (m,l,j,i,k) - tmp_mat(l,k,j)
+          enddo
+        enddo
+      enddo
+      !$OMP END PARALLEL DO
+
     enddo
+
+    deallocate(rm_grad_ik)
+    deallocate(rk_grad_im)
+    deallocate(lk_grad_mi)
+    deallocate(lm_grad_ik)
+
   enddo
- !$OMP END DO
+
- !$OMP END PARALLEL
+  deallocate(tmp_mat)
+
+  deallocate(orb_mat)
 
   call wall_time(wall1)
-  print *, ' wall time for three_e_5_idx_exch12_bi_ort', wall1 - wall0
+  print *, ' wall time for three_e_5_idx_bi_ort', wall1 - wall0
+  call print_memory_usage()
 
 END_PROVIDER
 
-! ---
-

src/bi_ort_ints/three_body_ijmkl_old.irp.f

@@ -0,0 +1,295 @@
+
+! ---
+
+BEGIN_PROVIDER [ double precision, three_e_5_idx_direct_bi_ort_old, (mo_num, mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF DOUBLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_5_idx_direct_bi_ort_old(m,l,j,k,i) = <mlk|-L|mji> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  END_DOC
+
+  implicit none
+  integer          :: i, j, k, m, l
+  double precision :: integral, wall1, wall0
+
+  three_e_5_idx_direct_bi_ort_old = 0.d0
+  print *, ' Providing the three_e_5_idx_direct_bi_ort_old ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+ !$OMP PARALLEL                     &
+ !$OMP DEFAULT (NONE)               &
+ !$OMP PRIVATE (i,j,k,m,l,integral) &
+ !$OMP SHARED (mo_num,three_e_5_idx_direct_bi_ort_old)
+ !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do l = 1, mo_num
+          do m = 1, mo_num
+            call give_integrals_3_body_bi_ort(m, l, k, m, j, i, integral)
+            three_e_5_idx_direct_bi_ort_old(m,l,j,k,i) = -1.d0 * integral
+          enddo
+        enddo
+      enddo
+    enddo
+  enddo
+ !$OMP END DO
+ !$OMP END PARALLEL
+
+  call wall_time(wall1)
+  print *, ' wall time for three_e_5_idx_direct_bi_ort_old', wall1 - wall0
+
+END_PROVIDER
+
+! ---
+
+BEGIN_PROVIDER [ double precision, three_e_5_idx_cycle_1_bi_ort_old, (mo_num, mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE FIRST CYCLIC PERMUTATION TERMS OF DOUBLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_5_idx_cycle_1_bi_ort_old(m,l,j,k,i) = <mlk|-L|jim> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  !
+  END_DOC
+
+  implicit none
+  integer          :: i, j, k, m, l
+  double precision :: integral, wall1, wall0
+
+  three_e_5_idx_cycle_1_bi_ort_old = 0.d0
+  print *, ' Providing the three_e_5_idx_cycle_1_bi_ort_old ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+ !$OMP PARALLEL                     &
+ !$OMP DEFAULT (NONE)               &
+ !$OMP PRIVATE (i,j,k,m,l,integral) &
+ !$OMP SHARED (mo_num,three_e_5_idx_cycle_1_bi_ort_old)
+ !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do l = 1, mo_num
+          do m = 1, mo_num
+            call give_integrals_3_body_bi_ort(m, l, k, j, i, m, integral)
+            three_e_5_idx_cycle_1_bi_ort_old(m,l,j,k,i) = -1.d0 * integral
+          enddo
+        enddo
+      enddo
+    enddo
+  enddo
+ !$OMP END DO
+ !$OMP END PARALLEL
+
+  call wall_time(wall1)
+  print *, ' wall time for three_e_5_idx_cycle_1_bi_ort_old', wall1 - wall0
+
+END_PROVIDER
+
+! ---
+
+BEGIN_PROVIDER [ double precision, three_e_5_idx_cycle_2_bi_ort_old, (mo_num, mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE FIRST CYCLIC PERMUTATION TERMS OF DOUBLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_5_idx_cycle_2_bi_ort_old(m,l,j,k,i) = <mlk|-L|imj> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  !
+  END_DOC
+
+  implicit none
+  integer          :: i, j, k, m, l
+  double precision :: integral, wall1, wall0
+
+  three_e_5_idx_cycle_2_bi_ort_old = 0.d0
+  print *, ' Providing the three_e_5_idx_cycle_2_bi_ort_old ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+ !$OMP PARALLEL                     &
+ !$OMP DEFAULT (NONE)               &
+ !$OMP PRIVATE (i,j,k,m,l,integral) &
+ !$OMP SHARED (mo_num,three_e_5_idx_cycle_2_bi_ort_old)
+ !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do m = 1, mo_num
+          do l = 1, mo_num
+            call give_integrals_3_body_bi_ort(m, l, k, i, m, j, integral)
+            three_e_5_idx_cycle_2_bi_ort_old(m,l,j,k,i) = -1.d0 * integral
+          enddo
+        enddo
+      enddo
+    enddo
+  enddo
+ !$OMP END DO
+ !$OMP END PARALLEL
+
+  call wall_time(wall1)
+  print *, ' wall time for three_e_5_idx_cycle_2_bi_ort_old', wall1 - wall0
+
+END_PROVIDER
+
+! ---
+
+BEGIN_PROVIDER [ double precision, three_e_5_idx_exch23_bi_ort_old, (mo_num, mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF DOUBLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_5_idx_exch23_bi_ort_old(m,l,j,k,i) = <mlk|-L|jmi> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  !
+  END_DOC
+
+  implicit none
+  integer          :: i, j, k, m, l
+  double precision :: integral, wall1, wall0
+
+  three_e_5_idx_exch23_bi_ort_old = 0.d0
+  print *, ' Providing the three_e_5_idx_exch23_bi_ort_old ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+ !$OMP PARALLEL                     &
+ !$OMP DEFAULT (NONE)               &
+ !$OMP PRIVATE (i,j,k,m,l,integral) &
+ !$OMP SHARED (mo_num,three_e_5_idx_exch23_bi_ort_old)
+ !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do l = 1, mo_num
+          do m = 1, mo_num
+            call give_integrals_3_body_bi_ort(m, l, k, j, m, i, integral)
+            three_e_5_idx_exch23_bi_ort_old(m,l,j,k,i) = -1.d0 * integral
+          enddo
+        enddo
+      enddo
+    enddo
+  enddo
+ !$OMP END DO
+ !$OMP END PARALLEL
+
+  call wall_time(wall1)
+  print *, ' wall time for three_e_5_idx_exch23_bi_ort_old', wall1 - wall0
+
+END_PROVIDER
+
+! ---
+
+BEGIN_PROVIDER [ double precision, three_e_5_idx_exch13_bi_ort_old, (mo_num, mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF DOUBLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_5_idx_exch13_bi_ort_old(m,l,j,k,i) = <mlk|-L|ijm> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  !
+  END_DOC
+
+  implicit none
+  integer          :: i, j, k, m, l
+  double precision :: integral, wall1, wall0
+
+  three_e_5_idx_exch13_bi_ort_old = 0.d0
+  print *, ' Providing the three_e_5_idx_exch13_bi_ort_old ...'
+  call wall_time(wall0)
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+
+ !$OMP PARALLEL                     &
+ !$OMP DEFAULT (NONE)               &
+ !$OMP PRIVATE (i,j,k,m,l,integral) &
+ !$OMP SHARED (mo_num,three_e_5_idx_exch13_bi_ort_old)
+ !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do l = 1, mo_num
+          do m = 1, mo_num
+            call give_integrals_3_body_bi_ort(m, l, k, i, j, m, integral)
+            three_e_5_idx_exch13_bi_ort_old(m,l,j,k,i) = -1.d0 * integral
+          enddo
+        enddo
+      enddo
+    enddo
+  enddo
+ !$OMP END DO
+ !$OMP END PARALLEL
+
+  call wall_time(wall1)
+  print *, ' wall time for three_e_5_idx_exch13_bi_ort_old', wall1 - wall0
+
+END_PROVIDER
+
+! ---
+
+BEGIN_PROVIDER [ double precision, three_e_5_idx_exch12_bi_ort_old, (mo_num, mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  !
+  ! matrix element of the -L  three-body operator FOR THE DIRECT TERMS OF DOUBLE EXCITATIONS AND BI ORTHO MOs
+  !
+  ! three_e_5_idx_exch12_bi_ort_old(m,l,j,k,i) = <mlk|-L|mij> ::: notice that i is the RIGHT MO and k is the LEFT MO
+  !
+  ! notice the -1 sign: in this way three_e_3_idx_direct_bi_ort can be directly used to compute Slater rules with a + sign
+  !
+  END_DOC
+
+  implicit none
+  integer          :: i, j, k, m, l
+  double precision :: integral, wall1, wall0
+
+  provide mos_r_in_r_array_transp mos_l_in_r_array_transp
+  PROVIDE mo_l_coef mo_r_coef int2_grad1_u12_bimo_t
+
+  three_e_5_idx_exch12_bi_ort_old = 0.d0
+  print *, ' Providing the three_e_5_idx_exch12_bi_ort_old ...'
+  call wall_time(wall0)
+
+ !$OMP PARALLEL                     &
+ !$OMP DEFAULT (NONE)               &
+ !$OMP PRIVATE (i,j,k,m,l,integral) &
+ !$OMP SHARED (mo_num,three_e_5_idx_exch12_bi_ort_old)
+ !$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
+  do i = 1, mo_num
+    do k = 1, mo_num
+      do j = 1, mo_num
+        do l = 1, mo_num
+          do m = 1, mo_num
+            call give_integrals_3_body_bi_ort(m, l, k, m, i, j, integral)
+            three_e_5_idx_exch12_bi_ort_old(m,l,j,k,i) = -1.d0 * integral
+          enddo
+        enddo
+      enddo
+    enddo
+  enddo
+ !$OMP END DO
+ !$OMP END PARALLEL
+
+  call wall_time(wall1)
+  print *, ' wall time for three_e_5_idx_exch12_bi_ort_old', wall1 - wall0
+
+END_PROVIDER
+

src/bi_ort_ints/three_body_ints_bi_ort.irp.f

@@ -57,6 +57,7 @@ BEGIN_PROVIDER [ double precision, three_body_ints_bi_ort, (mo_num, mo_num, mo_n
 
   call wall_time(wall1)
   print *, ' wall time for three_body_ints_bi_ort', wall1 - wall0
+  call print_memory_usage()
 ! if(write_three_body_ints_bi_ort)then
 !  print*,'Writing three_body_ints_bi_ort on disk ...'
 !  call write_array_6_index_tensor(mo_num,three_body_ints_bi_ort,name_file)
@@ -79,28 +80,31 @@ subroutine give_integrals_3_body_bi_ort(n, l, k, m, j, i, integral)
   integer,          intent(in)  :: n, l, k, m, j, i
   double precision, intent(out) :: integral
   integer                       :: ipoint
-  double precision              :: weight
+  double precision              :: weight, tmp
 
   PROVIDE mo_l_coef mo_r_coef
   PROVIDE int2_grad1_u12_bimo_t
 
   integral = 0.d0
+  ! (n, l, k, m, j, i)
   do ipoint = 1, n_points_final_grid
-    weight = final_weight_at_r_vector(ipoint)                                                                          
 
-    integral += weight * mos_l_in_r_array_transp(ipoint,k) * mos_r_in_r_array_transp(ipoint,i) & 
+    tmp =     mos_l_in_r_array_transp(ipoint,k) * mos_r_in_r_array_transp(ipoint,i) &
               * ( int2_grad1_u12_bimo_t(ipoint,1,n,m) * int2_grad1_u12_bimo_t(ipoint,1,l,j)    &
                 + int2_grad1_u12_bimo_t(ipoint,2,n,m) * int2_grad1_u12_bimo_t(ipoint,2,l,j)    &
                 + int2_grad1_u12_bimo_t(ipoint,3,n,m) * int2_grad1_u12_bimo_t(ipoint,3,l,j) )
-    integral += weight * mos_l_in_r_array_transp(ipoint,l) * mos_r_in_r_array_transp(ipoint,j) & 
+
+    tmp = tmp + mos_l_in_r_array_transp(ipoint,l) * mos_r_in_r_array_transp(ipoint,j) &
               * ( int2_grad1_u12_bimo_t(ipoint,1,n,m) * int2_grad1_u12_bimo_t(ipoint,1,k,i)    &
                 + int2_grad1_u12_bimo_t(ipoint,2,n,m) * int2_grad1_u12_bimo_t(ipoint,2,k,i)    &
                 + int2_grad1_u12_bimo_t(ipoint,3,n,m) * int2_grad1_u12_bimo_t(ipoint,3,k,i) )
-    integral += weight * mos_l_in_r_array_transp(ipoint,n) * mos_r_in_r_array_transp(ipoint,m) &
+
+    tmp = tmp + mos_l_in_r_array_transp(ipoint,n) * mos_r_in_r_array_transp(ipoint,m) &
               * ( int2_grad1_u12_bimo_t(ipoint,1,l,j) * int2_grad1_u12_bimo_t(ipoint,1,k,i)    &
                 + int2_grad1_u12_bimo_t(ipoint,2,l,j) * int2_grad1_u12_bimo_t(ipoint,2,k,i)    &
                 + int2_grad1_u12_bimo_t(ipoint,3,l,j) * int2_grad1_u12_bimo_t(ipoint,3,k,i) )
 
+    integral = integral + tmp * final_weight_at_r_vector(ipoint)
   enddo
 
 end subroutine give_integrals_3_body_bi_ort

src/bi_ortho_mos/bi_ort_mos_in_r.irp.f

@@ -116,7 +116,7 @@ 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_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

src/ccsd/EZFIO.cfg

@@ -0,0 +1,11 @@
+[energy]
+type: double precision
+doc: CCSD energy
+interface: ezfio
+
+[energy_t]
+type: double precision
+doc: CCSD(T) energy
+interface: ezfio
+
+

src/ccsd/ccsd_space_orb_sub.irp.f

@@ -16,20 +16,16 @@ subroutine run_ccsd_space_orb
   double precision, allocatable :: all_err(:,:), all_t(:,:)
   integer, allocatable          :: list_occ(:), list_vir(:)
   integer(bit_kind)             :: det(N_int,2)
-  integer                       :: nO, nV, nOa, nOb, nVa, nVb, n_spin(4)
+  integer                       :: nO, nV, nOa, nVa
 
-  PROVIDE mo_two_e_integrals_in_map
+!  PROVIDE mo_two_e_integrals_in_map
 
   det = psi_det(:,:,cc_ref)
   print*,'Reference determinant:'
   call print_det(det,N_int)
 
-  ! Extract number of occ/vir alpha/beta spin orbitals
+  nOa = cc_nOa
-  !call extract_n_spin(det,n_spin)
+  nVa = cc_nVa
-  nOa = cc_nOa !n_spin(1)
-  nOb = cc_nOb !n_spin(2)
-  nVa = cc_nVa !n_spin(3)
-  nVb = cc_nVb !n_spin(4)
 
   ! Check that the reference is a closed shell determinant
   if (cc_ref_is_open_shell) then
@@ -109,7 +105,7 @@ subroutine run_ccsd_space_orb
       call update_t1(nO,nV,cc_space_f_o,cc_space_f_v,r1,t1)
       call update_t2(nO,nV,cc_space_f_o,cc_space_f_v,r2,t2)
     else
-      print*,'Unkonw cc_method_method: '//cc_update_method
+      print*,'Unkown cc_method_method: '//cc_update_method
     endif
 
     call update_tau_space(nO,nV,t1,t2,tau)
@@ -139,8 +135,11 @@ subroutine run_ccsd_space_orb
   write(*,'(A15,1pE10.2,A3)')' Conv        = ', max_r
   print*,''
 
-  call write_t1(nO,nV,t1)
+  if (write_amplitudes) then
-  call write_t2(nO,nV,t2)
+    call write_t1(nO,nV,t1)
+    call write_t2(nO,nV,t2)
+    call ezfio_set_utils_cc_io_amplitudes('Read')
+  endif
 
   ! Deallocation
   if (cc_update_method == 'diis') then
@@ -151,6 +150,7 @@ subroutine run_ccsd_space_orb
 
   ! CCSD(T)
   double precision :: e_t
+  e_t = 0.d0
 
   if (cc_par_t .and. elec_alpha_num + elec_beta_num > 2) then
 
@@ -169,8 +169,13 @@ subroutine run_ccsd_space_orb
     ! New
     print*,'Computing (T) correction...'
     call wall_time(ta)
-    call ccsd_par_t_space_v2(nO,nV,t1,t2,cc_space_f_o,cc_space_f_v &
+!    call ccsd_par_t_space_v3(nO,nV,t1,t2,cc_space_f_o,cc_space_f_v &
+!         ,cc_space_v_vvvo,cc_space_v_vvoo,cc_space_v_vooo,e_t)
+
+    e_t = uncorr_energy + energy ! For print in next call
+    call ccsd_par_t_space_stoch(nO,nV,t1,t2,cc_space_f_o,cc_space_f_v &
          ,cc_space_v_vvvo,cc_space_v_vvoo,cc_space_v_vooo,e_t)
+
     call wall_time(tb)
     print*,'Time: ',tb-ta, ' s'
 
@@ -181,8 +186,7 @@ subroutine run_ccsd_space_orb
     print*,''
   endif
 
-  print*,'Reference determinant:'
+  call save_energy(uncorr_energy + energy, e_t)
-  call print_det(det,N_int)
 
   deallocate(t1,t2)
 
@@ -211,8 +215,8 @@ subroutine ccsd_energy_space(nO,nV,tau,t1,energy)
   !$omp default(none)
   e = 0d0
   !$omp do
-  do i = 1, nO
+  do a = 1, nV
-    do a = 1, nV
+    do i = 1, nO
       e = e + 2d0 * cc_space_f_vo(a,i) * t1(i,a)
     enddo
   enddo
@@ -255,7 +259,7 @@ subroutine update_tau_space(nO,nV,t1,t2,tau)
   !$OMP SHARED(nO,nV,tau,t2,t1) &
   !$OMP PRIVATE(i,j,a,b) &
   !$OMP DEFAULT(NONE)
-  !$OMP DO collapse(3)
+  !$OMP DO
   do b = 1, nV
     do a = 1, nV
       do j = 1, nO
@@ -373,7 +377,7 @@ subroutine compute_r1_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r1,max_r1)
   !$omp shared(nO,nV,X_voov,t2,t1) &
   !$omp private(u,beta,i,a) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do beta = 1, nV
     do u = 1, nO
       do i = 1, nO
@@ -412,7 +416,7 @@ subroutine compute_r1_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r1,max_r1)
   !$omp shared(nO,nV,cc_space_v_ovov,cc_space_v_voov,X_ovov) &
   !$omp private(u,beta,i,a) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do beta = 1, nV
     do u = 1, nO
       do a = 1, nv
@@ -452,7 +456,7 @@ subroutine compute_r1_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r1,max_r1)
   !$omp shared(nO,nV,cc_space_v_vvov,W_vvov,T_vvoo,tau) &
   !$omp private(b,beta,i,a) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do beta = 1, nV
     do i = 1, nO
       do b = 1, nV
@@ -464,11 +468,11 @@ subroutine compute_r1_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r1,max_r1)
   enddo
   !$omp end do nowait
 
-  !$omp do collapse(3)
+  !$omp do
-  do i = 1, nO
+  do u = 1, nO
-    do b = 1, nV
+    do i = 1, nO
-      do a = 1, nV
+      do b = 1, nV
-        do u = 1, nO
+        do a = 1, nV
           T_vvoo(a,b,i,u) = tau(i,u,a,b)
         enddo
       enddo
@@ -504,8 +508,8 @@ subroutine compute_r1_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r1,max_r1)
   !$omp shared(nO,nV,cc_space_v_vooo,W_oovo) &
   !$omp private(u,a,i,j) &
   !$omp default(none)
-  !$omp do collapse(3)
   do u = 1, nO
+    !$omp do
     do a = 1, nV
       do j = 1, nO
         do i = 1, nO
@@ -513,8 +517,8 @@ subroutine compute_r1_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r1,max_r1)
         enddo
       enddo
     enddo
+    !$omp end do nowait
   enddo
-  !$omp end do
   !$omp end parallel
 
   call dgemm('T','N', nO, nV, nO*nO*nV, &
@@ -527,9 +531,7 @@ subroutine compute_r1_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r1,max_r1)
   max_r1 = 0d0
   do a = 1, nV
     do i = 1, nO
-      if (dabs(r1(i,a)) > max_r1) then
+      max_r1 = max(dabs(r1(i,a)), max_r1)
-        max_r1 = dabs(r1(i,a))
-      endif
     enddo
   enddo
 
@@ -657,7 +659,7 @@ subroutine compute_H_vv(nO,nV,t1,t2,tau,H_vv)
   ! H_vv(a,beta) = H_vv(a,beta) - cc_space_w_vvoo(a,b,i,j) * tau(i,j,beta,b)
   ! H_vv(a,beta) = H_vv(a,beta) - cc_space_w_vvoo(a,b,i,j) * tmp_tau(b,i,j,beta)
 
-  !$omp do collapse(3)
+  !$omp do
   do beta = 1, nV
     do j = 1, nO
       do i = 1, nO
@@ -727,7 +729,7 @@ subroutine compute_H_vo(nO,nV,t1,t2,H_vo)
   ! H_vo(a,i) = H_vo(a,i) + cc_space_w_vvoo(a,b,i,j) * t1(j,b)
   ! H_vo(a,i) = H_vo(a,i) + w(a,i,j,b) * t1(j,b)
 
-  !$omp do collapse(3)
+  !$omp do
   do b = 1, nV
     do j = 1, nO
       do i = 1, nO
@@ -765,7 +767,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
 
   ! internal
   double precision, allocatable :: g_occ(:,:), g_vir(:,:), J1(:,:,:,:), K1(:,:,:,:)
-  double precision, allocatable :: A1(:,:,:,:), B1(:,:,:,:)
+  double precision, allocatable :: A1(:,:,:,:), B1_gam(:,:,:)
   integer                       :: u,v,i,j,beta,gam,a,b
 
   allocate(g_occ(nO,nO), g_vir(nV,nV))
@@ -787,7 +789,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,r2,cc_space_v_oovv) &
   !$omp private(u,v,gam,beta) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do v = 1, nO
@@ -835,13 +837,18 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !  enddo
   !enddo
 
-  allocate(B1(nV,nV,nV,nV))
+!  allocate(B1(nV,nV,nV,nV))
-  call compute_B1(nO,nV,t1,t2,B1)
+!  call compute_B1(nO,nV,t1,t2,B1)
-  call dgemm('N','N',nO*nO,nV*nV,nV*nV, &
+  allocate(B1_gam(nV,nV,nV))
-             1d0, tau, size(tau,1) * size(tau,2), &
+  do gam=1,nV
-                  B1 , size(B1,1) * size(B1,2), &
+    call compute_B1_gam(nO,nV,t1,t2,B1_gam,gam)
-             1d0, r2, size(r2,1) * size(r2,2))
+    call dgemm('N','N',nO*nO,nV,nV*nV, &
-  deallocate(B1)
+                1d0, tau, size(tau,1) * size(tau,2), &
+                     B1_gam        , size(B1_gam,1) * size(B1_gam,2), &
+                1d0, r2(1,1,1,gam), size(r2,1) * size(r2,2))
+  enddo
+  deallocate(B1_gam)
+
 
   !do gam = 1, nV
   !  do beta = 1, nV
@@ -863,7 +870,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,t2,X_oovv) &
   !$omp private(u,v,gam,a) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do a = 1, nV
     do gam = 1, nV
       do v = 1, nO
@@ -885,7 +892,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,r2,Y_oovv) &
   !$omp private(u,v,gam,beta) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do v = 1, nO
@@ -921,7 +928,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,r2,X_oovv) &
   !$omp private(u,v,gam,beta) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do v = 1, nO
@@ -957,7 +964,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,X_vovv,cc_space_v_ovvv) &
   !$omp private(u,a,gam,beta) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do u = 1, nO
@@ -979,7 +986,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,r2,Y_oovv) &
   !$omp private(u,v,gam,beta) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do v = 1, nO
@@ -1014,8 +1021,8 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,X_vovo,cc_space_v_ovov) &
   !$omp private(u,v,gam,i) &
   !$omp default(none)
-  !$omp do collapse(3)
   do i = 1, nO
+    !$omp do
     do gam = 1, nV
       do u = 1, nO
         do a = 1, nV
@@ -1023,8 +1030,8 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
         enddo
       enddo
     enddo
+    !$omp end do nowait
   enddo
-  !$omp end do
   !$omp end parallel
 
   call dgemm('N','N',nV*nO*nV,nV,nO, &
@@ -1041,7 +1048,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,r2,X_oovv) &
   !$omp private(u,v,gam,beta) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do v = 1, nO
@@ -1079,7 +1086,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,r2,X_oovv) &
   !$omp private(u,v,gam,beta) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do v = 1, nO
@@ -1116,8 +1123,8 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,X_vovo,cc_space_v_ovvo) &
   !$omp private(a,v,gam,i) &
   !$omp default(none)
-  !$omp do collapse(3)
   do i = 1, nO
+    !$omp do
     do gam = 1, nV
       do v = 1, nO
         do a = 1, nV
@@ -1125,8 +1132,8 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
         enddo
       enddo
     enddo
+    !$omp end do nowait
   enddo
-  !$omp end do
   !$omp end parallel
 
   call dgemm('N','N',nO,nO*nV*nO,nV, &
@@ -1143,7 +1150,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,r2,X_oovv) &
   !$omp private(u,v,gam,beta) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do v = 1, nO
@@ -1182,19 +1189,19 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,X_ovvo,Y_voov,K1,J1,t2) &
   !$omp private(u,v,gam,beta,i,a) &
   !$omp default(none)
-  !$omp do collapse(3)
   do i = 1, nO
+    !$omp do
     do a = 1, nV
       do beta = 1, nV
         do u = 1, nO
-          X_ovvo(u,beta,a,i) = 0.5d0 * (2d0 * J1(u,a,beta,i) - K1(u,a,i,beta))
+          X_ovvo(u,beta,a,i) = (J1(u,a,beta,i) - 0.5d0 * K1(u,a,i,beta))
         enddo
       enddo
     enddo
+    !$omp end do nowait
   enddo
-  !$omp end do nowait
 
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do v = 1, nO
       do i = 1, nO
@@ -1216,7 +1223,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,r2,Z_ovov) &
   !$omp private(u,v,gam,beta) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do v = 1, nO
@@ -1252,7 +1259,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,r2,K1,X_ovov,Y_ovov,t2) &
   !$omp private(u,a,i,beta,gam) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do beta = 1, nV
     do u = 1, nO
       do a = 1, nV
@@ -1264,7 +1271,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   enddo
   !$omp end do nowait
 
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do v = 1, nO
       do a = 1, nV
@@ -1286,7 +1293,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,r2,Z_ovov) &
   !$omp private(u,v,gam,beta) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do v = 1, nO
@@ -1319,7 +1326,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,K1,X_ovov,Z_ovov,t2) &
   !$omp private(u,v,gam,beta,i,a) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do a = 1, nV
     do i = 1, nO
       do gam = 1, nV
@@ -1331,7 +1338,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   enddo
   !$omp end do nowait
 
-  !$omp do collapse(3)
+  !$omp do
   do beta = 1, nV
     do v = 1, nO
       do a = 1, nV
@@ -1353,7 +1360,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,r2,Z_ovov) &
   !$omp private(u,v,gam,beta) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do v = 1, nO
@@ -1373,7 +1380,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
   !$omp shared(nO,nV,r2) &
   !$omp private(i,j,a,b) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do b = 1, nV
     do a = 1, nV
       do j = 1, nO
@@ -1391,9 +1398,7 @@ subroutine compute_r2_space(nO,nV,t1,t2,tau,H_oo,H_vv,H_vo,r2,max_r2)
     do a = 1, nV
       do j = 1, nO
         do i = 1, nO
-          if (dabs(r2(i,j,a,b)) > max_r2) then
+          max_r2 = max(r2(i,j,a,b), max_r2)
-            max_r2 = dabs(r2(i,j,a,b))
-          endif
         enddo
       enddo
     enddo
@@ -1448,7 +1453,7 @@ subroutine compute_A1(nO,nV,t1,t2,tau,A1)
   !$omp shared(nO,nV,A1,cc_space_v_oooo,cc_space_v_ovoo,X_vooo) &
   !$omp private(u,v,i,j) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do collapse(2)
   do j = 1, nO
     do i = 1, nO
       do v = 1, nO
@@ -1462,7 +1467,7 @@ subroutine compute_A1(nO,nV,t1,t2,tau,A1)
 
   ! A1(u,v,i,j) += cc_space_v_ovoo(u,a,i,j) * t1(v,a) &
 
-  !$omp do collapse(3)
+  !$omp do collapse(2)
   do j = 1, nO
     do i = 1, nO
       do u = 1, nO
@@ -1484,7 +1489,7 @@ subroutine compute_A1(nO,nV,t1,t2,tau,A1)
   !$omp shared(nO,nV,A1,Y_oooo) &
   !$omp private(u,v,i,j) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do collapse(2)
   do j = 1, nO
     do i = 1, nO
       do v = 1, nO
@@ -1515,6 +1520,90 @@ end
 
 ! B1
 
+subroutine compute_B1_gam(nO,nV,t1,t2,B1,gam)
+
+  implicit none
+
+  integer, intent(in)           :: nO,nV,gam
+  double precision, intent(in)  :: t1(nO, nV)
+  double precision, intent(in)  :: t2(nO, nO, nV, nV)
+  double precision, intent(out) :: B1(nV, nV, nV)
+
+  integer :: a,tmp_a,b,k,l,c,d,tmp_c,tmp_d,i,j,u,v, beta
+
+!  do beta = 1, nV
+!    do b = 1, nV
+!      do a = 1, nV
+!        B1(a,b,beta) = cc_space_v_vvvv(a,b,beta,gam)
+!
+!        do i = 1, nO
+!          B1(a,b,beta) = B1(a,b,beta) &
+!          - cc_space_v_vvvo(a,b,beta,i) * t1(i,gam) &
+!          - cc_space_v_vvov(a,b,i,gam) * t1(i,beta)
+!        enddo
+!
+!      enddo
+!    enddo
+!  enddo
+
+  double precision, allocatable :: X_vvvo(:,:,:), Y_vvvv(:,:,:)
+  allocate(X_vvvo(nV,nV,nO), Y_vvvv(nV,nV,nV))
+!  ! B1(a,b,beta,gam) = cc_space_v_vvvv(a,b,beta,gam)
+  !$omp parallel &
+  !$omp shared(nO,nV,B1,cc_space_v_vvvv,cc_space_v_vvov,X_vvvo,gam) &
+  !$omp private(a,b,beta) &
+  !$omp default(none)
+  !$omp do
+    do beta = 1, nV
+      do b = 1, nV
+        do a = 1, nV
+          B1(a,b,beta) = cc_space_v_vvvv(a,b,beta,gam)
+        enddo
+      enddo
+    enddo
+  !$omp end do nowait
+  do i = 1, nO
+    !$omp do
+      do b = 1, nV
+        do a = 1, nV
+          X_vvvo(a,b,i) = cc_space_v_vvov(a,b,i,gam)
+        enddo
+      enddo
+    !$omp end do nowait
+  enddo
+  !$omp end parallel
+
+!  ! B1(a,b,beta) -= cc_space_v_vvvo(a,b,beta,i) * t1(i,gam) &
+  call dgemm('N','N', nV*nV*nV, 1, nO, &
+             -1d0, cc_space_v_vvvo, size(cc_space_v_vvvo,1) * size(cc_space_v_vvvo,2) * size(cc_space_v_vvvo,3), &
+                   t1(1,gam), size(t1,1), &
+              1d0, B1      , size(B1,1) * size(B1,2) * size(B1,3))
+
+  ! B1(a,b,beta,gam) -= cc_space_v_vvov(a,b,i,gam) * t1(i,beta)
+  call dgemm('N','N', nV*nV, nV, nO, &
+             -1d0, X_vvvo, size(X_vvvo,1) * size(X_vvvo,2), &
+                   t1    , size(t1,1), &
+              0d0, Y_vvvv, size(Y_vvvv,1) * size(Y_vvvv,2))
+
+  !$omp parallel &
+  !$omp shared(nV,B1,Y_vvvv,gam) &
+  !$omp private(a,b,beta) &
+  !$omp default(none)
+  !$omp do
+  do beta = 1, nV
+    do b = 1, nV
+      do a = 1, nV
+        B1(a,b,beta) = B1(a,b,beta) + Y_vvvv(a,b,beta)
+      enddo
+    enddo
+  enddo
+  !$omp end do
+  !$omp end parallel
+
+  deallocate(X_vvvo,Y_vvvv)
+
+end
+
 subroutine compute_B1(nO,nV,t1,t2,B1)
 
   implicit none
@@ -1553,7 +1642,7 @@ subroutine compute_B1(nO,nV,t1,t2,B1)
   !$omp shared(nO,nV,B1,cc_space_v_vvvv,cc_space_v_vvov,X_vvvo) &
   !$omp private(a,b,beta,gam) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do b = 1, nV
@@ -1564,8 +1653,8 @@ subroutine compute_B1(nO,nV,t1,t2,B1)
     enddo
   enddo
   !$omp end do nowait
-  !$omp do collapse(3)
   do i = 1, nO
+    !$omp do
     do gam = 1, nV
       do b = 1, nV
         do a = 1, nV
@@ -1573,8 +1662,8 @@ subroutine compute_B1(nO,nV,t1,t2,B1)
         enddo
       enddo
     enddo
+    !$omp end do nowait
   enddo
-  !$omp end do
   !$omp end parallel
 
   ! B1(a,b,beta,gam) -= cc_space_v_vvvo(a,b,beta,i) * t1(i,gam) &
@@ -1594,7 +1683,7 @@ subroutine compute_B1(nO,nV,t1,t2,B1)
   !$omp shared(nV,B1,Y_vvvv) &
   !$omp private(a,b,beta,gam) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do gam = 1, nV
     do beta = 1, nV
       do b = 1, nV
@@ -1658,7 +1747,7 @@ subroutine compute_g_occ(nO,nV,t1,t2,H_oo,g_occ)
   enddo
   !$omp end do
 
-  !$omp do collapse(1)
+  !$omp do
   do i = 1, nO
     do j = 1, nO
       do a = 1, nV
@@ -1720,7 +1809,7 @@ subroutine compute_g_vir(nO,nV,t1,t2,H_vv,g_vir)
   enddo
   !$omp end do
 
-  !$omp do collapse(1)
+  !$omp do
   do beta = 1, nV
     do i = 1, nO
       do b = 1, nV
@@ -1788,8 +1877,8 @@ subroutine compute_J1(nO,nV,t1,t2,v_ovvo,v_ovoo,v_vvvo,v_vvoo,J1)
   !$omp shared(nO,nV,J1,v_ovvo,v_ovoo,X_ovoo) &
   !$omp private(i,j,a,u,beta) &
   !$omp default(none)
-  !$omp do collapse(3)
   do i = 1, nO
+    !$omp do
     do beta = 1, nV
       do a = 1, nV
         do u = 1, nO
@@ -1797,10 +1886,10 @@ subroutine compute_J1(nO,nV,t1,t2,v_ovvo,v_ovoo,v_vvvo,v_vvoo,J1)
         enddo
       enddo
     enddo
+    !$omp end do nowait
   enddo
-  !$omp end do nowait
 
-  !$omp do collapse(3)
+  !$omp do collapse(2)
   do j = 1, nO
     do i = 1, nO
       do a = 1, nV
@@ -1822,8 +1911,8 @@ subroutine compute_J1(nO,nV,t1,t2,v_ovvo,v_ovoo,v_vvvo,v_vvoo,J1)
   !$omp shared(nO,nV,J1,Y_ovov) &
   !$omp private(i,beta,a,u) &
   !$omp default(none)
-  !$omp do collapse(3)
   do i = 1, nO
+    !$omp do
     do beta = 1, nV
       do a = 1, nV
         do u = 1, nO
@@ -1831,8 +1920,8 @@ subroutine compute_J1(nO,nV,t1,t2,v_ovvo,v_ovoo,v_vvvo,v_vvoo,J1)
         enddo
       enddo
     enddo
+    !$omp end do nowait
   enddo
-  !$omp end do
   !$omp end parallel
   deallocate(X_ovoo)
 
@@ -1849,7 +1938,7 @@ subroutine compute_J1(nO,nV,t1,t2,v_ovvo,v_ovoo,v_vvvo,v_vvoo,J1)
   !$omp shared(nO,nV,t2,t1,Y_ovov,X_voov,v_vvoo) &
   !$omp private(i,beta,a,u,b,j) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do b = 1, nV
     do j = 1, nO
       do beta = 1, nV
@@ -1861,7 +1950,7 @@ subroutine compute_J1(nO,nV,t1,t2,v_ovvo,v_ovoo,v_vvvo,v_vvoo,J1)
   enddo
   !$omp end do nowait
 
-  !$omp do collapse(3)
+  !$omp do
   do b = 1, nV
     do j = 1, nO
       do i = 1, nO
@@ -1886,8 +1975,8 @@ subroutine compute_J1(nO,nV,t1,t2,v_ovvo,v_ovoo,v_vvvo,v_vvoo,J1)
   !$omp shared(nO,nV,J1,Z_ovvo,t2,Y_vovo,v_vvoo,X_ovvo) &
   !$omp private(i,beta,a,u,j,b) &
   !$omp default(none)
-  !$omp do collapse(3)
   do i = 1, nO
+    !$omp do
     do beta = 1, nV
       do a = 1, nV
         do u = 1, nO
@@ -1895,12 +1984,12 @@ subroutine compute_J1(nO,nV,t1,t2,v_ovvo,v_ovoo,v_vvvo,v_vvoo,J1)
         enddo
       enddo
     enddo
+    !$omp end do nowait
   enddo
-  !$omp end do nowait
 
   !+ 0.5d0 * (2d0 * cc_space_v_vvoo(a,b,i,j) - cc_space_v_vvoo(b,a,i,j)) * t2(u,j,beta,b)
-  !$omp do collapse(3)
   do j = 1, nO
+    !$omp do
     do b = 1, nV
       do i = 1, nO
         do a = 1, nV
@@ -1908,11 +1997,11 @@ subroutine compute_J1(nO,nV,t1,t2,v_ovvo,v_ovoo,v_vvvo,v_vvoo,J1)
         enddo
       enddo
     enddo
+    !$omp end do nowait
   enddo
-  !$omp end do nowait
 
-  !$omp do collapse(3)
   do j = 1, nO
+    !$omp do
     do b = 1, nV
       do beta = 1, nV
         do u = 1, nO
@@ -1920,8 +2009,8 @@ subroutine compute_J1(nO,nV,t1,t2,v_ovvo,v_ovoo,v_vvvo,v_vvoo,J1)
         enddo
       enddo
     enddo
+    !$omp end do nowait
   enddo
-  !$omp end do
   !$omp end parallel
 
   call dgemm('N','T',nO*nV,nV*nO,nV*nO, &
@@ -1933,8 +2022,8 @@ subroutine compute_J1(nO,nV,t1,t2,v_ovvo,v_ovoo,v_vvvo,v_vvoo,J1)
   !$omp shared(nO,nV,J1,Z_ovvo) &
   !$omp private(i,beta,a,u) &
   !$omp default(none)
-  !$omp do collapse(3)
   do i = 1, nO
+    !$omp do
     do beta = 1, nV
       do a = 1, nV
         do u = 1, nO
@@ -1942,8 +2031,8 @@ subroutine compute_J1(nO,nV,t1,t2,v_ovvo,v_ovoo,v_vvvo,v_vvoo,J1)
         enddo
       enddo
     enddo
+    !$omp end do nowait
   enddo
-  !$omp end do
   !$omp end parallel
 
   deallocate(X_ovvo,Z_ovvo,Y_ovov)
@@ -2003,7 +2092,7 @@ subroutine compute_K1(nO,nV,t1,t2,v_ovoo,v_vvoo,v_ovov,v_vvov,K1)
   !$omp shared(nO,nV,K1,X,Y,v_vvoo,v_ovov,t1,t2) &
   !$omp private(i,beta,a,u,j,b) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
   do beta = 1, nV
     do i = 1, nO
       do a = 1, nV
@@ -2015,8 +2104,8 @@ subroutine compute_K1(nO,nV,t1,t2,v_ovoo,v_vvoo,v_ovov,v_vvov,K1)
   enddo
   !$omp end do nowait
 
-  !$omp do collapse(3)
   do i = 1, nO
+    !$omp do
     do a = 1, nV
       do j = 1, nO
         do b = 1, nV
@@ -2024,11 +2113,11 @@ subroutine compute_K1(nO,nV,t1,t2,v_ovoo,v_vvoo,v_ovov,v_vvov,K1)
         enddo
       enddo
     enddo
+    !$omp end do nowait
   enddo
-  !$omp end do nowait
 
-  !$omp do collapse(3)
   do j = 1, nO
+    !$omp do
     do b = 1, nV
       do beta = 1, nV
         do u = 1, nO
@@ -2036,8 +2125,8 @@ subroutine compute_K1(nO,nV,t1,t2,v_ovoo,v_vvoo,v_ovov,v_vvov,K1)
         enddo
       enddo
     enddo
+    !$omp end do
   enddo
-  !$omp end do
   !$omp end parallel
 
   call dgemm('N','N',nO*nV*nO,nV,nO, &
@@ -2060,7 +2149,7 @@ subroutine compute_K1(nO,nV,t1,t2,v_ovoo,v_vvoo,v_ovov,v_vvov,K1)
   !$omp shared(nO,nV,K1,Z) &
   !$omp private(i,beta,a,u) &
   !$omp default(none)
-  !$omp do collapse(3)
+  !$omp do
    do beta = 1, nV
     do i = 1, nO
       do a = 1, nV

src/ccsd/ccsd_spin_orb_sub.irp.f

@@ -269,8 +269,11 @@ subroutine run_ccsd_spin_orb
   write(*,'(A15,1pE10.2,A3)')' Conv        = ', max_r
   print*,''
 
-  call write_t1(nO,nV,t1)
+  if (write_amplitudes) then
-  call write_t2(nO,nV,t2)
+    call write_t1(nO,nV,t1)
+    call write_t2(nO,nV,t2)
+    call ezfio_set_utils_cc_io_amplitudes('Read')
+  endif
 
   ! Deallocate
   if (cc_update_method == 'diis') then
@@ -284,8 +287,9 @@ subroutine run_ccsd_spin_orb
   deallocate(v_ovoo,v_oovo)
   deallocate(v_ovvo,v_ovov,v_oovv)
   
+  double precision :: t_corr
+  t_corr = 0.d0
   if (cc_par_t .and. elec_alpha_num  +elec_beta_num > 2) then
-    double precision :: t_corr
     print*,'CCSD(T) calculation...'
     call wall_time(ta)
     !allocate(v_vvvo(nV,nV,nV,nO))
@@ -307,8 +311,8 @@ subroutine run_ccsd_spin_orb
     write(*,'(A15,F18.12,A3)') ' Correlation = ', energy + t_corr, ' Ha'
     print*,''
   endif
-  print*,'Reference determinant:'
+
-  call print_det(det,N_int)
+  call save_energy(uncorr_energy + energy, t_corr)
   
   deallocate(f_oo,f_ov,f_vv,f_o,f_v)
   deallocate(v_ooov,v_vvoo,t1,t2)

src/ccsd/ccsd_t_space_orb_abc.irp.f

@@ -10,51 +10,43 @@ subroutine ccsd_par_t_space_v3(nO,nV,t1,t2,f_o,f_v,v_vvvo,v_vvoo,v_vooo,energy)
   double precision, intent(in)  :: v_vvvo(nV,nV,nV,nO), v_vvoo(nV,nV,nO,nO), v_vooo(nV,nO,nO,nO)
   double precision, intent(out) :: energy
 
-  double precision, allocatable :: W(:,:,:,:,:,:)
+  double precision, allocatable :: X_vovv(:,:,:,:), X_ooov(:,:,:,:), X_oovv(:,:,:,:)
-  double precision, allocatable :: V(:,:,:,:,:,:)
+  double precision, allocatable :: T_voov(:,:,:,:), T_oovv(:,:,:,:)
-  double precision, allocatable :: W_abc(:,:,:), V_abc(:,:,:)
-  double precision, allocatable :: W_cab(:,:,:), W_cba(:,:,:)
-  double precision, allocatable :: W_bca(:,:,:), V_cba(:,:,:)
-  double precision, allocatable :: X_vvvo(:,:,:,:), X_ovoo(:,:,:,:), X_vvoo(:,:,:,:)
-  double precision, allocatable :: T_vvoo(:,:,:,:), T_ovvo(:,:,:,:), T_vo(:,:)
   integer                       :: i,j,k,l,a,b,c,d
-  double precision              :: e,ta,tb, delta, delta_abc
+  double precision              :: e,ta,tb
 
-  !allocate(W(nV,nV,nV,nO,nO,nO))
+  call set_multiple_levels_omp(.False.)
-  !allocate(V(nV,nV,nV,nO,nO,nO))
+
-  allocate(W_abc(nO,nO,nO), V_abc(nO,nO,nO), W_cab(nO,nO,nO))
+  allocate(X_vovv(nV,nO,nV,nV), X_ooov(nO,nO,nO,nV), X_oovv(nO,nO,nV,nV))
-  allocate(W_bca(nO,nO,nO), V_cba(nO,nO,nO), W_cba(nO,nO,nO))
+  allocate(T_voov(nV,nO,nO,nV),T_oovv(nO,nO,nV,nV))
-  allocate(X_vvvo(nV,nV,nV,nO), X_ovoo(nO,nV,nO,nO), X_vvoo(nV,nV,nO,nO))
-  allocate(T_vvoo(nV,nV,nO,nO), T_ovvo(nO,nV,nV,nO), T_vo(nV,nO))
 
-  ! Temporary arrays
   !$OMP PARALLEL &
-  !$OMP SHARED(nO,nV,T_vvoo,T_ovvo,T_vo,X_vvvo,X_ovoo,X_vvoo, &
+  !$OMP SHARED(nO,nV,T_voov,T_oovv,X_vovv,X_ooov,X_oovv, &
   !$OMP t1,t2,v_vvvo,v_vooo,v_vvoo) &
   !$OMP PRIVATE(a,b,c,d,i,j,k,l) &
   !$OMP DEFAULT(NONE)
 
   !v_vvvo(b,a,d,i) * t2(k,j,c,d) &
-  !X_vvvo(d,b,a,i) * T_vvoo(d,c,k,j)
+  !X_vovv(d,i,b,a,i) * T_voov(d,j,c,k)
 
-  !$OMP DO collapse(3)
+  !$OMP DO
-  do i = 1, nO
+  do a = 1, nV
-    do a = 1, nV
+    do b = 1, nV
-      do b = 1, nV
+      do i = 1, nO
         do d = 1, nV
-          X_vvvo(d,b,a,i) = v_vvvo(b,a,d,i)
+          X_vovv(d,i,b,a) = v_vvvo(b,a,d,i)
         enddo
       enddo
     enddo
   enddo
   !$OMP END DO nowait
 
-  !$OMP DO collapse(3)
+  !$OMP DO
-  do j = 1, nO
+  do c = 1, nV
-    do k = 1, nO
+    do j = 1, nO
-      do c = 1, nV
+      do k = 1, nO
         do d = 1, nV
-          T_vvoo(d,c,k,j) = t2(k,j,c,d)
+          T_voov(d,k,j,c) = t2(k,j,c,d)
         enddo
       enddo
     enddo
@@ -62,191 +54,399 @@ subroutine ccsd_par_t_space_v3(nO,nV,t1,t2,f_o,f_v,v_vvvo,v_vvoo,v_vooo,energy)
   !$OMP END DO nowait
 
   !v_vooo(c,j,k,l) * t2(i,l,a,b) &
-  !X_ovoo(l,c,j,k) * T_ovvo(l,a,b,i) &
+  !X_ooov(l,j,k,c) * T_oovv(l,i,a,b) &
 
-  !$OMP DO collapse(3)
+  !$OMP DO
-  do k = 1, nO
+  do c = 1, nV
-    do j = 1, nO
-      do c = 1, nV
-        do l = 1, nO
-           X_ovoo(l,c,j,k) = v_vooo(c,j,k,l)
-        enddo
-      enddo
-    enddo
-  enddo
-  !$OMP END DO nowait
-
-  !$OMP DO collapse(3)
-  do i = 1, nO
-    do b = 1, nV
-      do a = 1, nV
-        do l = 1, nO
-          T_ovvo(l,a,b,i) = t2(i,l,a,b)
-        enddo
-      enddo
-    enddo
-  enddo
-  !$OMP END DO nowait
-
-  !v_vvoo(b,c,j,k) * t1(i,a) &
-  !X_vvoo(b,c,k,j) * T1_vo(a,i) &
-
-  !$OMP DO collapse(3)
-  do j = 1, nO
     do k = 1, nO
-      do c = 1, nV
+      do j = 1, nO
-        do b = 1, nV
+        do l = 1, nO
-          X_vvoo(b,c,k,j) = v_vvoo(b,c,j,k)
+           X_ooov(l,j,k,c) = v_vooo(c,j,k,l)
         enddo
       enddo
     enddo
   enddo
   !$OMP END DO nowait
 
-  !$OMP DO collapse(1)
+  !$OMP DO
-  do i = 1, nO
+  do b = 1, nV
     do a = 1, nV
-      T_vo(a,i) = t1(i,a)
+      do i = 1, nO
+        do l = 1, nO
+          T_oovv(l,i,a,b) = t2(i,l,a,b)
+        enddo
+      enddo
     enddo
   enddo
-  !$OMP END DO
+  !$OMP END DO nowait
-  !$OMP END PARALLEL
 
-  call wall_time(ta)
+  !X_oovv(j,k,b,c) * T1_vo(a,i) &
-  energy = 0d0
+
+  !$OMP DO
   do c = 1, nV
     do b = 1, nV
-      do a = 1, nV
+      do k = 1, nO
-        delta_abc = f_v(a) + f_v(b) + f_v(c)
+        do j = 1, nO
-        call form_w_abc(nO,nV,a,b,c,T_vvoo,T_ovvo,X_vvvo,X_ovoo,W_abc)
+          X_oovv(j,k,b,c) = v_vvoo(b,c,j,k)
-        call form_w_abc(nO,nV,b,c,a,T_vvoo,T_ovvo,X_vvvo,X_ovoo,W_bca)
-        call form_w_abc(nO,nV,c,a,b,T_vvoo,T_ovvo,X_vvvo,X_ovoo,W_cab)
-        call form_w_abc(nO,nV,c,b,a,T_vvoo,T_ovvo,X_vvvo,X_ovoo,W_cba)
-
-        call form_v_abc(nO,nV,a,b,c,T_vo,X_vvoo,W_abc,V_abc)
-        call form_v_abc(nO,nV,c,b,a,T_vo,X_vvoo,W_cba,V_cba)
-        !$OMP PARALLEL                                               &
-            !$OMP SHARED(energy,nO,a,b,c,W_abc,W_cab,W_bca,V_abc,V_cba,f_o,f_v,delta_abc)&
-            !$OMP PRIVATE(i,j,k,e,delta)                             &
-            !$OMP DEFAULT(NONE)
-        e = 0d0
-        !$OMP DO
-        do i = 1, nO
-          do j = 1, nO
-            do k = 1, nO
-              delta = 1d0 / (f_o(i) + f_o(j) + f_o(k) - delta_abc)
-              !energy = energy + (4d0 * W(i,j,k,a,b,c) + W(i,j,k,b,c,a) + W(i,j,k,c,a,b)) * (V(i,j,k,a,b,c) - V(i,j,k,c,b,a)) / (cc_space_f_o(i) + cc_space_f_o(j) + cc_space_f_o(k) - cc_space_f_v(a) - cc_space_f_v(b) - cc_space_f_v(c))  !delta_ooovvv(i,j,k,a,b,c)
-              e = e + (4d0 * W_abc(i,j,k) + W_bca(i,j,k) + W_cab(i,j,k))&
-                  * (V_abc(i,j,k) - V_cba(i,j,k)) * delta
-            enddo
-          enddo
         enddo
-        !$OMP END DO NOWAIT
-        !$OMP CRITICAL
-        energy = energy + e
-        !$OMP END CRITICAL
-        !$OMP END PARALLEL
       enddo
     enddo
-    call wall_time(tb)
-    write(*,'(F12.2,A5,F12.2,A2)') dble(i)/dble(nO)*100d0, '% in ', tb - ta, ' s'
   enddo
+  !$OMP END DO nowait
 
-  energy = energy / 3d0
+  !$OMP END PARALLEL
 
-  deallocate(W_abc,V_abc,W_cab,V_cba,W_bca,X_vvvo,X_ovoo,T_vvoo,T_ovvo,T_vo)
+  double precision, external :: ccsd_t_task_aba
-  !deallocate(V,W)
+  double precision, external :: ccsd_t_task_abc
+
+  !$OMP PARALLEL PRIVATE(a,b,c,e) DEFAULT(SHARED)
+  e = 0d0
+  !$OMP DO SCHEDULE(dynamic)
+  do a = 1, nV
+    do b = a+1, nV
+      do c = b+1, nV
+        e = e + ccsd_t_task_abc(a,b,c,nO,nV,t1,T_oovv,T_voov, &
+                        X_ooov,X_oovv,X_vovv,f_o,f_v)
+      enddo
+
+      e = e + ccsd_t_task_aba(a,b,nO,nV,t1,T_oovv,T_voov, &
+                      X_ooov,X_oovv,X_vovv,f_o,f_v)
+
+      e = e + ccsd_t_task_aba(b,a,nO,nV,t1,T_oovv,T_voov, &
+                      X_ooov,X_oovv,X_vovv,f_o,f_v)
+
+    enddo
+  enddo
+  !$OMP END DO NOWAIT
+
+  !$OMP CRITICAL
+  energy = energy + e
+  !$OMP END CRITICAL
+
+  !$OMP END PARALLEL
+
+  energy = energy / 3.d0
+
+  deallocate(X_vovv,X_ooov,T_voov,T_oovv)
 end
 
 
-subroutine form_w_abc(nO,nV,a,b,c,T_vvoo,T_ovvo,X_vvvo,X_ovoo,W_abc)
+double precision function ccsd_t_task_abc(a,b,c,nO,nV,t1,T_oovv,T_voov,&
+      X_ooov,X_oovv,X_vovv,f_o,f_v) result(e)
+  implicit none
+  integer, intent(in)              :: nO,nV,a,b,c
+  double precision, intent(in)     :: t1(nO,nV), f_o(nO), f_v(nV)
+  double precision, intent(in)     :: X_oovv(nO,nO,nV,nV)
+  double precision, intent(in)     :: T_voov(nV,nO,nO,nV), T_oovv(nO,nO,nV,nV)
+  double precision, intent(in)     :: X_vovv(nV,nO,nV,nV), X_ooov(nO,nO,nO,nV)
+
+  double precision :: delta, delta_abc
+  integer  :: i,j,k
+
+  double precision, allocatable :: W_abc(:,:,:), W_cab(:,:,:), W_bca(:,:,:)
+  double precision, allocatable :: W_bac(:,:,:), W_cba(:,:,:), W_acb(:,:,:)
+  double precision, allocatable :: V_abc(:,:,:), V_cab(:,:,:), V_bca(:,:,:)
+  double precision, allocatable :: V_bac(:,:,:), V_cba(:,:,:), V_acb(:,:,:)
+
+  allocate( W_abc(nO,nO,nO), W_cab(nO,nO,nO), W_bca(nO,nO,nO), &
+            W_bac(nO,nO,nO), W_cba(nO,nO,nO), W_acb(nO,nO,nO), &
+            V_abc(nO,nO,nO), V_cab(nO,nO,nO), V_bca(nO,nO,nO), &
+            V_bac(nO,nO,nO), V_cba(nO,nO,nO), V_acb(nO,nO,nO) )
+
+  call form_w_abc(nO,nV,a,b,c,T_voov,T_oovv,X_vovv,X_ooov,W_abc,W_cba,W_bca,W_cab,W_bac,W_acb)
+
+  call form_v_abc(nO,nV,a,b,c,t1,X_oovv,W_abc,V_abc,W_cba,V_cba,W_bca,V_bca,W_cab,V_cab,W_bac,V_bac,W_acb,V_acb)
+
+  delta_abc = f_v(a) + f_v(b) + f_v(c)
+  e = 0.d0
+
+  do k = 1, nO
+    do j = 1, nO
+      do i = 1, nO
+        delta = 1.d0 / (f_o(i) + f_o(j) + f_o(k) - delta_abc)
+        e = e + delta * (                                    &
+            (4d0 * (W_abc(i,j,k) - W_cba(i,j,k)) +           &
+            W_bca(i,j,k) - W_bac(i,j,k)  +                   &
+            W_cab(i,j,k) - W_acb(i,j,k)  ) * (V_abc(i,j,k) - V_cba(i,j,k)) +&
+            (4d0 * (W_acb(i,j,k) - W_bca(i,j,k)) +           &
+            W_cba(i,j,k) - W_cab(i,j,k)  +                   &
+            W_bac(i,j,k) - W_abc(i,j,k)  ) * (V_acb(i,j,k) - V_bca(i,j,k)) +&
+            (4d0 * (W_bac(i,j,k) - W_cab(i,j,k)) +           &
+            W_acb(i,j,k) - W_abc(i,j,k)  +                   &
+            W_cba(i,j,k) - W_bca(i,j,k)  ) * (V_bac(i,j,k) - V_cab(i,j,k)) )
+      enddo
+    enddo
+  enddo
+
+  deallocate(W_abc, W_cab, W_bca, W_bac, W_cba, W_acb, &
+             V_abc, V_cab, V_bca, V_bac, V_cba, V_acb )
+
+end
+
+double precision function ccsd_t_task_aba(a,b,nO,nV,t1,T_oovv,T_voov,&
+      X_ooov,X_oovv,X_vovv,f_o,f_v) result(e)
+  implicit none
+  integer, intent(in)              :: nO,nV,a,b
+  double precision, intent(in)     :: t1(nO,nV), f_o(nO), f_v(nV)
+  double precision, intent(in)     :: X_oovv(nO,nO,nV,nV)
+  double precision, intent(in)     :: T_voov(nV,nO,nO,nV), T_oovv(nO,nO,nV,nV)
+  double precision, intent(in)     :: X_vovv(nV,nO,nV,nV), X_ooov(nO,nO,nO,nV)
+
+  double precision :: delta, delta_abc
+  integer  :: i,j,k
+
+  double precision, allocatable :: W_abc(:,:,:), W_cab(:,:,:), W_bca(:,:,:)
+  double precision, allocatable :: W_bac(:,:,:), W_cba(:,:,:), W_acb(:,:,:)
+  double precision, allocatable :: V_abc(:,:,:), V_cab(:,:,:), V_bca(:,:,:)
+  double precision, allocatable :: V_bac(:,:,:), V_cba(:,:,:), V_acb(:,:,:)
+
+  allocate( W_abc(nO,nO,nO), W_cab(nO,nO,nO), W_bca(nO,nO,nO), &
+            W_bac(nO,nO,nO), W_cba(nO,nO,nO), W_acb(nO,nO,nO), &
+            V_abc(nO,nO,nO), V_cab(nO,nO,nO), V_bca(nO,nO,nO), &
+            V_bac(nO,nO,nO), V_cba(nO,nO,nO), V_acb(nO,nO,nO) )
+
+  call form_w_abc(nO,nV,a,b,a,T_voov,T_oovv,X_vovv,X_ooov,W_abc,W_cba,W_bca,W_cab,W_bac,W_acb)
+
+  call form_v_abc(nO,nV,a,b,a,t1,X_oovv,W_abc,V_abc,W_cba,V_cba,W_bca,V_bca,W_cab,V_cab,W_bac,V_bac,W_acb,V_acb)
+
+  delta_abc = f_v(a) + f_v(b) + f_v(a)
+  e = 0.d0
+
+  do k = 1, nO
+    do j = 1, nO
+      do i = 1, nO
+        delta = 1.d0 / (f_o(i) + f_o(j) + f_o(k) - delta_abc)
+        e = e + delta * (                                    &
+               (4d0 * W_abc(i,j,k) + W_bca(i,j,k) + W_cab(i,j,k)) * (V_abc(i,j,k) - V_cba(i,j,k)) + &
+               (4d0 * W_acb(i,j,k) + W_cba(i,j,k) + W_bac(i,j,k)) * (V_acb(i,j,k) - V_bca(i,j,k)) + &
+               (4d0 * W_bac(i,j,k) + W_acb(i,j,k) + W_cba(i,j,k)) * (V_bac(i,j,k) - V_cab(i,j,k)) )
+
+      enddo
+    enddo
+  enddo
+
+  deallocate(W_abc, W_cab, W_bca, W_bac, W_cba, W_acb, &
+             V_abc, V_cab, V_bca, V_bac, V_cba, V_acb )
+
+end
+
+subroutine form_w_abc(nO,nV,a,b,c,T_voov,T_oovv,X_vovv,X_ooov,W_abc,W_cba,W_bca,W_cab,W_bac,W_acb)
 
   implicit none
 
   integer, intent(in)           :: nO,nV,a,b,c
-  !double precision, intent(in) :: t2(nO,nO,nV,nV)
+  double precision, intent(in)  :: T_voov(nV,nO,nO,nV), T_oovv(nO,nO,nV,nV)
-  double precision, intent(in)  :: T_vvoo(nV,nV,nO,nO), T_ovvo(nO,nV,nV,nO)
+  double precision, intent(in)  :: X_vovv(nV,nO,nV,nV), X_ooov(nO,nO,nO,nV)
-  double precision, intent(in)  :: X_vvvo(nV,nV,nV,nO), X_ovoo(nO,nV,nO,nO)
   double precision, intent(out) :: W_abc(nO,nO,nO)
+  double precision, intent(out) :: W_cba(nO,nO,nO)
+  double precision, intent(out) :: W_bca(nO,nO,nO)
+  double precision, intent(out) :: W_cab(nO,nO,nO)
+  double precision, intent(out) :: W_bac(nO,nO,nO)
+  double precision, intent(out) :: W_acb(nO,nO,nO)
 
   integer :: l,i,j,k,d
+  double precision, allocatable, dimension(:,:,:,:) :: W_ikj
+  double precision, allocatable :: X(:,:,:,:)
 
+  allocate(W_ikj(nO,nO,nO,6))
+  allocate(X(nV,nO,nO,3))
 
-  !$OMP PARALLEL &
+  do k=1,nO
-  !$OMP SHARED(nO,nV,a,b,c,T_vvoo,T_ovvo,X_vvvo,X_ovoo,W_abc) &
+    do i=1,nO
-  !$OMP PRIVATE(i,j,k,d,l) &
+      do d=1,nV
-  !$OMP DEFAULT(NONE)
+        X(d,i,k,1) = T_voov(d,k,i,a)
-
+        X(d,i,k,2) = T_voov(d,k,i,b)
-  !$OMP DO collapse(3)
+        X(d,i,k,3) = T_voov(d,k,i,c)
-  do k = 1, nO
-    do j = 1, nO
-      do i = 1, nO
-        W_abc(i,j,k) = 0.d0
-
-        do d = 1, nV
-          W_abc(i,j,k) = W_abc(i,j,k) &
-                 + X_vvvo(d,b,a,i) * T_vvoo(d,c,k,j) &
-                 + X_vvvo(d,c,a,i) * T_vvoo(d,b,j,k) &
-                 + X_vvvo(d,a,c,k) * T_vvoo(d,b,j,i) &
-                 + X_vvvo(d,b,c,k) * T_vvoo(d,a,i,j) &
-                 + X_vvvo(d,c,b,j) * T_vvoo(d,a,i,k) &
-                 + X_vvvo(d,a,b,j) * T_vvoo(d,c,k,i)
-
-        enddo
-
-        do l = 1, nO
-          W_abc(i,j,k) = W_abc(i,j,k) &
-              - T_ovvo(l,a,b,i) * X_ovoo(l,c,j,k) &
-              - T_ovvo(l,a,c,i) * X_ovoo(l,b,k,j) & ! bc kj
-              - T_ovvo(l,c,a,k) * X_ovoo(l,b,i,j) & ! prev ac ik
-              - T_ovvo(l,c,b,k) * X_ovoo(l,a,j,i) & ! prev ab ij
-              - T_ovvo(l,b,c,j) * X_ovoo(l,a,k,i) & ! prev bc kj
-              - T_ovvo(l,b,a,j) * X_ovoo(l,c,i,k) ! prev ac ik
-        enddo
-
       enddo
     enddo
   enddo
-  !$OMP END DO
+
-  !$OMP END PARALLEL
+!   X_vovv(d,i,c,a) * T_voov(d,j,k,b) : i jk
+
+  call dgemm('T','N', nO, nO*nO, nV, 1.d0, X_vovv(1,1,c,a), nV, T_voov(1,1,1,b), nV, 0.d0, W_abc, nO)
+  call dgemm('T','N', nO, nO*nO, nV, 1.d0, X_vovv(1,1,c,b), nV, T_voov(1,1,1,a), nV, 0.d0, W_bac, nO)
+  call dgemm('T','N', nO, nO*nO, nV, 1.d0, X_vovv(1,1,a,c), nV, T_voov(1,1,1,b), nV, 0.d0, W_cba, nO)
+  call dgemm('T','N', nO, nO*nO, nV, 1.d0, X_vovv(1,1,a,b), nV, T_voov(1,1,1,c), nV, 0.d0, W_bca, nO)
+  call dgemm('T','N', nO, nO*nO, nV, 1.d0, X_vovv(1,1,b,c), nV, T_voov(1,1,1,a), nV, 0.d0, W_cab, nO)
+  call dgemm('T','N', nO, nO*nO, nV, 1.d0, X_vovv(1,1,b,a), nV, T_voov(1,1,1,c), nV, 0.d0, W_acb, nO)
+
+!   T_voov(d,i,j,a) * X_vovv(d,k,b,c) : ij k
+
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, T_voov(1,1,1,a), nV, X_vovv(1,1,b,c), nV, 1.d0, W_abc, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, T_voov(1,1,1,b), nV, X_vovv(1,1,a,c), nV, 1.d0, W_bac, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, T_voov(1,1,1,c), nV, X_vovv(1,1,b,a), nV, 1.d0, W_cba, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, T_voov(1,1,1,b), nV, X_vovv(1,1,c,a), nV, 1.d0, W_bca, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, T_voov(1,1,1,c), nV, X_vovv(1,1,a,b), nV, 1.d0, W_cab, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, T_voov(1,1,1,a), nV, X_vovv(1,1,c,b), nV, 1.d0, W_acb, nO*nO)
 
 
+!   X_vovv(d,k,a,c) * T_voov(d,j,i,b) : k ji
+
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, X(1,1,1,2), nV, X_vovv(1,1,a,c), nV, 1.d0, W_abc, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, X(1,1,1,1), nV, X_vovv(1,1,b,c), nV, 1.d0, W_bac, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, X(1,1,1,2), nV, X_vovv(1,1,c,a), nV, 1.d0, W_cba, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, X(1,1,1,3), nV, X_vovv(1,1,b,a), nV, 1.d0, W_bca, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, X(1,1,1,1), nV, X_vovv(1,1,c,b), nV, 1.d0, W_cab, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, X(1,1,1,3), nV, X_vovv(1,1,a,b), nV, 1.d0, W_acb, nO*nO)
+
+!   X_vovv(d,i,b,a) * T_voov(d,k,j,c) : i kj
+
+  call dgemm('T','N', nO, nO*nO, nV, 1.d0, X_vovv(1,1,b,a), nV, X(1,1,1,3), nV, 1.d0, W_abc, nO)
+  call dgemm('T','N', nO, nO*nO, nV, 1.d0, X_vovv(1,1,a,b), nV, X(1,1,1,3), nV, 1.d0, W_bac, nO)
+  call dgemm('T','N', nO, nO*nO, nV, 1.d0, X_vovv(1,1,b,c), nV, X(1,1,1,1), nV, 1.d0, W_cba, nO)
+  call dgemm('T','N', nO, nO*nO, nV, 1.d0, X_vovv(1,1,c,b), nV, X(1,1,1,1), nV, 1.d0, W_bca, nO)
+  call dgemm('T','N', nO, nO*nO, nV, 1.d0, X_vovv(1,1,a,c), nV, X(1,1,1,2), nV, 1.d0, W_cab, nO)
+  call dgemm('T','N', nO, nO*nO, nV, 1.d0, X_vovv(1,1,c,a), nV, X(1,1,1,2), nV, 1.d0, W_acb, nO)
+
+!  T_voov(d,k,i,c) * X_vovv(d,j,a,b) : ki j
+
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, X(1,1,1,3), nV, X_vovv(1,1,a,b), nV, 0.d0, W_ikj(1,1,1,1), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, X(1,1,1,3), nV, X_vovv(1,1,b,a), nV, 0.d0, W_ikj(1,1,1,2), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, X(1,1,1,1), nV, X_vovv(1,1,c,b), nV, 0.d0, W_ikj(1,1,1,3), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, X(1,1,1,1), nV, X_vovv(1,1,b,c), nV, 0.d0, W_ikj(1,1,1,4), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, X(1,1,1,2), nV, X_vovv(1,1,c,a), nV, 0.d0, W_ikj(1,1,1,5), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, X(1,1,1,2), nV, X_vovv(1,1,a,c), nV, 0.d0, W_ikj(1,1,1,6), nO*nO)
+
+!   T_voov(d,i,k,a) * X_vovv(d,j,c,b) : ik j
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, T_voov(1,1,1,a), nV, X_vovv(1,1,c,b), nV, 1.d0, W_ikj(1,1,1,1), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, T_voov(1,1,1,b), nV, X_vovv(1,1,c,a), nV, 1.d0, W_ikj(1,1,1,2), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, T_voov(1,1,1,c), nV, X_vovv(1,1,a,b), nV, 1.d0, W_ikj(1,1,1,3), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, T_voov(1,1,1,b), nV, X_vovv(1,1,a,c), nV, 1.d0, W_ikj(1,1,1,4), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, T_voov(1,1,1,c), nV, X_vovv(1,1,b,a), nV, 1.d0, W_ikj(1,1,1,5), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nV, 1.d0, T_voov(1,1,1,a), nV, X_vovv(1,1,b,c), nV, 1.d0, W_ikj(1,1,1,6), nO*nO)
+
+  deallocate(X)
+
+  allocate(X(nO,nO,nO,3))
+
+  do k=1,nO
+    do j=1,nO
+      do l=1,nO
+        X(l,j,k,1) = X_ooov(l,k,j,a)
+        X(l,j,k,2) = X_ooov(l,k,j,b)
+        X(l,j,k,3) = X_ooov(l,k,j,c)
+      enddo
+    enddo
+  enddo
+
+
+!   - T_oovv(l,i,a,b) * X_ooov(l,j,k,c) : i jk
+  call dgemm('T','N', nO, nO*nO, nO, -1.d0, T_oovv(1,1,a,b), nO, X_ooov(1,1,1,c), nO, 1.d0, W_abc, nO)
+  call dgemm('T','N', nO, nO*nO, nO, -1.d0, T_oovv(1,1,b,a), nO, X_ooov(1,1,1,c), nO, 1.d0, W_bac, nO)
+  call dgemm('T','N', nO, nO*nO, nO, -1.d0, T_oovv(1,1,c,b), nO, X_ooov(1,1,1,a), nO, 1.d0, W_cba, nO)
+  call dgemm('T','N', nO, nO*nO, nO, -1.d0, T_oovv(1,1,b,c), nO, X_ooov(1,1,1,a), nO, 1.d0, W_bca, nO)
+  call dgemm('T','N', nO, nO*nO, nO, -1.d0, T_oovv(1,1,c,a), nO, X_ooov(1,1,1,b), nO, 1.d0, W_cab, nO)
+  call dgemm('T','N', nO, nO*nO, nO, -1.d0, T_oovv(1,1,a,c), nO, X_ooov(1,1,1,b), nO, 1.d0, W_acb, nO)
+
+!   - T_oovv(l,i,a,c) * X_ooov(l,k,j,b) : i kj
+  call dgemm('T','N', nO, nO*nO, nO, -1.d0, T_oovv(1,1,a,c), nO, X(1,1,1,2), nO, 1.d0, W_abc, nO)
+  call dgemm('T','N', nO, nO*nO, nO, -1.d0, T_oovv(1,1,b,c), nO, X(1,1,1,1), nO, 1.d0, W_bac, nO)
+  call dgemm('T','N', nO, nO*nO, nO, -1.d0, T_oovv(1,1,c,a), nO, X(1,1,1,2), nO, 1.d0, W_cba, nO)
+  call dgemm('T','N', nO, nO*nO, nO, -1.d0, T_oovv(1,1,b,a), nO, X(1,1,1,3), nO, 1.d0, W_bca, nO)
+  call dgemm('T','N', nO, nO*nO, nO, -1.d0, T_oovv(1,1,c,b), nO, X(1,1,1,1), nO, 1.d0, W_cab, nO)
+  call dgemm('T','N', nO, nO*nO, nO, -1.d0, T_oovv(1,1,a,b), nO, X(1,1,1,3), nO, 1.d0, W_acb, nO)
+
+!   - X_ooov(l,i,j,b) * T_oovv(l,k,c,a) : ij k
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X_ooov(1,1,1,b), nO, T_oovv(1,1,c,a), nO, 1.d0, W_abc, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X_ooov(1,1,1,a), nO, T_oovv(1,1,c,b), nO, 1.d0, W_bac, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X_ooov(1,1,1,b), nO, T_oovv(1,1,a,c), nO, 1.d0, W_cba, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X_ooov(1,1,1,c), nO, T_oovv(1,1,a,b), nO, 1.d0, W_bca, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X_ooov(1,1,1,a), nO, T_oovv(1,1,b,c), nO, 1.d0, W_cab, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X_ooov(1,1,1,c), nO, T_oovv(1,1,b,a), nO, 1.d0, W_acb, nO*nO)
+
+!   - X_ooov(l,j,i,a) * T_oovv(l,k,c,b) : ji k
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X(1,1,1,1), nO, T_oovv(1,1,c,b), nO, 1.d0, W_abc, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X(1,1,1,2), nO, T_oovv(1,1,c,a), nO, 1.d0, W_bac, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X(1,1,1,3), nO, T_oovv(1,1,a,b), nO, 1.d0, W_cba, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X(1,1,1,2), nO, T_oovv(1,1,a,c), nO, 1.d0, W_bca, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X(1,1,1,3), nO, T_oovv(1,1,b,a), nO, 1.d0, W_cab, nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X(1,1,1,1), nO, T_oovv(1,1,b,c), nO, 1.d0, W_acb, nO*nO)
+
+!   - X_ooov(l,k,i,a) * T_oovv(l,j,b,c) : ki j
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X(1,1,1,1), nO, T_oovv(1,1,b,c), nO, 1.d0, W_ikj(1,1,1,1), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X(1,1,1,2), nO, T_oovv(1,1,a,c), nO, 1.d0, W_ikj(1,1,1,2), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X(1,1,1,3), nO, T_oovv(1,1,b,a), nO, 1.d0, W_ikj(1,1,1,3), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X(1,1,1,2), nO, T_oovv(1,1,c,a), nO, 1.d0, W_ikj(1,1,1,4), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X(1,1,1,3), nO, T_oovv(1,1,a,b), nO, 1.d0, W_ikj(1,1,1,5), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X(1,1,1,1), nO, T_oovv(1,1,c,b), nO, 1.d0, W_ikj(1,1,1,6), nO*nO)
+
+!   - X_ooov(l,i,k,c) * T_oovv(l,j,b,a) : ik j
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X_ooov(1,1,1,c), nO, T_oovv(1,1,b,a), nO, 1.d0, W_ikj(1,1,1,1), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X_ooov(1,1,1,c), nO, T_oovv(1,1,a,b), nO, 1.d0, W_ikj(1,1,1,2), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X_ooov(1,1,1,a), nO, T_oovv(1,1,b,c), nO, 1.d0, W_ikj(1,1,1,3), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X_ooov(1,1,1,a), nO, T_oovv(1,1,c,b), nO, 1.d0, W_ikj(1,1,1,4), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X_ooov(1,1,1,b), nO, T_oovv(1,1,a,c), nO, 1.d0, W_ikj(1,1,1,5), nO*nO)
+  call dgemm('T','N', nO*nO, nO, nO, -1.d0, X_ooov(1,1,1,b), nO, T_oovv(1,1,c,a), nO, 1.d0, W_ikj(1,1,1,6), nO*nO)
+
+  do k=1,nO
+    do j=1,nO
+      do i=1,nO
+        W_abc(i,j,k) = W_abc(i,j,k) + W_ikj(i,k,j,1)
+        W_bac(i,j,k) = W_bac(i,j,k) + W_ikj(i,k,j,2)
+        W_cba(i,j,k) = W_cba(i,j,k) + W_ikj(i,k,j,3)
+        W_bca(i,j,k) = W_bca(i,j,k) + W_ikj(i,k,j,4)
+        W_cab(i,j,k) = W_cab(i,j,k) + W_ikj(i,k,j,5)
+        W_acb(i,j,k) = W_acb(i,j,k) + W_ikj(i,k,j,6)
+      enddo
+    enddo
+  enddo
+
+  deallocate(X,W_ikj)
 end
 
 
 ! V_abc
 
-subroutine form_v_abc(nO,nV,a,b,c,T_vo,X_vvoo,W,V)
+subroutine form_v_abc(nO,nV,a,b,c,T_ov,X_oovv,W_abc,V_abc,W_cba,V_cba,W_bca,V_bca,W_cab,V_cab,W_bac,V_bac,W_acb,V_acb)
 
 implicit none
 
   integer, intent(in)           :: nO,nV,a,b,c
-  !double precision, intent(in)  :: t1(nO,nV)
+  double precision, intent(in)  :: T_ov(nO,nV)
-  double precision, intent(in)  :: T_vo(nV,nO)
+  double precision, intent(in)  :: X_oovv(nO,nO,nV,nV)
-  double precision, intent(in)  :: X_vvoo(nV,nV,nO,nO)
+  double precision, intent(in)  :: W_abc(nO,nO,nO), W_cab(nO,nO,nO), W_bca(nO,nO,nO)
-  double precision, intent(in)  :: W(nO,nO,nO)
+  double precision, intent(in)  :: W_bac(nO,nO,nO), W_cba(nO,nO,nO), W_acb(nO,nO,nO)
-  double precision, intent(out) :: V(nO,nO,nO)
+  double precision, intent(out) :: V_abc(nO,nO,nO), V_cab(nO,nO,nO), V_bca(nO,nO,nO)
+  double precision, intent(out) :: V_bac(nO,nO,nO), V_cba(nO,nO,nO), V_acb(nO,nO,nO)
 
   integer :: i,j,k
 
-  !$OMP PARALLEL &
-  !$OMP SHARED(nO,nV,a,b,c,T_vo,X_vvoo,W,V) &
-  !$OMP PRIVATE(i,j,k) &
-  !$OMP DEFAULT(NONE)
-  !$OMP DO collapse(2)
   do k = 1, nO
     do j = 1, nO
       do i = 1, nO
-        !V(i,j,k,a,b,c) = V(i,j,k,a,b,c) + W(i,j,k,a,b,c) &
+        V_abc(i,j,k) = W_abc(i,j,k) &
-        V(i,j,k) = W(i,j,k) &
+           + X_oovv(j,k,b,c) * T_ov(i,a) &
-           + X_vvoo(b,c,k,j) * T_vo(a,i) &
+           + X_oovv(i,k,a,c) * T_ov(j,b) &
-           + X_vvoo(a,c,k,i) * T_vo(b,j) &
+           + X_oovv(i,j,a,b) * T_ov(k,c)
-           + X_vvoo(a,b,j,i) * T_vo(c,k)
+
+        V_cba(i,j,k) = W_cba(i,j,k) &
+           + X_oovv(j,k,b,a) * T_ov(i,c) &
+           + X_oovv(i,k,c,a) * T_ov(j,b) &
+           + X_oovv(i,j,c,b) * T_ov(k,a)
+
+        V_bca(i,j,k) = W_bca(i,j,k) &
+           + X_oovv(j,k,c,a) * T_ov(i,b) &
+           + X_oovv(i,k,b,a) * T_ov(j,c) &
+           + X_oovv(i,j,b,c) * T_ov(k,a)
+
+        V_cab(i,j,k) = W_cab(i,j,k) &
+           + X_oovv(j,k,a,b) * T_ov(i,c) &
+           + X_oovv(i,k,c,b) * T_ov(j,a) &
+           + X_oovv(i,j,c,a) * T_ov(k,b)
+
+        V_bac(i,j,k) = W_bac(i,j,k) &
+           + X_oovv(j,k,a,c) * T_ov(i,b) &
+           + X_oovv(i,k,b,c) * T_ov(j,a) &
+           + X_oovv(i,j,b,a) * T_ov(k,c)
+
+        V_acb(i,j,k) = W_acb(i,j,k) &
+           + X_oovv(j,k,c,b) * T_ov(i,a) &
+           + X_oovv(i,k,a,b) * T_ov(j,c) &
+           + X_oovv(i,j,a,c) * T_ov(k,b)
+
       enddo
     enddo
   enddo
-  !$OMP END DO
-  !$OMP END PARALLEL
 
 end
 

src/ccsd/ccsd_t_space_orb_stoch.irp.f

@@ -0,0 +1,375 @@
+! Main
+subroutine ccsd_par_t_space_stoch(nO,nV,t1,t2,f_o,f_v,v_vvvo,v_vvoo,v_vooo,energy)
+
+  implicit none
+
+  integer, intent(in)           :: nO,nV
+  double precision, intent(in)  :: t1(nO,nV), f_o(nO), f_v(nV)
+  double precision, intent(in)  :: t2(nO,nO,nV,nV)
+  double precision, intent(in)  :: v_vvvo(nV,nV,nV,nO), v_vvoo(nV,nV,nO,nO), v_vooo(nV,nO,nO,nO)
+  double precision, intent(inout) :: energy
+
+  double precision, allocatable :: X_vovv(:,:,:,:), X_ooov(:,:,:,:), X_oovv(:,:,:,:)
+  double precision, allocatable :: T_voov(:,:,:,:), T_oovv(:,:,:,:)
+  integer                       :: i,j,k,l,a,b,c,d
+  double precision              :: e,ta,tb,eccsd
+
+  eccsd = energy
+  call set_multiple_levels_omp(.False.)
+
+  allocate(X_vovv(nV,nO,nV,nV), X_ooov(nO,nO,nO,nV), X_oovv(nO,nO,nV,nV))
+  allocate(T_voov(nV,nO,nO,nV),T_oovv(nO,nO,nV,nV))
+
+  !$OMP PARALLEL &
+  !$OMP SHARED(nO,nV,T_voov,T_oovv,X_vovv,X_ooov,X_oovv, &
+  !$OMP t1,t2,v_vvvo,v_vooo,v_vvoo) &
+  !$OMP PRIVATE(a,b,c,d,i,j,k,l) &
+  !$OMP DEFAULT(NONE)
+
+  !v_vvvo(b,a,d,i) * t2(k,j,c,d) &
+  !X_vovv(d,i,b,a,i) * T_voov(d,j,c,k)
+
+  !$OMP DO
+  do a = 1, nV
+    do b = 1, nV
+      do i = 1, nO
+        do d = 1, nV
+          X_vovv(d,i,b,a) = v_vvvo(b,a,d,i)
+        enddo
+      enddo
+    enddo
+  enddo
+  !$OMP END DO nowait
+
+  !$OMP DO
+  do c = 1, nV
+    do j = 1, nO
+      do k = 1, nO
+        do d = 1, nV
+          T_voov(d,k,j,c) = t2(k,j,c,d)
+        enddo
+      enddo
+    enddo
+  enddo
+  !$OMP END DO nowait
+
+  !v_vooo(c,j,k,l) * t2(i,l,a,b) &
+  !X_ooov(l,j,k,c) * T_oovv(l,i,a,b) &
+
+  !$OMP DO
+  do c = 1, nV
+    do k = 1, nO
+      do j = 1, nO
+        do l = 1, nO
+           X_ooov(l,j,k,c) = v_vooo(c,j,k,l)
+        enddo
+      enddo
+    enddo
+  enddo
+  !$OMP END DO nowait
+
+  !$OMP DO
+  do b = 1, nV
+    do a = 1, nV
+      do i = 1, nO
+        do l = 1, nO
+          T_oovv(l,i,a,b) = t2(i,l,a,b)
+        enddo
+      enddo
+    enddo
+  enddo
+  !$OMP END DO nowait
+
+  !X_oovv(j,k,b,c) * T1_vo(a,i) &
+
+  !$OMP DO
+  do c = 1, nV
+    do b = 1, nV
+      do k = 1, nO
+        do j = 1, nO
+          X_oovv(j,k,b,c) = v_vvoo(b,c,j,k)
+        enddo
+      enddo
+    enddo
+  enddo
+  !$OMP END DO nowait
+
+  !$OMP END PARALLEL
+
+  double precision, external :: ccsd_t_task_aba
+  double precision, external :: ccsd_t_task_abc
+!  logical, external :: omp_test_lock
+
+  double precision, allocatable :: memo(:), Pabc(:), waccu(:)
+  integer*8, allocatable :: sampled(:)
+!  integer(omp_lock_kind), allocatable :: lock(:)
+  integer*2       , allocatable :: abc(:,:)
+  integer*8                     :: Nabc, i8,kiter
+  integer*8, allocatable :: iorder(:)
+  double precision :: eocc
+  double precision :: norm
+  integer :: isample
+
+
+  ! Prepare table of triplets (a,b,c)
+
+  Nabc = (int(nV,8) * int(nV+1,8) * int(nV+2,8))/6_8 - nV
+  allocate (memo(Nabc), sampled(Nabc), Pabc(Nabc), waccu(0:Nabc))
+  allocate (abc(4,Nabc), iorder(Nabc)) !, lock(Nabc))
+
+!  eocc = 3.d0/dble(nO) * sum(f_o(1:nO))
+  Nabc = 0_8
+  do a = 1, nV
+    do b = a+1, nV
+      do c = b+1, nV
+        Nabc = Nabc + 1_8
+        Pabc(Nabc) = -1.d0/(f_v(a) + f_v(b) + f_v(c))
+        abc(1,Nabc) = int(a,2)
+        abc(2,Nabc) = int(b,2)
+        abc(3,Nabc) = int(c,2)
+      enddo
+
+      Nabc = Nabc + 1_8
+      abc(1,Nabc) = int(a,2)
+      abc(2,Nabc) = int(b,2)
+      abc(3,Nabc) = int(a,2)
+      Pabc(Nabc) = -1.d0/(2.d0*f_v(a) + f_v(b))
+
+      Nabc = Nabc + 1_8
+      abc(1,Nabc) = int(b,2)
+      abc(2,Nabc) = int(a,2)
+      abc(3,Nabc) = int(b,2)
+      Pabc(Nabc) = -1.d0/(f_v(a) + 2.d0*f_v(b))
+    enddo
+  enddo
+
+  do i8=1,Nabc
+   iorder(i8) = i8
+  enddo
+
+  ! Sort triplets in decreasing Pabc
+  call dsort_big(Pabc, iorder, Nabc)
+
+  ! Normalize
+  norm = 0.d0
+  do i8=Nabc,1,-1
+    norm = norm + Pabc(i8)
+  enddo
+  norm = 1.d0/norm
+  do i8=1,Nabc
+    Pabc(i8) = Pabc(i8) * norm
+  enddo
+
+  call i8set_order_big(abc, iorder, Nabc)
+
+
+  ! Cumulative distribution for sampling
+  waccu(Nabc) = 0.d0
+  do i8=Nabc-1,1,-1
+   waccu(i8) = waccu(i8+1) - Pabc(i8+1)
+  enddo
+  waccu(:) = waccu(:) + 1.d0
+  waccu(0) = 0.d0
+
+  logical :: converged, do_comp
+  double precision :: eta, variance, error, sample
+  double precision :: t00, t01
+  integer*8 :: ieta, Ncomputed
+  integer*8, external :: binary_search
+
+  integer :: nbuckets
+  nbuckets = 100
+
+  double precision, allocatable :: wsum(:)
+  allocate(wsum(nbuckets))
+
+  converged = .False.
+  Ncomputed = 0_8
+
+  energy = 0.d0
+  variance = 0.d0
+  memo(:) = 0.d0
+  sampled(:) = -1_8
+
+  integer*8 :: ileft, iright, imin
+  ileft = 1_8
+  iright = Nabc
+  integer*8, allocatable :: bounds(:,:)
+
+  allocate (bounds(2,nbuckets))
+  do isample=1,nbuckets
+    eta = 1.d0/dble(nbuckets) * dble(isample)
+    ieta = binary_search(waccu,eta,Nabc)
+    bounds(1,isample) = ileft
+    bounds(2,isample) = ieta
+    ileft = ieta+1
+    wsum(isample) = sum( Pabc(bounds(1,isample):bounds(2,isample) ) )
+  enddo
+
+  Pabc(:) = 1.d0/Pabc(:)
+
+  print '(A)', ''
+  print '(A)', ' +----------------------+--------------+----------+'
+  print '(A)', ' |      E(CCSD(T))      |   Error      |     %    |'
+  print '(A)', ' +----------------------+--------------+----------+'
+
+
+  call wall_time(t00)
+  imin = 1_8
+  !$OMP PARALLEL                                                     &
+      !$OMP PRIVATE(ieta,eta,a,b,c,kiter,isample)                    &
+      !$OMP DEFAULT(SHARED)
+
+  do kiter=1,Nabc
+
+    !$OMP MASTER
+    do while (imin <= Nabc)
+      if (sampled(imin)>-1_8) then
+        imin = imin+1
+      else
+        exit
+      endif
+    enddo
+
+    ! Deterministic part
+    if (imin < Nabc) then
+      ieta=imin
+      sampled(ieta) = 0_8
+      a = abc(1,ieta)
+      b = abc(2,ieta)
+      c = abc(3,ieta)
+      Ncomputed += 1_8
+      !$OMP TASK DEFAULT(SHARED) FIRSTPRIVATE(a,b,c,ieta)
+      if (a/=c) then
+        memo(ieta) = ccsd_t_task_abc(a,b,c,nO,nV,t1,T_oovv,T_voov, &
+            X_ooov,X_oovv,X_vovv,f_o,f_v) / 3.d0
+      else
+        memo(ieta) =  ccsd_t_task_aba(a,b,nO,nV,t1,T_oovv,T_voov,  &
+            X_ooov,X_oovv,X_vovv,f_o,f_v) / 3.d0
+      endif
+      !$OMP END TASK
+    endif
+
+    ! Stochastic part
+    call random_number(eta)
+    do isample=1,nbuckets
+      if (imin >= bounds(2,isample)) then
+        cycle
+      endif
+      ieta = binary_search(waccu,(eta + dble(isample-1))/dble(nbuckets),Nabc)
+
+      if (sampled(ieta) == -1_8) then
+        sampled(ieta) = 0_8
+        a = abc(1,ieta)
+        b = abc(2,ieta)
+        c = abc(3,ieta)
+        Ncomputed += 1_8
+        !$OMP TASK DEFAULT(SHARED) FIRSTPRIVATE(a,b,c,ieta)
+        if (a/=c) then
+          memo(ieta) = ccsd_t_task_abc(a,b,c,nO,nV,t1,T_oovv,T_voov, &
+              X_ooov,X_oovv,X_vovv,f_o,f_v) / 3.d0
+        else
+          memo(ieta) =  ccsd_t_task_aba(a,b,nO,nV,t1,T_oovv,T_voov,  &
+              X_ooov,X_oovv,X_vovv,f_o,f_v) / 3.d0
+        endif
+        !$OMP END TASK
+      endif
+      sampled(ieta) = sampled(ieta)+1_8
+
+    enddo
+
+    call wall_time(t01)
+    if ((t01-t00 > 1.0d0).or.(imin >= Nabc)) then
+      t00 = t01
+
+      !$OMP TASKWAIT
+
+      double precision :: ET, ET2
+      double precision :: energy_stoch, energy_det
+      double precision :: scale
+      double precision :: w
+      double precision :: tmp
+      energy_stoch = 0.d0
+      energy_det   = 0.d0
+      norm = 0.d0
+      scale = 1.d0
+      ET = 0.d0
+      ET2 = 0.d0
+
+
+      do isample=1,nbuckets
+        if (imin >= bounds(2,isample)) then
+          energy_det = energy_det + sum(memo(bounds(1,isample):bounds(2,isample)))
+          scale = scale - wsum(isample)
+        else
+          exit
+        endif
+      enddo
+
+      isample = min(isample,nbuckets)
+      do ieta=bounds(1,isample), Nabc
+          w = dble(max(sampled(ieta),0_8))
+          tmp = w * memo(ieta) * Pabc(ieta)
+          ET = ET + tmp
+          ET2 = ET2 + tmp * memo(ieta) * Pabc(ieta)
+          norm = norm + w
+      enddo
+      norm = norm/scale
+      if (norm > 0.d0) then
+        energy_stoch = ET / norm
+        variance = ET2 / norm - energy_stoch*energy_stoch
+      endif
+
+      energy = energy_det + energy_stoch
+
+      print '('' | '',F20.8, '' | '', E12.4,'' | '', F8.2,'' |'')', eccsd+energy, dsqrt(variance/(norm-1.d0)), 100.*real(Ncomputed)/real(Nabc)
+    endif
+    !$OMP END MASTER
+    if (imin >= Nabc) exit
+  enddo
+
+  !$OMP END PARALLEL
+  print '(A)', ' +----------------------+--------------+----------+'
+  print '(A)', ''
+
+  deallocate(X_vovv,X_ooov,T_voov,T_oovv)
+end
+
+
+
+integer*8 function binary_search(arr, key, sze)
+    implicit none
+    BEGIN_DOC
+! Searches the key in array arr(1:sze) between l_in and r_in, and returns its index
+    END_DOC
+    integer*8 :: sze, i, j, mid
+    double precision :: arr(0:sze)
+    double precision :: key
+
+    if ( key < arr(1) ) then
+      binary_search = 0_8
+      return 
+    end if
+
+    if ( key >= arr(sze) ) then
+      binary_search = sze
+      return 
+    end if
+
+    i = 0_8
+    j = sze + 1_8
+
+    do while (.True.)
+      mid = (i + j) / 2_8
+      if ( key >= arr(mid) ) then
+        i = mid
+      else
+        j = mid
+      end if
+      if (j-i <= 1_8) then
+        binary_search = i
+        return
+      endif
+    end do
+end function binary_search
+

src/ccsd/save_energy.irp.f

@@ -0,0 +1,13 @@
+subroutine save_energy(E,ET)
+  implicit none
+  BEGIN_DOC
+! Saves the energy in |EZFIO|.
+  END_DOC
+  double precision, intent(in) :: E, ET
+  call ezfio_set_ccsd_energy(E)
+  if (ET /= 0.d0) then
+    call ezfio_set_ccsd_energy_t(E+ET)
+  endif
+end
+
+

src/cipsi/selection.irp.f

@@ -76,6 +76,8 @@ subroutine select_connected(i_generator,E0,pt2_data,b,subset,csubset)
 
   double precision, allocatable  :: fock_diag_tmp(:,:)
 
+  if (csubset == 0) return
+
   allocate(fock_diag_tmp(2,mo_num+1))
 
   call build_fock_tmp(fock_diag_tmp,psi_det_generators(1,1,i_generator),N_int)
@@ -86,6 +88,10 @@ subroutine select_connected(i_generator,E0,pt2_data,b,subset,csubset)
       particle_mask(k,1) = iand(generators_bitmask(k,1,s_part), not(psi_det_generators(k,1,i_generator)) )
       particle_mask(k,2) = iand(generators_bitmask(k,2,s_part), not(psi_det_generators(k,2,i_generator)) )
   enddo
+  if ((subset == 1).and.(sum(hole_mask(:,2)) == 0_bit_kind)) then
+     ! No beta electron to excite
+     call select_singles(i_generator,hole_mask,particle_mask,fock_diag_tmp,E0,pt2_data,b)
+  endif
   call select_singles_and_doubles(i_generator,hole_mask,particle_mask,fock_diag_tmp,E0,pt2_data,b,subset,csubset)
   deallocate(fock_diag_tmp)
 end subroutine
@@ -140,7 +146,7 @@ subroutine select_singles_and_doubles(i_generator,hole_mask,particle_mask,fock_d
   use selection_types
   implicit none
   BEGIN_DOC
-!            WARNING /!\ : It is assumed that the generators and selectors are psi_det_sorted
+!  WARNING /!\ : It is assumed that the generators and selectors are psi_det_sorted
   END_DOC
 
   integer, intent(in)            :: i_generator, subset, csubset
@@ -177,6 +183,7 @@ subroutine select_singles_and_doubles(i_generator,hole_mask,particle_mask,fock_d
   monoAdo = .true.
   monoBdo = .true.
 
+  if (csubset == 0) return
 
   do k=1,N_int
     hole    (k,1) = iand(psi_det_generators(k,1,i_generator), hole_mask(k,1))
@@ -234,7 +241,6 @@ subroutine select_singles_and_doubles(i_generator,hole_mask,particle_mask,fock_d
   enddo
 
   ! Iterate on 0S alpha, and find betas TQ such that exc_degree <= 4
-  ! Remove also contributions < 1.d-20)
   do j=1,N_det_alpha_unique
     call get_excitation_degree_spin(psi_det_alpha_unique(1,j),       &
         psi_det_generators(1,1,i_generator), nt, N_int)
@@ -477,7 +483,9 @@ subroutine select_singles_and_doubles(i_generator,hole_mask,particle_mask,fock_d
       do s2=s1,2
         sp = s1
 
-        if(s1 /= s2) sp = 3
+        if(s1 /= s2) then
+          sp = 3
+        endif
 
         ib = 1
         if(s1 == s2) ib = i1+1
@@ -525,7 +533,10 @@ subroutine select_singles_and_doubles(i_generator,hole_mask,particle_mask,fock_d
   deallocate(preinteresting, prefullinteresting, interesting, fullinteresting)
   deallocate(banned, bannedOrb,mat)
 end subroutine
-subroutine fill_buffer_double(i_generator, sp, h1, h2, bannedOrb, banned, fock_diag_tmp, E0, pt2_data, mat, buf)
+
+BEGIN_TEMPLATE
+
+subroutine fill_buffer_$DOUBLE(i_generator, sp, h1, h2, bannedOrb, banned, fock_diag_tmp, E0, pt2_data, mat, buf)
   use bitmasks
   use selection_types
   implicit none
@@ -559,7 +570,20 @@ subroutine fill_buffer_double(i_generator, sp, h1, h2, bannedOrb, banned, fock_d
     s1 = sp
     s2 = sp
   end if
-  call apply_holes(psi_det_generators(1,1,i_generator), s1, h1, s2, h2, mask, ok, N_int)
+
+  if ($IS_DOUBLE) then
+    if (h2 == 0) then
+       print *, 'h2=0 in '//trim(irp_here)
+       stop
+    endif
+    call apply_holes(psi_det_generators(1,1,i_generator), s1, h1, s2, h2, mask, ok, N_int)
+  else
+    if (h2 /= 0) then
+       print *, 'h2 /= in '//trim(irp_here)
+       stop
+    endif
+    call apply_hole(psi_det_generators(1,1,i_generator), s1, h1, mask, ok, N_int)
+  endif
   E_shift = 0.d0
 
   if (h0_type == 'CFG') then
@@ -567,12 +591,15 @@ subroutine fill_buffer_double(i_generator, sp, h1, h2, bannedOrb, banned, fock_d
     E_shift = psi_det_Hii(i_generator) - psi_configuration_Hii(j)
   endif
 
-  do p1=1,mo_num
+  $DO_p1
-    if(bannedOrb(p1, s1)) cycle
+! do p1=1,mo_num
+
+    if (bannedOrb(p1, s1)) cycle
     ib = 1
     if(sp /= 3) ib = p1+1
 
-    do p2=ib,mo_num
+    $DO_p2
+  ! do p2=ib,mo_num
 
 ! -----
 ! /!\ Generating only single excited determinants doesn't work because a
@@ -581,9 +608,10 @@ subroutine fill_buffer_double(i_generator, sp, h1, h2, bannedOrb, banned, fock_d
 ! detected as already generated when generating in the future with a
 ! double excitation.
 ! -----
-
+      if ($IS_DOUBLE) then
-      if(bannedOrb(p2, s2)) cycle
+        if(bannedOrb(p2, s2)) cycle
-      if(banned(p1,p2)) cycle
+        if(banned(p1,p2)) cycle
+      endif
 
       if(pseudo_sym)then
         if(dabs(mat(1, p1, p2)).lt.thresh_sym)then
@@ -593,7 +621,11 @@ subroutine fill_buffer_double(i_generator, sp, h1, h2, bannedOrb, banned, fock_d
 
       val = maxval(abs(mat(1:N_states, p1, p2)))
       if( val == 0d0) cycle
-      call apply_particles(mask, s1, p1, s2, p2, det, ok, N_int)
+      if ($IS_DOUBLE) then
+        call apply_particles(mask, s1, p1, s2, p2, det, ok, N_int)
+      else
+        call apply_particle(mask, s1, p1, det, ok, N_int)
+      endif
 
       if (do_only_cas) then
         integer, external :: number_of_holes, number_of_particles
@@ -854,10 +886,19 @@ subroutine fill_buffer_double(i_generator, sp, h1, h2, bannedOrb, banned, fock_d
       if(w <= buf%mini) then
         call add_to_selection_buffer(buf, det, w)
       end if
-    end do
+  ! enddo
-  end do
+    $ENDDO_p1
+! enddo
+  $ENDDO_p2
 end
 
+SUBST [ DOUBLE , DO_p1 , ENDDO_p1 , DO_p2 , ENDDO_p2 , IS_DOUBLE ]
+
+double ; do p1=1,mo_num ; enddo ; do p2=ib,mo_num ; enddo ; .True.  ;;
+single ; do p1=1,mo_num ; enddo ; p2=1            ;       ; .False. ;;
+
+END_TEMPLATE
+
 subroutine splash_pq(mask, sp, det, i_gen, N_sel, bannedOrb, banned, mat, interesting)
   use bitmasks
   implicit none
@@ -879,6 +920,7 @@ subroutine splash_pq(mask, sp, det, i_gen, N_sel, bannedOrb, banned, mat, intere
 
   PROVIDE psi_selectors_coef_transp psi_det_sorted
   mat = 0d0
+  p=0
 
   do i=1,N_int
     negMask(i,1) = not(mask(i,1))
@@ -1432,7 +1474,7 @@ subroutine get_d0(gen, phasemask, bannedOrb, banned, mat, mask, h, p, sp, coefs)
     p1 = p(1,sp)
     p2 = p(2,sp)
     do puti=1, mo_num
-      if(bannedOrb(puti, sp)) cycle
+      if (bannedOrb(puti, sp)) cycle
       call get_mo_two_e_integrals(puti,p2,p1,mo_num,hij_cache1,mo_integrals_map)
       call get_mo_two_e_integrals(puti,p1,p2,mo_num,hij_cache2,mo_integrals_map)
       do putj=puti+1, mo_num
@@ -1443,7 +1485,7 @@ subroutine get_d0(gen, phasemask, bannedOrb, banned, mat, mask, h, p, sp, coefs)
           call i_h_j(gen, det, N_int, hij)
           if (hij == 0.d0) cycle
         else
-          hij = (mo_two_e_integral(p1, p2, puti, putj) -  mo_two_e_integral(p2, p1, puti, putj))
+          hij = hij_cache1(putj) - hij_cache2(putj)
           if (hij == 0.d0) cycle
           hij = hij * get_phase_bi(phasemask, sp, sp, puti, p1 , putj, p2, N_int)
         end if
@@ -1503,7 +1545,7 @@ subroutine spot_isinwf(mask, det, i_gen, N, banned, fullMatch, interesting)
   use bitmasks
   implicit none
   BEGIN_DOC
-! Identify the determinants in det which are in the internal space. These are
+! Identify the determinants in det that are in the internal space. These are
 ! the determinants that can be produced by creating two particles on the mask.
   END_DOC
 
@@ -1531,7 +1573,7 @@ subroutine spot_isinwf(mask, det, i_gen, N, banned, fullMatch, interesting)
       if(iand(det(j,2,i), mask(j,2)) /= mask(j, 2)) cycle genl
     end do
 
-    ! If det(i) < det(i_gen), it hs already been considered
+    ! If det(i) < det(i_gen), it has already been considered
     if(interesting(i) < i_gen) then
       fullMatch = .true.
       return
@@ -1582,352 +1624,4 @@ end
 
 
 
-! OLD unoptimized routines for debugging
-! ======================================
-
-subroutine get_d0_reference(gen, phasemask, bannedOrb, banned, mat, mask, h, p, sp, coefs)
-  use bitmasks
-  implicit none
-
-  integer(bit_kind), intent(in) :: gen(N_int, 2), mask(N_int, 2)
-  integer(bit_kind), intent(in) :: phasemask(N_int,2)
-  logical, intent(in) :: bannedOrb(mo_num, 2), banned(mo_num, mo_num,2)
-  integer(bit_kind) :: det(N_int, 2)
-  double precision, intent(in) :: coefs(N_states)
-  double precision, intent(inout) :: mat(N_states, mo_num, mo_num)
-  integer, intent(in) :: h(0:2,2), p(0:4,2), sp
-
-  integer :: i, j, s, h1, h2, p1, p2, puti, putj
-  double precision :: hij, phase
-  double precision, external :: get_phase_bi, mo_two_e_integral
-  logical :: ok
-
-  integer :: bant
-  bant = 1
-
-
-  if(sp == 3) then ! AB
-    h1 = p(1,1)
-    h2 = p(1,2)
-    do p1=1, mo_num
-      if(bannedOrb(p1, 1)) cycle
-      do p2=1, mo_num
-        if(bannedOrb(p2,2)) cycle
-        if(banned(p1, p2, bant)) cycle ! rentable?
-        if(p1 == h1 .or. p2 == h2) then
-          call apply_particles(mask, 1,p1,2,p2, det, ok, N_int)
-          call i_h_j(gen, det, N_int, hij)
-        else
-          phase = get_phase_bi(phasemask, 1, 2, h1, p1, h2, p2, N_int)
-          hij = mo_two_e_integral(p1, p2, h1, h2) * phase
-        end if
-        mat(:, p1, p2) = mat(:, p1, p2) + coefs(:) * hij
-      end do
-    end do
-  else ! AA BB
-    p1 = p(1,sp)
-    p2 = p(2,sp)
-    do puti=1, mo_num
-      if(bannedOrb(puti, sp)) cycle
-      do putj=puti+1, mo_num
-        if(bannedOrb(putj, sp)) cycle
-        if(banned(puti, putj, bant)) cycle ! rentable?
-        if(puti == p1 .or. putj == p2 .or. puti == p2 .or. putj == p1) then
-          call apply_particles(mask, sp,puti,sp,putj, det, ok, N_int)
-          call i_h_j(gen, det, N_int, hij)
-        else
-          hij = (mo_two_e_integral(p1, p2, puti, putj) -  mo_two_e_integral(p2, p1, puti, putj))* get_phase_bi(phasemask, sp, sp, puti, p1 , putj, p2, N_int)
-        end if
-        mat(:, puti, putj) = mat(:, puti, putj) + coefs(:) * hij
-      end do
-    end do
-  end if
-end
-
-subroutine get_d1_reference(gen, phasemask, bannedOrb, banned, mat, mask, h, p, sp, coefs)
-  use bitmasks
-  implicit none
-
-  integer(bit_kind), intent(in)  :: mask(N_int, 2), gen(N_int, 2)
-  integer(bit_kind), intent(in)  :: phasemask(N_int,2)
-  logical, intent(in)            :: bannedOrb(mo_num, 2), banned(mo_num, mo_num,2)
-  integer(bit_kind)              :: det(N_int, 2)
-  double precision, intent(in)   :: coefs(N_states)
-  double precision, intent(inout) :: mat(N_states, mo_num, mo_num)
-  integer, intent(in)            :: h(0:2,2), p(0:4,2), sp
-  double precision               :: hij, tmp_row(N_states, mo_num), tmp_row2(N_states, mo_num)
-  double precision, external     :: get_phase_bi, mo_two_e_integral
-  logical                        :: ok
-
-  logical, allocatable           :: lbanned(:,:)
-  integer                        :: puti, putj, ma, mi, s1, s2, i, i1, i2, j
-  integer                        :: hfix, pfix, h1, h2, p1, p2, ib
-
-  integer, parameter             :: turn2(2) = (/2,1/)
-  integer, parameter             :: turn3(2,3) = reshape((/2,3,  1,3, 1,2/), (/2,3/))
-
-  integer                        :: bant
-
-
-  allocate (lbanned(mo_num, 2))
-  lbanned = bannedOrb
-
-  do i=1, p(0,1)
-    lbanned(p(i,1), 1) = .true.
-  end do
-  do i=1, p(0,2)
-    lbanned(p(i,2), 2) = .true.
-  end do
-
-  ma = 1
-  if(p(0,2) >= 2) ma = 2
-  mi = turn2(ma)
-
-  bant = 1
-
-  if(sp == 3) then
-    !move MA
-    if(ma == 2) bant = 2
-    puti = p(1,mi)
-    hfix = h(1,ma)
-    p1 = p(1,ma)
-    p2 = p(2,ma)
-    if(.not. bannedOrb(puti, mi)) then
-      tmp_row = 0d0
-      do putj=1, hfix-1
-        if(lbanned(putj, ma) .or. banned(putj, puti,bant)) cycle
-        hij = (mo_two_e_integral(p1, p2, putj, hfix)-mo_two_e_integral(p2,p1,putj,hfix)) * get_phase_bi(phasemask, ma, ma, putj, p1, hfix, p2, N_int)
-        tmp_row(1:N_states,putj) = tmp_row(1:N_states,putj) + hij * coefs(1:N_states)
-      end do
-      do putj=hfix+1, mo_num
-        if(lbanned(putj, ma) .or. banned(putj, puti,bant)) cycle
-        hij = (mo_two_e_integral(p1, p2, hfix, putj)-mo_two_e_integral(p2,p1,hfix,putj)) * get_phase_bi(phasemask, ma, ma, hfix, p1, putj, p2, N_int)
-        tmp_row(1:N_states,putj) = tmp_row(1:N_states,putj) + hij * coefs(1:N_states)
-      end do
-
-      if(ma == 1) then
-        mat(1:N_states,1:mo_num,puti) = mat(1:N_states,1:mo_num,puti) + tmp_row(1:N_states,1:mo_num)
-      else
-        mat(1:N_states,puti,1:mo_num) = mat(1:N_states,puti,1:mo_num) + tmp_row(1:N_states,1:mo_num)
-      end if
-    end if
-
-    !MOVE MI
-    pfix = p(1,mi)
-    tmp_row = 0d0
-    tmp_row2 = 0d0
-    do puti=1,mo_num
-      if(lbanned(puti,mi)) cycle
-      !p1 fixed
-      putj = p1
-      if(.not. banned(putj,puti,bant)) then
-        hij = mo_two_e_integral(p2,pfix,hfix,puti) * get_phase_bi(phasemask, ma, mi, hfix, p2, puti, pfix, N_int)
-        tmp_row(:,puti) = tmp_row(:,puti) + hij * coefs(:)
-      end if
-
-      putj = p2
-      if(.not. banned(putj,puti,bant)) then
-        hij = mo_two_e_integral(p1,pfix,hfix,puti) * get_phase_bi(phasemask, ma, mi, hfix, p1, puti, pfix, N_int)
-        tmp_row2(:,puti) = tmp_row2(:,puti) + hij * coefs(:)
-      end if
-    end do
-
-    if(mi == 1) then
-      mat(:,:,p1) = mat(:,:,p1) + tmp_row(:,:)
-      mat(:,:,p2) = mat(:,:,p2) + tmp_row2(:,:)
-    else
-      mat(:,p1,:) = mat(:,p1,:) + tmp_row(:,:)
-      mat(:,p2,:) = mat(:,p2,:) + tmp_row2(:,:)
-    end if
-  else
-    if(p(0,ma) == 3) then
-      do i=1,3
-        hfix = h(1,ma)
-        puti = p(i, ma)
-        p1 = p(turn3(1,i), ma)
-        p2 = p(turn3(2,i), ma)
-        tmp_row = 0d0
-        do putj=1,hfix-1
-          if(lbanned(putj,ma) .or. banned(puti,putj,1)) cycle
-          hij = (mo_two_e_integral(p1, p2, putj, hfix)-mo_two_e_integral(p2,p1,putj,hfix)) * get_phase_bi(phasemask, ma, ma, putj, p1, hfix, p2, N_int)
-          tmp_row(:,putj) = tmp_row(:,putj) + hij * coefs(:)
-        end do
-        do putj=hfix+1,mo_num
-          if(lbanned(putj,ma) .or. banned(puti,putj,1)) cycle
-          hij = (mo_two_e_integral(p1, p2, hfix, putj)-mo_two_e_integral(p2,p1,hfix,putj)) * get_phase_bi(phasemask, ma, ma, hfix, p1, putj, p2, N_int)
-          tmp_row(:,putj) = tmp_row(:,putj) + hij * coefs(:)
-        end do
-
-        mat(:, :puti-1, puti) = mat(:, :puti-1, puti) + tmp_row(:,:puti-1)
-        mat(:, puti, puti:) = mat(:, puti, puti:) + tmp_row(:,puti:)
-      end do
-    else
-      hfix = h(1,mi)
-      pfix = p(1,mi)
-      p1 = p(1,ma)
-      p2 = p(2,ma)
-      tmp_row = 0d0
-      tmp_row2 = 0d0
-      do puti=1,mo_num
-        if(lbanned(puti,ma)) cycle
-        putj = p2
-        if(.not. banned(puti,putj,1)) then
-          hij = mo_two_e_integral(pfix, p1, hfix, puti) * get_phase_bi(phasemask, mi, ma, hfix, pfix, puti, p1, N_int)
-          tmp_row(:,puti) = tmp_row(:,puti) + hij * coefs(:)
-        end if
-
-        putj = p1
-        if(.not. banned(puti,putj,1)) then
-          hij = mo_two_e_integral(pfix, p2, hfix, puti) * get_phase_bi(phasemask, mi, ma, hfix, pfix, puti, p2, N_int)
-          tmp_row2(:,puti) = tmp_row2(:,puti) + hij * coefs(:)
-        end if
-      end do
-      mat(:,:p2-1,p2) = mat(:,:p2-1,p2) + tmp_row(:,:p2-1)
-      mat(:,p2,p2:) = mat(:,p2,p2:) + tmp_row(:,p2:)
-      mat(:,:p1-1,p1) = mat(:,:p1-1,p1) + tmp_row2(:,:p1-1)
-      mat(:,p1,p1:) = mat(:,p1,p1:) + tmp_row2(:,p1:)
-    end if
-  end if
-  deallocate(lbanned)
-
- !! MONO
-    if(sp == 3) then
-      s1 = 1
-      s2 = 2
-    else
-      s1 = sp
-      s2 = sp
-    end if
-
-    do i1=1,p(0,s1)
-      ib = 1
-      if(s1 == s2) ib = i1+1
-      do i2=ib,p(0,s2)
-        p1 = p(i1,s1)
-        p2 = p(i2,s2)
-        if(bannedOrb(p1, s1) .or. bannedOrb(p2, s2) .or. banned(p1, p2, 1)) cycle
-        call apply_particles(mask, s1, p1, s2, p2, det, ok, N_int)
-        call i_h_j(gen, det, N_int, hij)
-        mat(:, p1, p2) = mat(:, p1, p2) + coefs(:) * hij
-      end do
-    end do
-end
-
-subroutine get_d2_reference(gen, phasemask, bannedOrb, banned, mat, mask, h, p, sp, coefs)
-  use bitmasks
-  implicit none
-
-  integer(bit_kind), intent(in) :: mask(N_int, 2), gen(N_int, 2)
-  integer(bit_kind), intent(in) :: phasemask(2,N_int)
-  logical, intent(in) :: bannedOrb(mo_num, 2), banned(mo_num, mo_num,2)
-  double precision, intent(in) :: coefs(N_states)
-  double precision, intent(inout) :: mat(N_states, mo_num, mo_num)
-  integer, intent(in) :: h(0:2,2), p(0:4,2), sp
-
-  double precision, external :: get_phase_bi, mo_two_e_integral
-
-  integer :: i, j, tip, ma, mi, puti, putj
-  integer :: h1, h2, p1, p2, i1, i2
-  double precision :: hij, phase
-
-  integer, parameter:: turn2d(2,3,4) = reshape((/0,0, 0,0, 0,0,  3,4, 0,0, 0,0,  2,4, 1,4, 0,0,  2,3, 1,3, 1,2 /), (/2,3,4/))
-  integer, parameter :: turn2(2) = (/2, 1/)
-  integer, parameter :: turn3(2,3) = reshape((/2,3,  1,3, 1,2/), (/2,3/))
-
-  integer :: bant
-  bant = 1
-
-  tip = p(0,1) * p(0,2)
-
-  ma = sp
-  if(p(0,1) > p(0,2)) ma = 1
-  if(p(0,1) < p(0,2)) ma = 2
-  mi = mod(ma, 2) + 1
-
-  if(sp == 3) then
-    if(ma == 2) bant = 2
-
-    if(tip == 3) then
-      puti = p(1, mi)
-      do i = 1, 3
-        putj = p(i, ma)
-        if(banned(putj,puti,bant)) cycle
-        i1 = turn3(1,i)
-        i2 = turn3(2,i)
-        p1 = p(i1, ma)
-        p2 = p(i2, ma)
-        h1 = h(1, ma)
-        h2 = h(2, ma)
-
-        hij = (mo_two_e_integral(p1, p2, h1, h2) - mo_two_e_integral(p2,p1, h1, h2)) * get_phase_bi(phasemask, ma, ma, h1, p1, h2, p2, N_int)
-        if(ma == 1) then
-          mat(:, putj, puti) = mat(:, putj, puti) + coefs(:) * hij
-        else
-          mat(:, puti, putj) = mat(:, puti, putj) + coefs(:) * hij
-        end if
-      end do
-    else
-      h1 = h(1,1)
-      h2 = h(1,2)
-      do j = 1,2
-        putj = p(j, 2)
-        p2 = p(turn2(j), 2)
-        do i = 1,2
-          puti = p(i, 1)
-
-          if(banned(puti,putj,bant)) cycle
-          p1 = p(turn2(i), 1)
-
-          hij = mo_two_e_integral(p1, p2, h1, h2) * get_phase_bi(phasemask, 1, 2, h1, p1, h2, p2,N_int)
-          mat(:, puti, putj) = mat(:, puti, putj) + coefs(:) * hij
-        end do
-      end do
-    end if
-
-  else
-    if(tip == 0) then
-      h1 = h(1, ma)
-      h2 = h(2, ma)
-      do i=1,3
-      puti = p(i, ma)
-      do j=i+1,4
-        putj = p(j, ma)
-        if(banned(puti,putj,1)) cycle
-
-        i1 = turn2d(1, i, j)
-        i2 = turn2d(2, i, j)
-        p1 = p(i1, ma)
-        p2 = p(i2, ma)
-        hij = (mo_two_e_integral(p1, p2, h1, h2) - mo_two_e_integral(p2,p1, h1, h2)) * get_phase_bi(phasemask, ma, ma, h1, p1, h2, p2,N_int)
-        mat(:, puti, putj) = mat(:, puti, putj) + coefs(:) * hij
-      end do
-      end do
-    else if(tip == 3) then
-      h1 = h(1, mi)
-      h2 = h(1, ma)
-      p1 = p(1, mi)
-      do i=1,3
-        puti = p(turn3(1,i), ma)
-        putj = p(turn3(2,i), ma)
-        if(banned(puti,putj,1)) cycle
-        p2 = p(i, ma)
-
-        hij = mo_two_e_integral(p1, p2, h1, h2) * get_phase_bi(phasemask, mi, ma, h1, p1, h2, p2,N_int)
-        mat(:, min(puti, putj), max(puti, putj)) = mat(:, min(puti, putj), max(puti, putj)) + coefs(:) * hij
-      end do
-    else ! tip == 4
-      puti = p(1, sp)
-      putj = p(2, sp)
-      if(.not. banned(puti,putj,1)) then
-        p1 = p(1, mi)
-        p2 = p(2, mi)
-        h1 = h(1, mi)
-        h2 = h(2, mi)
-        hij = (mo_two_e_integral(p1, p2, h1, h2) - mo_two_e_integral(p2,p1, h1, h2)) * get_phase_bi(phasemask, mi, mi, h1, p1, h2, p2,N_int)
-        mat(:, puti, putj) = mat(:, puti, putj) + coefs(:) * hij
-      end if
-    end if
-  end if
-end
-
 

src/cipsi/selection_old.irp.f

@@ -0,0 +1,350 @@
+
+! OLD unoptimized routines for debugging
+! ======================================
+
+subroutine get_d0_reference(gen, phasemask, bannedOrb, banned, mat, mask, h, p, sp, coefs)
+  use bitmasks
+  implicit none
+
+  integer(bit_kind), intent(in) :: gen(N_int, 2), mask(N_int, 2)
+  integer(bit_kind), intent(in) :: phasemask(N_int,2)
+  logical, intent(in) :: bannedOrb(mo_num, 2), banned(mo_num, mo_num,2)
+  integer(bit_kind) :: det(N_int, 2)
+  double precision, intent(in) :: coefs(N_states)
+  double precision, intent(inout) :: mat(N_states, mo_num, mo_num)
+  integer, intent(in) :: h(0:2,2), p(0:4,2), sp
+
+  integer :: i, j, s, h1, h2, p1, p2, puti, putj
+  double precision :: hij, phase
+  double precision, external :: get_phase_bi, mo_two_e_integral
+  logical :: ok
+
+  integer :: bant
+  bant = 1
+
+
+  if(sp == 3) then ! AB
+    h1 = p(1,1)
+    h2 = p(1,2)
+    do p1=1, mo_num
+      if(bannedOrb(p1, 1)) cycle
+      do p2=1, mo_num
+        if(bannedOrb(p2,2)) cycle
+        if(banned(p1, p2, bant)) cycle ! rentable?
+        if(p1 == h1 .or. p2 == h2) then
+          call apply_particles(mask, 1,p1,2,p2, det, ok, N_int)
+          call i_h_j(gen, det, N_int, hij)
+        else
+          phase = get_phase_bi(phasemask, 1, 2, h1, p1, h2, p2, N_int)
+          hij = mo_two_e_integral(p1, p2, h1, h2) * phase
+        end if
+        mat(:, p1, p2) = mat(:, p1, p2) + coefs(:) * hij
+      end do
+    end do
+  else ! AA BB
+    p1 = p(1,sp)
+    p2 = p(2,sp)
+    do puti=1, mo_num
+!      do not cycle here? otherwise singles will be missed??
+      if(bannedOrb(puti, sp)) cycle
+      do putj=puti+1, mo_num
+        if(bannedOrb(putj, sp)) cycle
+        if(banned(puti, putj, bant)) cycle ! rentable?
+        if(puti == p1 .or. putj == p2 .or. puti == p2 .or. putj == p1) then
+          call apply_particles(mask, sp,puti,sp,putj, det, ok, N_int)
+          call i_h_j(gen, det, N_int, hij)
+        else
+          hij = (mo_two_e_integral(p1, p2, puti, putj) -  mo_two_e_integral(p2, p1, puti, putj))* get_phase_bi(phasemask, sp, sp, puti, p1 , putj, p2, N_int)
+        end if
+        mat(:, puti, putj) = mat(:, puti, putj) + coefs(:) * hij
+      end do
+    end do
+  end if
+end
+
+subroutine get_d1_reference(gen, phasemask, bannedOrb, banned, mat, mask, h, p, sp, coefs)
+  use bitmasks
+  implicit none
+
+  integer(bit_kind), intent(in)  :: mask(N_int, 2), gen(N_int, 2)
+  integer(bit_kind), intent(in)  :: phasemask(N_int,2)
+  logical, intent(in)            :: bannedOrb(mo_num, 2), banned(mo_num, mo_num,2)
+  integer(bit_kind)              :: det(N_int, 2)
+  double precision, intent(in)   :: coefs(N_states)
+  double precision, intent(inout) :: mat(N_states, mo_num, mo_num)
+  integer, intent(in)            :: h(0:2,2), p(0:4,2), sp
+  double precision               :: hij, tmp_row(N_states, mo_num), tmp_row2(N_states, mo_num)
+  double precision, external     :: get_phase_bi, mo_two_e_integral
+  logical                        :: ok
+
+  logical, allocatable           :: lbanned(:,:)
+  integer                        :: puti, putj, ma, mi, s1, s2, i, i1, i2, j
+  integer                        :: hfix, pfix, h1, h2, p1, p2, ib
+
+  integer, parameter             :: turn2(2) = (/2,1/)
+  integer, parameter             :: turn3(2,3) = reshape((/2,3,  1,3, 1,2/), (/2,3/))
+
+  integer                        :: bant
+
+
+  allocate (lbanned(mo_num, 2))
+  lbanned = bannedOrb
+
+  do i=1, p(0,1)
+    lbanned(p(i,1), 1) = .true.
+  end do
+  do i=1, p(0,2)
+    lbanned(p(i,2), 2) = .true.
+  end do
+
+  ma = 1
+  if(p(0,2) >= 2) ma = 2
+  mi = turn2(ma)
+
+  bant = 1
+
+  if(sp == 3) then
+    !move MA
+    if(ma == 2) bant = 2
+    puti = p(1,mi)
+    hfix = h(1,ma)
+    p1 = p(1,ma)
+    p2 = p(2,ma)
+    if(.not. bannedOrb(puti, mi)) then
+      tmp_row = 0d0
+      do putj=1, hfix-1
+        if(lbanned(putj, ma) .or. banned(putj, puti,bant)) cycle
+        hij = (mo_two_e_integral(p1, p2, putj, hfix)-mo_two_e_integral(p2,p1,putj,hfix)) * get_phase_bi(phasemask, ma, ma, putj, p1, hfix, p2, N_int)
+        tmp_row(1:N_states,putj) = tmp_row(1:N_states,putj) + hij * coefs(1:N_states)
+      end do
+      do putj=hfix+1, mo_num
+        if(lbanned(putj, ma) .or. banned(putj, puti,bant)) cycle
+        hij = (mo_two_e_integral(p1, p2, hfix, putj)-mo_two_e_integral(p2,p1,hfix,putj)) * get_phase_bi(phasemask, ma, ma, hfix, p1, putj, p2, N_int)
+        tmp_row(1:N_states,putj) = tmp_row(1:N_states,putj) + hij * coefs(1:N_states)
+      end do
+
+      if(ma == 1) then
+        mat(1:N_states,1:mo_num,puti) = mat(1:N_states,1:mo_num,puti) + tmp_row(1:N_states,1:mo_num)
+      else
+        mat(1:N_states,puti,1:mo_num) = mat(1:N_states,puti,1:mo_num) + tmp_row(1:N_states,1:mo_num)
+      end if
+    end if
+
+    !MOVE MI
+    pfix = p(1,mi)
+    tmp_row = 0d0
+    tmp_row2 = 0d0
+    do puti=1,mo_num
+      if(lbanned(puti,mi)) cycle
+      !p1 fixed
+      putj = p1
+      if(.not. banned(putj,puti,bant)) then
+        hij = mo_two_e_integral(p2,pfix,hfix,puti) * get_phase_bi(phasemask, ma, mi, hfix, p2, puti, pfix, N_int)
+        tmp_row(:,puti) = tmp_row(:,puti) + hij * coefs(:)
+      end if
+
+      putj = p2
+      if(.not. banned(putj,puti,bant)) then
+        hij = mo_two_e_integral(p1,pfix,hfix,puti) * get_phase_bi(phasemask, ma, mi, hfix, p1, puti, pfix, N_int)
+        tmp_row2(:,puti) = tmp_row2(:,puti) + hij * coefs(:)
+      end if
+    end do
+
+    if(mi == 1) then
+      mat(:,:,p1) = mat(:,:,p1) + tmp_row(:,:)
+      mat(:,:,p2) = mat(:,:,p2) + tmp_row2(:,:)
+    else
+      mat(:,p1,:) = mat(:,p1,:) + tmp_row(:,:)
+      mat(:,p2,:) = mat(:,p2,:) + tmp_row2(:,:)
+    end if
+  else
+    if(p(0,ma) == 3) then
+      do i=1,3
+        hfix = h(1,ma)
+        puti = p(i, ma)
+        p1 = p(turn3(1,i), ma)
+        p2 = p(turn3(2,i), ma)
+        tmp_row = 0d0
+        do putj=1,hfix-1
+          if(lbanned(putj,ma) .or. banned(puti,putj,1)) cycle
+          hij = (mo_two_e_integral(p1, p2, putj, hfix)-mo_two_e_integral(p2,p1,putj,hfix)) * get_phase_bi(phasemask, ma, ma, putj, p1, hfix, p2, N_int)
+          tmp_row(:,putj) = tmp_row(:,putj) + hij * coefs(:)
+        end do
+        do putj=hfix+1,mo_num
+          if(lbanned(putj,ma) .or. banned(puti,putj,1)) cycle
+          hij = (mo_two_e_integral(p1, p2, hfix, putj)-mo_two_e_integral(p2,p1,hfix,putj)) * get_phase_bi(phasemask, ma, ma, hfix, p1, putj, p2, N_int)
+          tmp_row(:,putj) = tmp_row(:,putj) + hij * coefs(:)
+        end do
+
+        mat(:, :puti-1, puti) = mat(:, :puti-1, puti) + tmp_row(:,:puti-1)
+        mat(:, puti, puti:) = mat(:, puti, puti:) + tmp_row(:,puti:)
+      end do
+    else
+      hfix = h(1,mi)
+      pfix = p(1,mi)
+      p1 = p(1,ma)
+      p2 = p(2,ma)
+      tmp_row = 0d0
+      tmp_row2 = 0d0
+      do puti=1,mo_num
+        if(lbanned(puti,ma)) cycle
+        putj = p2
+        if(.not. banned(puti,putj,1)) then
+          hij = mo_two_e_integral(pfix, p1, hfix, puti) * get_phase_bi(phasemask, mi, ma, hfix, pfix, puti, p1, N_int)
+          tmp_row(:,puti) = tmp_row(:,puti) + hij * coefs(:)
+        end if
+
+        putj = p1
+        if(.not. banned(puti,putj,1)) then
+          hij = mo_two_e_integral(pfix, p2, hfix, puti) * get_phase_bi(phasemask, mi, ma, hfix, pfix, puti, p2, N_int)
+          tmp_row2(:,puti) = tmp_row2(:,puti) + hij * coefs(:)
+        end if
+      end do
+      mat(:,:p2-1,p2) = mat(:,:p2-1,p2) + tmp_row(:,:p2-1)
+      mat(:,p2,p2:) = mat(:,p2,p2:) + tmp_row(:,p2:)
+      mat(:,:p1-1,p1) = mat(:,:p1-1,p1) + tmp_row2(:,:p1-1)
+      mat(:,p1,p1:) = mat(:,p1,p1:) + tmp_row2(:,p1:)
+    end if
+  end if
+  deallocate(lbanned)
+
+ !! MONO
+    if(sp == 3) then
+      s1 = 1
+      s2 = 2
+    else
+      s1 = sp
+      s2 = sp
+    end if
+
+    do i1=1,p(0,s1)
+      ib = 1
+      if(s1 == s2) ib = i1+1
+      do i2=ib,p(0,s2)
+        p1 = p(i1,s1)
+        p2 = p(i2,s2)
+        if(bannedOrb(p1, s1) .or. bannedOrb(p2, s2) .or. banned(p1, p2, 1)) cycle
+        call apply_particles(mask, s1, p1, s2, p2, det, ok, N_int)
+        call i_h_j(gen, det, N_int, hij)
+        mat(:, p1, p2) = mat(:, p1, p2) + coefs(:) * hij
+      end do
+    end do
+end
+
+subroutine get_d2_reference(gen, phasemask, bannedOrb, banned, mat, mask, h, p, sp, coefs)
+  use bitmasks
+  implicit none
+
+  integer(bit_kind), intent(in) :: mask(N_int, 2), gen(N_int, 2)
+  integer(bit_kind), intent(in) :: phasemask(2,N_int)
+  logical, intent(in) :: bannedOrb(mo_num, 2), banned(mo_num, mo_num,2)
+  double precision, intent(in) :: coefs(N_states)
+  double precision, intent(inout) :: mat(N_states, mo_num, mo_num)
+  integer, intent(in) :: h(0:2,2), p(0:4,2), sp
+
+  double precision, external :: get_phase_bi, mo_two_e_integral
+
+  integer :: i, j, tip, ma, mi, puti, putj
+  integer :: h1, h2, p1, p2, i1, i2
+  double precision :: hij, phase
+
+  integer, parameter:: turn2d(2,3,4) = reshape((/0,0, 0,0, 0,0,  3,4, 0,0, 0,0,  2,4, 1,4, 0,0,  2,3, 1,3, 1,2 /), (/2,3,4/))
+  integer, parameter :: turn2(2) = (/2, 1/)
+  integer, parameter :: turn3(2,3) = reshape((/2,3,  1,3, 1,2/), (/2,3/))
+
+  integer :: bant
+  bant = 1
+
+  tip = p(0,1) * p(0,2)
+
+  ma = sp
+  if(p(0,1) > p(0,2)) ma = 1
+  if(p(0,1) < p(0,2)) ma = 2
+  mi = mod(ma, 2) + 1
+
+  if(sp == 3) then
+    if(ma == 2) bant = 2
+
+    if(tip == 3) then
+      puti = p(1, mi)
+      do i = 1, 3
+        putj = p(i, ma)
+        if(banned(putj,puti,bant)) cycle
+        i1 = turn3(1,i)
+        i2 = turn3(2,i)
+        p1 = p(i1, ma)
+        p2 = p(i2, ma)
+        h1 = h(1, ma)
+        h2 = h(2, ma)
+
+        hij = (mo_two_e_integral(p1, p2, h1, h2) - mo_two_e_integral(p2,p1, h1, h2)) * get_phase_bi(phasemask, ma, ma, h1, p1, h2, p2, N_int)
+        if(ma == 1) then
+          mat(:, putj, puti) = mat(:, putj, puti) + coefs(:) * hij
+        else
+          mat(:, puti, putj) = mat(:, puti, putj) + coefs(:) * hij
+        end if
+      end do
+    else
+      h1 = h(1,1)
+      h2 = h(1,2)
+      do j = 1,2
+        putj = p(j, 2)
+        p2 = p(turn2(j), 2)
+        do i = 1,2
+          puti = p(i, 1)
+
+          if(banned(puti,putj,bant)) cycle
+          p1 = p(turn2(i), 1)
+
+          hij = mo_two_e_integral(p1, p2, h1, h2) * get_phase_bi(phasemask, 1, 2, h1, p1, h2, p2,N_int)
+          mat(:, puti, putj) = mat(:, puti, putj) + coefs(:) * hij
+        end do
+      end do
+    end if
+
+  else
+    if(tip == 0) then
+      h1 = h(1, ma)
+      h2 = h(2, ma)
+      do i=1,3
+      puti = p(i, ma)
+      do j=i+1,4
+        putj = p(j, ma)
+        if(banned(puti,putj,1)) cycle
+
+        i1 = turn2d(1, i, j)
+        i2 = turn2d(2, i, j)
+        p1 = p(i1, ma)
+        p2 = p(i2, ma)
+        hij = (mo_two_e_integral(p1, p2, h1, h2) - mo_two_e_integral(p2,p1, h1, h2)) * get_phase_bi(phasemask, ma, ma, h1, p1, h2, p2,N_int)
+        mat(:, puti, putj) = mat(:, puti, putj) + coefs(:) * hij
+      end do
+      end do
+    else if(tip == 3) then
+      h1 = h(1, mi)
+      h2 = h(1, ma)
+      p1 = p(1, mi)
+      do i=1,3
+        puti = p(turn3(1,i), ma)
+        putj = p(turn3(2,i), ma)
+        if(banned(puti,putj,1)) cycle
+        p2 = p(i, ma)
+
+        hij = mo_two_e_integral(p1, p2, h1, h2) * get_phase_bi(phasemask, mi, ma, h1, p1, h2, p2,N_int)
+        mat(:, min(puti, putj), max(puti, putj)) = mat(:, min(puti, putj), max(puti, putj)) + coefs(:) * hij
+      end do
+    else ! tip == 4
+      puti = p(1, sp)
+      putj = p(2, sp)
+      if(.not. banned(puti,putj,1)) then
+        p1 = p(1, mi)
+        p2 = p(2, mi)
+        h1 = h(1, mi)
+        h2 = h(2, mi)
+        hij = (mo_two_e_integral(p1, p2, h1, h2) - mo_two_e_integral(p2,p1, h1, h2)) * get_phase_bi(phasemask, mi, mi, h1, p1, h2, p2,N_int)
+        mat(:, puti, putj) = mat(:, puti, putj) + coefs(:) * hij
+      end if
+    end if
+  end if
+end
+

src/cipsi/selection_singles.irp.f

@@ -0,0 +1,356 @@
+use bitmasks
+
+subroutine select_singles(i_gen,hole_mask,particle_mask,fock_diag_tmp,E0,pt2_data,buf)
+  use bitmasks
+  use selection_types
+  implicit none
+  BEGIN_DOC
+! Select determinants connected to i_det by H
+  END_DOC
+  integer, intent(in)             :: i_gen
+  integer(bit_kind), intent(in)   :: hole_mask(N_int,2), particle_mask(N_int,2)
+  double precision, intent(in)    :: fock_diag_tmp(mo_num)
+  double precision, intent(in)    :: E0(N_states)
+  type(pt2_type),   intent(inout) :: pt2_data
+  type(selection_buffer), intent(inout) :: buf
+
+  logical, allocatable            :: banned(:,:), bannedOrb(:)
+  double precision, allocatable   :: mat(:,:,:)
+  integer                         :: i, j, k
+  integer                         :: h1,h2,s1,s2,i1,i2,ib,sp
+  integer(bit_kind)               :: hole(N_int,2), particle(N_int,2), mask(N_int, 2)
+  logical                         :: fullMatch, ok
+
+
+  do k=1,N_int
+    hole    (k,1) = iand(psi_det_generators(k,1,i_gen), hole_mask(k,1))
+    hole    (k,2) = iand(psi_det_generators(k,2,i_gen), hole_mask(k,2))
+    particle(k,1) = iand(not(psi_det_generators(k,1,i_gen)), particle_mask(k,1))
+    particle(k,2) = iand(not(psi_det_generators(k,2,i_gen)), particle_mask(k,2))
+  enddo
+
+  allocate(banned(mo_num,mo_num), bannedOrb(mo_num), mat(N_states, mo_num, 1))
+  banned = .False.
+
+  ! Create lists of holes and particles
+  ! -----------------------------------
+
+  integer                        :: N_holes(2), N_particles(2)
+  integer                        :: hole_list(N_int*bit_kind_size,2)
+  integer                        :: particle_list(N_int*bit_kind_size,2)
+
+  call bitstring_to_list_ab(hole    , hole_list    , N_holes    , N_int)
+  call bitstring_to_list_ab(particle, particle_list, N_particles, N_int)
+
+  do sp=1,2
+    do i=1, N_holes(sp)
+      h1 = hole_list(i,sp)
+      call apply_hole(psi_det_generators(1,1,i_gen), sp, h1, mask, ok, N_int)
+      bannedOrb = .true.
+      do j=1,N_particles(sp)
+        bannedOrb(particle_list(j, sp)) = .false.
+      end do
+      call spot_hasBeen(mask, sp, psi_det_sorted, i_gen, N_det, bannedOrb, fullMatch)
+      if(fullMatch) cycle
+      mat = 0d0
+      call splash_p(mask, sp, psi_selectors(1,1,i_gen), psi_selectors_coef_transp(1,i_gen), N_det_selectors - i_gen + 1, bannedOrb, mat(1,1,1))
+      call fill_buffer_single(i_gen, sp, h1, 0, bannedOrb, banned, fock_diag_tmp, E0, pt2_data, mat, buf)
+    end do
+  enddo
+end subroutine
+
+
+subroutine spot_hasBeen(mask, sp, det, i_gen, N, banned, fullMatch)
+  use bitmasks
+  implicit none
+
+  integer(bit_kind),intent(in) :: mask(N_int, 2), det(N_int, 2, N)
+  integer, intent(in) :: i_gen, N, sp
+  logical, intent(inout) :: banned(mo_num)
+  logical, intent(out) :: fullMatch
+
+
+  integer :: i, j, na, nb, list(3), nt
+  integer(bit_kind) :: myMask(N_int, 2), negMask(N_int, 2)
+
+  fullMatch = .false.
+
+  do i=1,N_int
+    negMask(i,1) = not(mask(i,1))
+    negMask(i,2) = not(mask(i,2))
+  end do
+
+  genl : do i=1, N
+    nt = 0
+
+    do j=1, N_int
+      myMask(j, 1) = iand(det(j, 1, i), negMask(j, 1))
+      myMask(j, 2) = iand(det(j, 2, i), negMask(j, 2))
+      nt += popcnt(myMask(j, 1)) + popcnt(myMask(j, 2))
+    end do
+
+    if(nt > 3) cycle
+
+    if(nt <= 2 .and. i < i_gen) then
+      fullMatch = .true.
+      return
+    end if
+
+    call bitstring_to_list(myMask(1,sp), list(1), na, N_int)
+
+    if(nt == 3 .and. i < i_gen) then
+      do j=1,na
+        banned(list(j)) = .true.
+      end do
+    else if(nt == 1 .and. na == 1) then
+      banned(list(1)) = .true.
+    end if
+  end do genl
+end subroutine
+
+
+subroutine splash_p(mask, sp, det, coefs, N_sel, bannedOrb, vect)
+  use bitmasks
+  implicit none
+
+  integer(bit_kind),intent(in) :: mask(N_int, 2), det(N_int,2,N_sel)
+  double precision, intent(in) :: coefs(N_states, N_sel)
+  integer, intent(in) :: sp, N_sel
+  logical, intent(inout) :: bannedOrb(mo_num)
+  double precision, intent(inout)     :: vect(N_states, mo_num)
+
+  integer :: i, j, h(0:2,2), p(0:3,2), nt
+  integer(bit_kind) :: perMask(N_int, 2), mobMask(N_int, 2), negMask(N_int, 2)
+  integer(bit_kind) :: phasemask(N_int, 2)
+
+  do i=1,N_int
+    negMask(i,1) = not(mask(i,1))
+    negMask(i,2) = not(mask(i,2))
+  end do
+
+  do i=1, N_sel
+    nt = 0
+    do j=1,N_int
+      mobMask(j,1) = iand(negMask(j,1), det(j,1,i))
+      mobMask(j,2) = iand(negMask(j,2), det(j,2,i))
+      nt += popcnt(mobMask(j, 1)) + popcnt(mobMask(j, 2))
+    end do
+
+    if(nt > 3) cycle
+
+    do j=1,N_int
+      perMask(j,1) = iand(mask(j,1), not(det(j,1,i)))
+      perMask(j,2) = iand(mask(j,2), not(det(j,2,i)))
+    end do
+
+    call bitstring_to_list(perMask(1,1), h(1,1), h(0,1), N_int)
+    call bitstring_to_list(perMask(1,2), h(1,2), h(0,2), N_int)
+
+    call bitstring_to_list(mobMask(1,1), p(1,1), p(0,1), N_int)
+    call bitstring_to_list(mobMask(1,2), p(1,2), p(0,2), N_int)
+
+    call get_mask_phase(psi_det_sorted(1,1,i), phasemask, N_int)
+
+    if(nt == 3) then
+      call get_m2(det(1,1,i), phasemask, bannedOrb, vect, mask, h, p, sp, coefs(1, i))
+    else if(nt == 2) then
+      call get_m1(det(1,1,i), phasemask, bannedOrb, vect, mask, h, p, sp, coefs(1, i))
+    else
+      call get_m0(det(1,1,i), phasemask, bannedOrb, vect, mask, h, p, sp, coefs(1, i))
+    end if
+  end do
+end subroutine
+
+
+subroutine get_m2(gen, phasemask, bannedOrb, vect, mask, h, p, sp, coefs)
+  use bitmasks
+  implicit none
+
+  integer(bit_kind), intent(in) :: gen(N_int, 2), mask(N_int, 2)
+  integer(bit_kind), intent(in) :: phasemask(N_int, 2)
+  logical, intent(in) :: bannedOrb(mo_num)
+  double precision, intent(in) :: coefs(N_states)
+  double precision, intent(inout) :: vect(N_states, mo_num)
+  integer, intent(in) :: sp, h(0:2, 2), p(0:3, 2)
+  integer :: i, j, h1, h2, p1, p2, sfix, hfix, pfix, hmob, pmob, puti
+  double precision :: hij
+  double precision, external :: get_phase_bi, mo_two_e_integral
+
+  integer, parameter :: turn3_2(2,3) = reshape((/2,3, 1,3, 1,2/), (/2,3/))
+  integer, parameter :: turn2(2) = (/2,1/)
+
+  if(h(0,sp) == 2) then
+    h1 = h(1, sp)
+    h2 = h(2, sp)
+    do i=1,3
+      puti = p(i, sp)
+      if(bannedOrb(puti)) cycle
+      p1 = p(turn3_2(1,i), sp)
+      p2 = p(turn3_2(2,i), sp)
+      hij = mo_two_e_integral(p1, p2, h1, h2) - mo_two_e_integral(p2, p1, h1, h2)
+      hij *= get_phase_bi(phasemask, sp, sp, h1, p1, h2, p2)
+      vect(:, puti) += hij * coefs
+    end do
+  else if(h(0,sp) == 1) then
+    sfix = turn2(sp)
+    hfix = h(1,sfix)
+    pfix = p(1,sfix)
+    hmob = h(1,sp)
+    do j=1,2
+      puti = p(j, sp)
+      if(bannedOrb(puti)) cycle
+      pmob = p(turn2(j), sp)
+      hij = mo_two_e_integral(pfix, pmob, hfix, hmob)
+      hij *= get_phase_bi(phasemask, sp, sfix, hmob, pmob, hfix, pfix)
+      vect(:, puti) += hij * coefs
+    end do
+  else
+    puti = p(1,sp)
+    if(.not. bannedOrb(puti)) then
+      sfix = turn2(sp)
+      p1 = p(1,sfix)
+      p2 = p(2,sfix)
+      h1 = h(1,sfix)
+      h2 = h(2,sfix)
+      hij = (mo_two_e_integral(p1,p2,h1,h2) - mo_two_e_integral(p2,p1,h1,h2))
+      hij *= get_phase_bi(phasemask, sfix, sfix, h1, p1, h2, p2)
+      vect(:, puti) += hij * coefs
+    end if
+  end if
+end subroutine
+
+subroutine get_m1(gen, phasemask, bannedOrb, vect, mask, h, p, sp, coefs)
+  use bitmasks
+  implicit none
+
+  integer(bit_kind), intent(in) :: gen(N_int, 2), mask(N_int, 2)
+  integer(bit_kind), intent(in) :: phasemask(N_int, 2)
+  logical, intent(in) :: bannedOrb(mo_num)
+  double precision, intent(in) :: coefs(N_states)
+  double precision, intent(inout) :: vect(N_states, mo_num)
+  integer, intent(in) :: sp, h(0:2, 2), p(0:3, 2)
+  integer :: i, hole, p1, p2, sh
+  logical :: ok, lbanned(mo_num)
+  integer(bit_kind) :: det(N_int, 2)
+  double precision :: hij
+  double precision, external :: get_phase_bi,mo_two_e_integral
+
+  lbanned = bannedOrb
+  sh = 1
+  if(h(0,2) == 1) sh = 2
+  hole = h(1, sh)
+  lbanned(p(1,sp)) = .true.
+  if(p(0,sp) == 2) lbanned(p(2,sp)) = .true.
+  !print *, "SPm1", sp, sh
+
+  p1 = p(1, sp)
+
+  if(sp == sh) then
+    p2 = p(2, sp)
+    lbanned(p2) = .true.
+
+    do i=1,hole-1
+      if(lbanned(i)) cycle
+      hij = (mo_two_e_integral(p1, p2, i, hole) - mo_two_e_integral(p2, p1, i, hole))
+      hij *= get_phase_bi(phasemask, sp, sp, i, p1, hole, p2)
+      vect(:,i) += hij * coefs
+    end do
+    do i=hole+1,mo_num
+      if(lbanned(i)) cycle
+      hij = (mo_two_e_integral(p1, p2, hole, i) - mo_two_e_integral(p2, p1, hole, i))
+      hij *= get_phase_bi(phasemask, sp, sp, hole, p1, i, p2)
+      vect(:,i) += hij * coefs
+    end do
+
+    call apply_particle(mask, sp, p2, det, ok,  N_int)
+    call i_h_j(gen, det, N_int, hij)
+    vect(:, p2) += hij * coefs
+  else
+    p2 = p(1, sh)
+    do i=1,mo_num
+      if(lbanned(i)) cycle
+      hij = mo_two_e_integral(p1, p2, i, hole)
+      hij *= get_phase_bi(phasemask, sp, sh, i, p1, hole, p2)
+      vect(:,i) += hij * coefs
+    end do
+  end if
+
+  call apply_particle(mask, sp, p1, det, ok,  N_int)
+  call i_h_j(gen, det, N_int, hij)
+  vect(:, p1) += hij * coefs
+end subroutine
+
+subroutine get_m0(gen, phasemask, bannedOrb, vect, mask, h, p, sp, coefs)
+  use bitmasks
+  implicit none
+
+  integer(bit_kind), intent(in) :: gen(N_int, 2), mask(N_int, 2)
+  integer(bit_kind), intent(in) :: phasemask(N_int, 2)
+  logical, intent(in) :: bannedOrb(mo_num)
+  double precision, intent(in) :: coefs(N_states)
+  double precision, intent(inout) :: vect(N_states, mo_num)
+  integer, intent(in) :: sp, h(0:2, 2), p(0:3, 2)
+  integer :: i
+  logical :: ok, lbanned(mo_num)
+  integer(bit_kind) :: det(N_int, 2)
+  double precision :: hij
+
+  lbanned = bannedOrb
+  lbanned(p(1,sp)) = .true.
+  do i=1,mo_num
+    if(lbanned(i)) cycle
+    call apply_particle(mask, sp, i, det, ok, N_int)
+    call i_h_j(gen, det, N_int, hij)
+    vect(:, i) += hij * coefs
+  end do
+end subroutine
+
+
+
+!
+!subroutine fill_buffer_single(i_generator, sp, h1, bannedOrb, fock_diag_tmp, E0, pt2, vect, buf)
+!  use bitmasks
+!  use selection_types
+!  implicit none
+!
+!  integer, intent(in) :: i_generator, sp, h1
+!  double precision, intent(in) :: vect(N_states, mo_num)
+!  logical, intent(in) :: bannedOrb(mo_num)
+!  double precision, intent(in)           :: fock_diag_tmp(mo_num)
+!  double precision, intent(in)    :: E0(N_states)
+!  double precision, intent(inout) :: pt2(N_states)
+!  type(selection_buffer), intent(inout) :: buf
+!  logical :: ok
+!  integer :: s1, s2, p1, p2, ib, istate
+!  integer(bit_kind) :: mask(N_int, 2), det(N_int, 2)
+!  double precision :: e_pert, delta_E, val, Hii, max_e_pert, tmp
+!  double precision, external :: diag_H_mat_elem_fock
+!
+!
+!  call apply_hole(psi_det_generators(1,1,i_generator), sp, h1, mask, ok, N_int)
+!
+!  do p1=1,mo_num
+!    if(bannedOrb(p1)) cycle
+!    if(vect(1, p1) == 0d0) cycle
+!    call apply_particle(mask, sp, p1, det, ok, N_int)
+!
+!
+!    Hii = diag_H_mat_elem_fock(psi_det_generators(1,1,i_generator),det,fock_diag_tmp,N_int)
+!    max_e_pert = 0d0
+!
+!    do istate=1,N_states
+!      val = vect(istate, p1) + vect(istate, p1)
+!      delta_E = E0(istate) - Hii
+!      tmp = dsqrt(delta_E * delta_E + val * val)
+!      if (delta_E < 0.d0) then
+!        tmp = -tmp
+!      endif
+!      e_pert = 0.5d0 * ( tmp - delta_E)
+!      pt2(istate) += e_pert
+!      if(dabs(e_pert) > dabs(max_e_pert)) max_e_pert = e_pert
+!    end do
+!
+!    if(dabs(max_e_pert) > buf%mini) call add_to_selection_buffer(buf, det, max_e_pert)
+!  end do
+!end subroutine
+!

src/cipsi_tc_bi_ortho/selection.irp.f

@@ -868,7 +868,6 @@ subroutine fill_buffer_double(i_generator, sp, h1, h2, bannedOrb, banned, fock_d
         ! <det|H(j)|psi_0> and transpose
         ! -------------------------------------------
 
-!        call htilde_mu_mat_bi_ortho_tot(det, det, N_int, Hii)
         double precision               :: hmono, htwoe, hthree
         call diag_htilde_mu_mat_fock_bi_ortho(N_int, det, hmono, htwoe, hthree, hii)
         do istate = 1,N_states
@@ -878,8 +877,8 @@ subroutine fill_buffer_double(i_generator, sp, h1, h2, bannedOrb, banned, fock_d
             psi_h_alpha = 0.d0
             alpha_h_psi = 0.d0
             do iii = 1, N_det_selectors
-              call htilde_mu_mat_bi_ortho_tot(psi_selectors(1,1,iii), det, N_int, i_h_alpha)
+              call htilde_mu_mat_bi_ortho_tot_slow(psi_selectors(1,1,iii), det, N_int, i_h_alpha)
-              call htilde_mu_mat_bi_ortho_tot(det, psi_selectors(1,1,iii), N_int, alpha_h_i)
+              call htilde_mu_mat_bi_ortho_tot_slow(det, psi_selectors(1,1,iii), N_int, alpha_h_i)
               call get_excitation_degree(psi_selectors(1,1,iii), det,degree,N_int)
               if(degree == 0)then
                print*,'problem !!!'

src/cosgtos_ao_int/EZFIO.cfg

@@ -1,19 +0,0 @@
-[ao_expoim_cosgtos]
-type: double precision
-doc: imag part for Exponents for each primitive of each cosGTOs |AO|
-size: (ao_basis.ao_num,ao_basis.ao_prim_num_max)
-interface: ezfio, provider
-
-[use_cosgtos]
-type: logical
-doc: If true, use cosgtos for AO integrals
-interface: ezfio,provider,ocaml
-default: False
-
-[ao_integrals_threshold]
-type: Threshold
-doc: If | (pq|rs) | < `ao_integrals_threshold` then (pq|rs) is zero
-interface: ezfio,provider,ocaml
-default: 1.e-15
-ezfio_name: threshold_ao
-

src/cosgtos_ao_int/NEED

@@ -1,2 +0,0 @@
-ezfio_files
-ao_basis

src/cosgtos_ao_int/README.rst

@@ -1,4 +0,0 @@
-==============
-cosgtos_ao_int
-==============
-

src/cosgtos_ao_int/cosgtos_ao_int.irp.f

@@ -1,7 +0,0 @@
-program cosgtos_ao_int
-  implicit none
-  BEGIN_DOC
-! TODO : Put the documentation of the program here
-  END_DOC
-  print *, 'Hello world'
-end

src/davidson/davidson_parallel.irp.f

@@ -545,11 +545,6 @@ end
 
 
 
-
-
-
-
-
 integer function zmq_put_N_states_diag(zmq_to_qp_run_socket,worker_id)
   use f77_zmq
   implicit none

src/ezfio_files/00.create.bats

@@ -23,6 +23,34 @@ function run {
   qp set mo_two_e_ints io_mo_two_e_integrals "Write" 
 }
 
+@test "H2_1" {
+  run h2_1.xyz 1 0 cc-pvdz
+}
+
+@test "H2_3" {
+  run h2_3.xyz 3 0 cc-pvdz
+}
+
+@test "H3_2" {
+  run h3_2.xyz 2 0 cc-pvdz
+}
+
+@test "H3_4" {
+  run h3_4.xyz 4 0 cc-pvdz
+}
+
+@test "H4_1" {
+  run h4_1.xyz 1 0 cc-pvdz
+}
+
+@test "H4_3" {
+  run h4_3.xyz 3 0 cc-pvdz
+}
+
+@test "H4_5" {
+  run h4_5.xyz 5 0 cc-pvdz
+}
+
 
 @test "B-B" {
   qp set_file b2_stretched.ezfio

src/fci/40.fci.bats

@@ -10,8 +10,8 @@ function run() {
   qp set perturbation do_pt2 False
   qp set determinants n_det_max 8000
   qp set determinants n_states  1
-  qp set davidson threshold_davidson 1.e-10
+  qp set davidson_keywords threshold_davidson 1.e-10
-  qp set davidson n_states_diag 8
+  qp set davidson_keywords n_states_diag 8
   qp run fci
   energy1="$(ezfio get fci energy | tr '[]' ' ' | cut -d ',' -f 1)"
   eq $energy1 $1 $thresh
@@ -24,99 +24,134 @@ function run_stoch() {
   qp set perturbation do_pt2 True
   qp set determinants n_det_max $3
   qp set determinants n_states  1
-  qp set davidson threshold_davidson 1.e-10
+  qp set davidson_keywords threshold_davidson 1.e-10
-  qp set davidson n_states_diag 1
+  qp set davidson_keywords n_states_diag 1
   qp run fci
   energy1="$(ezfio get fci energy_pt2 | tr '[]' ' ' | cut -d ',' -f 1)"
   eq $energy1 $1 $thresh
 }
 
-@test "B-B" {
+@test "H2_1" { # 1s
+  qp set_file h2_1.ezfio
+  qp set perturbation pt2_max 0.
+  run_stoch -1.06415255 1.e-8 10000
+}
+
+@test "H2_3" { # 1s
+  qp set_file h2_3.ezfio
+  qp set perturbation pt2_max 0.
+  run_stoch -0.96029881 1.e-8 10000
+}
+
+@test "H3_2" { # 3s
+  qp set_file h3_2.ezfio
+  qp set perturbation pt2_max 0.
+  run_stoch -1.61003132 1.e-8 10000
+}
+
+@test "H3_4" { # 2s
+  qp set_file h3_4.ezfio
+  qp set perturbation pt2_max 0.
+  run_stoch -1.02434843 1.e-8 10000
+}
+
+@test "H4_1" { # 13s
+  qp set_file h4_1.ezfio
+  qp set perturbation pt2_max 0.
+  run_stoch -2.01675062 1.e-8 10000
+}
+
+@test "H4_3" { # 10s
+  qp set_file h4_3.ezfio
+  qp set perturbation pt2_max 0.
+  run_stoch -1.95927626 1.e-8 10000
+}
+
+@test "H4_5" { # 3s
+  qp set_file h4_5.ezfio
+  qp set perturbation pt2_max 0.
+  run_stoch -1.25852765 1.e-8 10000
+}
+
+@test "B-B" { # 10s
   qp set_file b2_stretched.ezfio
   qp set determinants n_det_max 10000
   qp set_frozen_core
   run_stoch -49.14103054419 3.e-4 10000
 }
 
-@test "F2" { # 4.07m
+@test "NH3" { # 8s
-  [[ -n $TRAVIS ]] && skip
-  qp set_file f2.ezfio
-  qp set_frozen_core
-  run_stoch -199.304922384814 3.e-3  100000
-}
-
-@test "NH3" { # 10.6657s
   qp set_file nh3.ezfio
   qp set_mo_class --core="[1-4]" --act="[5-72]"
   run -56.244753429144986  3.e-4  100000
 }
 
-@test "DHNO" { # 11.4721s
+@test "DHNO" { # 8s
   qp set_file dhno.ezfio
   qp set_mo_class --core="[1-7]" --act="[8-64]"
-  run -130.459020029816 3.e-4  100000
+  run -130.466208113547 3.e-4  100000
 }
 
-@test "HCO" { # 12.2868s
+@test "HCO" { # 32s
   qp set_file hco.ezfio
-  run -113.393356604085 1.e-3  100000
+  run -113.395751656985 1.e-3  100000
 }
 
-@test "H2O2" { # 12.9214s
+@test "H2O2" { # 21s
   qp set_file h2o2.ezfio
   qp set_mo_class --core="[1-2]" --act="[3-24]" --del="[25-38]"
   run -151.005848404095 1.e-3  100000
 }
 
-@test "HBO" { # 13.3144s
+@test "HBO" { # 18s
   [[ -n $TRAVIS ]] && skip
   qp set_file hbo.ezfio
-  run -100.213 1.5e-3  100000
+  run -100.214 1.5e-3  100000
 }
 
-@test "H2O" { # 11.3727s
+@test "H2O" { # 16s
   [[ -n $TRAVIS ]] && skip
   qp set_file h2o.ezfio
-  run -76.2361605151999 5.e-4  100000
+  run -76.238051555276  5.e-4  100000
 }
 
-@test "ClO" { # 13.3755s
+@test "ClO" { # 47s
   [[ -n $TRAVIS ]] && skip
   qp set_file clo.ezfio
-  run -534.546453546852 1.e-3  100000
+  run -534.548529710256 1.e-3  100000
 }
 
-@test "SO" { # 13.4952s
+@test "SO" { # 23s
   [[ -n $TRAVIS ]] && skip
   qp set_file so.ezfio
   run -26.015 3.e-3  100000
 }
 
-@test "H2S" { # 13.6745s
+@test "H2S" { # 37s
   [[ -n $TRAVIS ]] && skip
   qp set_file h2s.ezfio
-  run -398.859577605891 5.e-4  100000
+  run -398.864853669111 5.e-4  100000
 }
 
-@test "OH" { # 13.865s
+@test "OH" { # 12s
   [[ -n $TRAVIS ]] && skip
   qp set_file oh.ezfio
-  run -75.6121856748294 3.e-4   100000
+  run -75.615 1.5e-3   100000
 }
 
-@test "SiH2_3B1" { # 13.938ss
+@test "SiH2_3B1" { # 10s
   [[ -n $TRAVIS ]] && skip
   qp set_file sih2_3b1.ezfio
-  run -290.0175411299477 3.e-4  100000
+  run -290.0206626734517 3.e-4  100000
 }
 
-@test "H3COH" { # 14.7299s
+@test "H3COH" { # 33s
   [[ -n $TRAVIS ]] && skip
   qp set_file h3coh.ezfio
-  run -115.205632960026 1.e-3  100000
+  run -115.206784386204 1.e-3  100000
 }
 
-@test "SiH3" { # 15.99s
+@test "SiH3" { # 15s
   [[ -n $TRAVIS ]] && skip
   qp set_file sih3.ezfio
   run -5.572 1.e-3   100000
@@ -132,7 +167,7 @@ function run_stoch() {
 @test "ClF" { # 16.8864s
   [[ -n $TRAVIS ]] && skip
   qp set_file clf.ezfio
-  run -559.169748890031 1.5e-3  100000
+  run -559.174371468224 1.5e-3  100000
 }
 
 @test "SO2" { # 17.5645s
@@ -170,12 +205,11 @@ function run_stoch() {
   run -187.970184372047 1.6e-3  100000
 }
 
-
 @test "[Cu(NH3)4]2+" { # 25.0417s
   [[ -n $TRAVIS ]] && skip
   qp set_file cu_nh3_4_2plus.ezfio
   qp set_mo_class --core="[1-24]" --act="[25-45]" --del="[46-87]"
-  run -1862.9869374387192  3.e-04  100000
+  run -1862.98320066637   3.e-04  100000
 }
 
 @test "HCN" { # 20.3273s
@@ -185,3 +219,10 @@ function run_stoch() {
   run -93.078 2.e-3  100000
 }
 
+@test "F2" { # 4.07m
+  [[ -n $TRAVIS ]] && skip
+  qp set_file f2.ezfio
+  qp set_frozen_core
+  run_stoch -199.304922384814 3.e-3  100000
+}
+

src/fci_tc_bi/diagonalize_ci.irp.f

@@ -39,7 +39,7 @@ subroutine diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
    write(*,'(A28,X,I10,X,100(F16.8,X))')'Ndet,E,E+PT2,E+RPT2,|PT2|=',ndet,E_tc ,E_tc  + pt2_tmp/norm,E_tc  + rpt2_tmp/norm,abs_pt2
    print*,'*****'
   endif
-  psi_energy(1:N_states) = eigval_right_tc_bi_orth(1:N_states)
+  psi_energy(1:N_states) = eigval_right_tc_bi_orth(1:N_states) - nuclear_repulsion
   psi_s2(1:N_states) = s2_eigvec_tc_bi_orth(1:N_states)
 
   E_tc  = eigval_right_tc_bi_orth(1)

src/fci_tc_bi/fci_tc_bi_ortho.irp.f

@@ -54,14 +54,18 @@ subroutine run_cipsi_tc
 
   implicit none
 
-  if (.not.is_zmq_slave) then
+  if (.not. is_zmq_slave) then
+
     PROVIDE psi_det psi_coef mo_bi_ortho_tc_two_e mo_bi_ortho_tc_one_e
-    if(elec_alpha_num+elec_beta_num.ge.3)then
+
+    if(elec_alpha_num+elec_beta_num .ge. 3) then
       if(three_body_h_tc)then
-        call provide_all_three_ints_bi_ortho
+        call provide_all_three_ints_bi_ortho()
       endif
     endif
-    ! ---
+
+    FREE int2_grad1_u12_bimo_transp int2_grad1_u12_ao_transp
+
     write(json_unit,json_array_open_fmt) 'fci_tc'
 
     if (do_pt2) then
@@ -76,13 +80,16 @@ subroutine run_cipsi_tc
     call json_close
 
   else
+
     PROVIDE mo_bi_ortho_tc_one_e mo_bi_ortho_tc_two_e pt2_min_parallel_tasks
+
     if(elec_alpha_num+elec_beta_num.ge.3)then
       if(three_body_h_tc)then
         call provide_all_three_ints_bi_ortho
       endif
     endif
-    ! ---
+
+    FREE int2_grad1_u12_bimo_transp int2_grad1_u12_ao_transp
 
     call run_slave_cipsi
 

src/hartree_fock/10.hf.bats

@@ -43,11 +43,39 @@ python write_pt_charges.py ${EZFIO}
 qp set nuclei point_charges True
 qp run scf | tee ${EZFIO}.pt_charges.out
   energy="$(ezfio get hartree_fock energy)"
-good=-92.76613324421798
+good=-92.79920682236470
   eq $energy $good $thresh
 rm -rf $EZFIO
 }
 
+@test "H2_1" { # 1s
+  run h2_1.ezfio -1.005924963288527
+}
+
+@test "H2_3" { # 1s
+  run h2_3.ezfio -0.9591011604845440
+}
+
+@test "H3_2" { # 1s
+  run h3_2.ezfio -1.558273529860488
+}
+
+@test "H3_4" { # 1s
+  run h3_4.ezfio -1.0158684760025190
+}
+
+@test "H4_1" { # 1s
+  run h4_1.ezfio -1.932022805374405
+}
+
+@test "H4_3" { # 1s
+  run h4_3.ezfio -1.8948449927787350
+}
+
+@test "H4_5" { # 1s
+  run h4_5.ezfio -1.2408338805496990
+}
+
 @test "point charges" { 
  run_pt_charges
 }
@@ -56,6 +84,8 @@ rm -rf $EZFIO
   run hcn.ezfio -92.88717500035233
 }
 
+
+
 @test "B-B" { # 3s
   run b2_stretched.ezfio -48.9950585434279
 }

src/mo_optimization/my_providers.irp.f

@@ -1,141 +0,0 @@
-! Dimensions of MOs
-
-
-BEGIN_PROVIDER [ integer, n_mo_dim ]
-  implicit none
-  BEGIN_DOC
-  ! Number of different pairs (i,j) of MOs we can build,
-  ! with i>j
-  END_DOC
-
-  n_mo_dim = mo_num*(mo_num-1)/2
-
-END_PROVIDER
-
-BEGIN_PROVIDER [ integer, n_mo_dim_core ]
-  implicit none 
-  BEGIN_DOC
-  ! Number of different pairs (i,j) of core MOs we can build,
-  ! with i>j
-  END_DOC
-
-  n_mo_dim_core = dim_list_core_orb*(dim_list_core_orb-1)/2
-
-END_PROVIDER
-
-BEGIN_PROVIDER [ integer, n_mo_dim_act ]
-  implicit none
-  BEGIN_DOC
-  ! Number of different pairs (i,j) of active MOs we can build,
-  ! with i>j
-  END_DOC
-
-  n_mo_dim_act = dim_list_act_orb*(dim_list_act_orb-1)/2
-
-END_PROVIDER
-
-BEGIN_PROVIDER [ integer, n_mo_dim_inact ]
-  implicit none 
-  BEGIN_DOC
-  ! Number of different pairs (i,j) of inactive MOs we can build,
-  ! with i>j
-  END_DOC
-
-  n_mo_dim_inact = dim_list_inact_orb*(dim_list_inact_orb-1)/2
-
-END_PROVIDER
-
-BEGIN_PROVIDER [ integer, n_mo_dim_virt ]
-  implicit none 
-  BEGIN_DOC
-  ! Number of different pairs (i,j) of virtual MOs we can build,
-  ! with i>j
-  END_DOC
-
-  n_mo_dim_virt = dim_list_virt_orb*(dim_list_virt_orb-1)/2
-
-END_PROVIDER
-
-! Energies/criterions
-
-BEGIN_PROVIDER [ double precision, my_st_av_energy ]
-  implicit none
-  BEGIN_DOC
-  ! State average CI energy
-  END_DOC
-
-  !call update_st_av_ci_energy(my_st_av_energy)
-  call state_average_energy(my_st_av_energy)
-
-END_PROVIDER
-
-! With all the MOs
-
-BEGIN_PROVIDER [ double precision, my_gradient_opt, (n_mo_dim) ]
-&BEGIN_PROVIDER [ double precision, my_CC1_opt ]
-  implicit none
-  BEGIN_DOC
-  ! - Gradient of the energy with respect to the MO rotations, for all the MOs.
-  ! - Maximal element of the gradient in absolute value 
-  END_DOC
-
-  double precision :: norm_grad
-
-  PROVIDE mo_two_e_integrals_in_map
-
-  call gradient_opt(n_mo_dim, my_gradient_opt, my_CC1_opt, norm_grad)
-
-END_PROVIDER
-
-BEGIN_PROVIDER [ double precision, my_hessian_opt, (n_mo_dim, n_mo_dim) ]
-  implicit none
-  BEGIN_DOC
-  ! - Gradient of the energy with respect to the MO rotations, for all the MOs.
-  ! - Maximal element of the gradient in absolute value 
-  END_DOC
-
-  double precision, allocatable :: h_f(:,:,:,:)
-
-  PROVIDE mo_two_e_integrals_in_map
-
-  allocate(h_f(mo_num, mo_num, mo_num, mo_num))
-
-  call hessian_list_opt(n_mo_dim, my_hessian_opt, h_f)
-
-END_PROVIDER
-
-! With the list of active MOs
-! Can be generalized to any mo_class by changing the list/dimension
-
-BEGIN_PROVIDER [ double precision, my_gradient_list_opt, (n_mo_dim_act) ]
-&BEGIN_PROVIDER [ double precision, my_CC2_opt ]
-  implicit none
-  BEGIN_DOC
-  ! - Gradient of the energy with respect to the MO rotations, only for the active MOs !
-  ! - Maximal element of the gradient in absolute value 
-  END_DOC
-
-  double precision :: norm_grad
-
-  PROVIDE mo_two_e_integrals_in_map !one_e_dm_mo two_e_dm_mo mo_one_e_integrals 
-
-  call gradient_list_opt(n_mo_dim_act, dim_list_act_orb, list_act, my_gradient_list_opt, my_CC2_opt, norm_grad)
-
-END_PROVIDER
-
-BEGIN_PROVIDER [ double precision, my_hessian_list_opt, (n_mo_dim_act, n_mo_dim_act) ]
-  implicit none
-  BEGIN_DOC
-  ! - Gradient of the energy with respect to the MO rotations, only for the active MOs !
-  ! - Maximal element of the gradient in absolute value 
-  END_DOC
-
-  double precision, allocatable :: h_f(:,:,:,:)
-
-  PROVIDE mo_two_e_integrals_in_map
-
-  allocate(h_f(dim_list_act_orb, dim_list_act_orb, dim_list_act_orb, dim_list_act_orb))
-
-  call hessian_list_opt(n_mo_dim_act, dim_list_act_orb, list_act, my_hessian_list_opt, h_f)
-
-END_PROVIDER

src/mo_two_e_ints/cholesky.irp.f

@@ -6,11 +6,44 @@ BEGIN_PROVIDER [ double precision, cholesky_mo, (mo_num, mo_num, cholesky_ao_num
 
  integer :: k
 
+ call set_multiple_levels_omp(.False.)
+ print *, 'AO->MO Transformation of Cholesky vectors'
  !$OMP PARALLEL DO PRIVATE(k)
  do k=1,cholesky_ao_num
   call ao_to_mo(cholesky_ao(1,1,k),ao_num,cholesky_mo(1,1,k),mo_num)
  enddo
  !$OMP END PARALLEL DO
+ print *, ''
+
+END_PROVIDER
+
+BEGIN_PROVIDER [ double precision, cholesky_mo_transp, (cholesky_ao_num, mo_num, mo_num) ]
+ implicit none
+ BEGIN_DOC
+ ! Cholesky vectors in MO basis
+ END_DOC
+
+ integer :: i,j,k
+ double precision, allocatable :: buffer(:,:)
+
+ print *, 'AO->MO Transformation of Cholesky vectors  .'
+
+ call set_multiple_levels_omp(.False.)
+ !$OMP PARALLEL PRIVATE(i,j,k,buffer)
+ allocate(buffer(mo_num,mo_num))
+ !$OMP DO SCHEDULE(static)
+ do k=1,cholesky_ao_num
+  call ao_to_mo(cholesky_ao(1,1,k),ao_num,buffer,mo_num)
+  do j=1,mo_num
+    do i=1,mo_num
+      cholesky_mo_transp(k,i,j) = buffer(i,j)
+    enddo
+  enddo
+ enddo
+ !$OMP END DO
+ deallocate(buffer)
+ !$OMP END PARALLEL
+ print *, ''
 
 END_PROVIDER
 

src/mo_two_e_ints/integrals_3_index.irp.f

@@ -4,24 +4,68 @@
  BEGIN_DOC
  ! big_array_coulomb_integrals(j,i,k)  = <ij|kj> = (ik|jj)
  !
- ! big_array_exchange_integrals(i,j,k) = <ij|jk> = (ij|kj)
+ ! big_array_exchange_integrals(j,i,k) = <ij|jk> = (ij|kj)
  END_DOC
- integer :: i,j,k,l
+ integer :: i,j,k,l,a
  double precision :: get_two_e_integral
  double precision :: integral
 
- do k = 1, mo_num
+ if (do_ao_cholesky) then
-  do i = 1, mo_num
+
-   do j = 1, mo_num
+    double precision, allocatable :: buffer_jj(:,:), buffer(:,:,:)
-     l = j
+    allocate(buffer_jj(cholesky_ao_num,mo_num), buffer(mo_num,mo_num,mo_num))
-     integral = get_two_e_integral(i,j,k,l,mo_integrals_map)
+    do j=1,mo_num
-     big_array_coulomb_integrals(j,i,k) = integral
+      buffer_jj(:,j) = cholesky_mo_transp(:,j,j)
-     l = j
+    enddo
-     integral = get_two_e_integral(i,j,l,k,mo_integrals_map)
+
-     big_array_exchange_integrals(j,i,k) = integral
+    call dgemm('T','N', mo_num*mo_num,mo_num,cholesky_ao_num, 1.d0, &
+        cholesky_mo_transp, cholesky_ao_num, &
+        buffer_jj, cholesky_ao_num, 0.d0, &
+        buffer, mo_num*mo_num)
+
+    do k = 1, mo_num
+      do i = 1, mo_num
+        do j = 1, mo_num
+          big_array_coulomb_integrals(j,i,k) = buffer(i,k,j)
+        enddo
+      enddo
+    enddo
+    deallocate(buffer_jj)
+
+    allocate(buffer_jj(mo_num,mo_num))
+
+    do j = 1, mo_num
+
+      call dgemm('T','N',mo_num,mo_num,cholesky_ao_num, 1.d0, &
+        cholesky_mo_transp(1,1,j), cholesky_ao_num, &
+        cholesky_mo_transp(1,1,j), cholesky_ao_num, 0.d0, &
+        buffer_jj, mo_num)
+
+      do k=1,mo_num
+        do i=1,mo_num
+          big_array_exchange_integrals(j,i,k) = buffer_jj(i,k)
+       enddo
+     enddo
+    enddo
+
+    deallocate(buffer_jj)
+
+ else
+
+   do k = 1, mo_num
+     do i = 1, mo_num
+       do j = 1, mo_num
+         l = j
+         integral = get_two_e_integral(i,j,k,l,mo_integrals_map)
+         big_array_coulomb_integrals(j,i,k) = integral
+         l = j
+         integral = get_two_e_integral(i,j,l,k,mo_integrals_map)
+         big_array_exchange_integrals(j,i,k) = integral
+       enddo
+     enddo
    enddo
-  enddo
+
- enddo
+ endif
 
 END_PROVIDER
 

src/mo_two_e_ints/mo_bi_integrals.irp.f

@@ -1353,15 +1353,30 @@ END_PROVIDER
   integer                        :: i,j
   double precision               :: get_two_e_integral
 
-  PROVIDE mo_two_e_integrals_in_map
+
-  mo_two_e_integrals_jj = 0.d0
+  if (do_ao_cholesky) then
-  mo_two_e_integrals_jj_exchange = 0.d0
+    do j=1,mo_num
+      do i=1,mo_num
+        !TODO: use dgemm
+        mo_two_e_integrals_jj(i,j) = sum(cholesky_mo_transp(:,i,i)*cholesky_mo_transp(:,j,j))
+        mo_two_e_integrals_jj_exchange(i,j) = sum(cholesky_mo_transp(:,i,j)*cholesky_mo_transp(:,j,i))
+      enddo
+    enddo
+
+  else
+
+    do j=1,mo_num
+      do i=1,mo_num
+        mo_two_e_integrals_jj(i,j) = get_two_e_integral(i,j,i,j,mo_integrals_map)
+        mo_two_e_integrals_jj_exchange(i,j) = get_two_e_integral(i,j,j,i,mo_integrals_map)
+      enddo
+    enddo
+
+  endif
 
   do j=1,mo_num
     do i=1,mo_num
-      mo_two_e_integrals_jj(i,j) = get_two_e_integral(i,j,i,j,mo_integrals_map)
+        mo_two_e_integrals_jj_anti(i,j) = mo_two_e_integrals_jj(i,j) - mo_two_e_integrals_jj_exchange(i,j)
-      mo_two_e_integrals_jj_exchange(i,j) = get_two_e_integral(i,j,j,i,mo_integrals_map)
-      mo_two_e_integrals_jj_anti(i,j) = mo_two_e_integrals_jj(i,j) - mo_two_e_integrals_jj_exchange(i,j)
     enddo
   enddo
 

src/non_h_ints_mu/grad_squared.irp.f

@@ -231,6 +231,7 @@ BEGIN_PROVIDER [ double precision, grad12_j12, (ao_num, ao_num, n_points_final_g
   call wall_time(time0)
 
   PROVIDE j1b_type
+  PROVIDE int2_grad1u2_grad2u2_j1b2
 
   do ipoint = 1, n_points_final_grid
     tmp1 = v_1b(ipoint)
@@ -242,6 +243,8 @@ BEGIN_PROVIDER [ double precision, grad12_j12, (ao_num, ao_num, n_points_final_g
     enddo
   enddo
 
+  FREE int2_grad1u2_grad2u2_j1b2
+
   !if(j1b_type .eq. 0) then 
   !  grad12_j12 = 0.d0
   !  do ipoint = 1, n_points_final_grid
@@ -262,6 +265,7 @@ BEGIN_PROVIDER [ double precision, grad12_j12, (ao_num, ao_num, n_points_final_g
 
   call wall_time(time1)
   print*, ' Wall time for grad12_j12 = ', time1 - time0
+  call print_memory_usage()
 
 END_PROVIDER
 
@@ -278,6 +282,9 @@ BEGIN_PROVIDER [double precision, u12sq_j1bsq, (ao_num, ao_num, n_points_final_g
   print*, ' providing u12sq_j1bsq ...'
   call wall_time(time0)
 
+  ! do not free here
+  PROVIDE int2_u2_j1b2
+
   do ipoint = 1, n_points_final_grid
     tmp_x = v_1b_grad(1,ipoint)
     tmp_y = v_1b_grad(2,ipoint)
@@ -292,6 +299,7 @@ BEGIN_PROVIDER [double precision, u12sq_j1bsq, (ao_num, ao_num, n_points_final_g
 
   call wall_time(time1)
   print*, ' Wall time for u12sq_j1bsq = ', time1 - time0
+  call print_memory_usage()
 
 END_PROVIDER
 
@@ -310,6 +318,9 @@ BEGIN_PROVIDER [ double precision, u12_grad1_u12_j1b_grad1_j1b, (ao_num, ao_num,
   print*, ' providing u12_grad1_u12_j1b_grad1_j1b ...'
   call wall_time(time0)
 
+  PROVIDE int2_u_grad1u_j1b2
+  PROVIDE int2_u_grad1u_x_j1b2
+
   do ipoint = 1, n_points_final_grid
 
     x     = final_grid_points(1,ipoint)
@@ -340,14 +351,17 @@ BEGIN_PROVIDER [ double precision, u12_grad1_u12_j1b_grad1_j1b, (ao_num, ao_num,
     enddo
   enddo
 
+  FREE int2_u_grad1u_j1b2
+  FREE int2_u_grad1u_x_j1b2
+
   call wall_time(time1)
   print*, ' Wall time for u12_grad1_u12_j1b_grad1_j1b = ', time1 - time0
+  call print_memory_usage()
 
 END_PROVIDER
 
 ! ---
 
-
 BEGIN_PROVIDER [double precision, tc_grad_square_ao, (ao_num, ao_num, ao_num, ao_num)]
 
   BEGIN_DOC
@@ -401,6 +415,8 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao, (ao_num, ao_num, ao_num, ao
               , int2_grad1_u12_square_ao(1,1,1), ao_num*ao_num, b_mat(1,1,1), n_points_final_grid &
               , 0.d0, tc_grad_square_ao, ao_num*ao_num)
 
+    FREE int2_grad1_u12_square_ao
+
     ! ---
 
     if(((j1b_type .eq. 3) .or. (j1b_type .eq. 4)) .and. use_ipp) then
@@ -442,6 +458,8 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao, (ao_num, ao_num, ao_num, ao
       call dgemm( "N", "N", ao_num*ao_num, ao_num*ao_num, n_points_final_grid, 1.d0     &
                 , int2_u2_j1b2(1,1,1), ao_num*ao_num, b_mat(1,1,1), n_points_final_grid &
                 , 1.d0, tc_grad_square_ao, ao_num*ao_num)
+
+      FREE int2_u2_j1b2
     endif
 
     ! ---
@@ -478,6 +496,7 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao, (ao_num, ao_num, ao_num, ao
 
   call wall_time(time1)
   print*, ' Wall time for tc_grad_square_ao = ', time1 - time0
+  call print_memory_usage()
 
 END_PROVIDER
 

src/non_h_ints_mu/j12_nucl_utils.irp.f

@@ -35,7 +35,7 @@ BEGIN_PROVIDER [ double precision, v_1b, (n_points_final_grid)]
 
   elseif(j1b_type .eq. 4) then
 
-    ! v(r) = 1 - \sum_{a} \exp(-\alpha_a (r - r_a)^2)
+    ! v(r) = 1 - \sum_{a} \beta_a \exp(-\alpha_a (r - r_a)^2)
 
     do ipoint = 1, n_points_final_grid
 
@@ -51,7 +51,7 @@ BEGIN_PROVIDER [ double precision, v_1b, (n_points_final_grid)]
         dz = z - nucl_coord(j,3)
         d  = dx*dx + dy*dy + dz*dz
 
-        fact_r = fact_r - dexp(-a*d)
+        fact_r = fact_r - j1b_pen_coef(j) * dexp(-a*d)
       enddo
 
       v_1b(ipoint) = fact_r
@@ -125,7 +125,7 @@ BEGIN_PROVIDER [double precision, v_1b_grad, (3, n_points_final_grid)]
 
   elseif(j1b_type .eq. 4) then
 
-    ! v(r) = 1 - \sum_{a} \exp(-\alpha_a (r - r_a)^2)
+    ! v(r) = 1 - \sum_{a} \beta_a \exp(-\alpha_a (r - r_a)^2)
 
     do ipoint = 1, n_points_final_grid
 
@@ -144,7 +144,7 @@ BEGIN_PROVIDER [double precision, v_1b_grad, (3, n_points_final_grid)]
         r2 = dx*dx + dy*dy + dz*dz
 
         a = j1b_pen(j)
-        e = a * dexp(-a * r2)
+        e = a * j1b_pen_coef(j) * dexp(-a * r2)
 
         ax_der += e * dx
         ay_der += e * dy

src/non_h_ints_mu/jast_deriv.irp.f

@@ -187,6 +187,19 @@ end function j12_mu
 
 subroutine grad1_j12_mu(r1, r2, grad)
 
+  BEGIN_DOC
+!  gradient of j(mu(r1,r2),r12) form of jastrow. 
+!
+! if mu(r1,r2) = cst ---> j1b_type < 200 and 
+!
+!  d/dx1 j(mu,r12) = 0.5 * (1 - erf(mu *r12))/r12 * (x1 - x2)
+!
+! if mu(r1,r2) /= cst ---> 200 < j1b_type < 300 and 
+!
+! d/dx1 j(mu(r1,r2),r12) = exp(-(mu(r1,r2)*r12)**2) /(2 *sqrt(pi) * mu(r1,r2)**2 )   d/dx1 mu(r1,r2) 
+!
+!                          + 0.5 * (1 - erf(mu(r1,r2) *r12))/r12 * (x1 - x2)
+  END_DOC
   include 'constants.include.F'
 
   implicit none
@@ -283,7 +296,7 @@ double precision function j1b_nucl(r)
       d = ( (r(1) - nucl_coord(i,1)) * (r(1) - nucl_coord(i,1)) &
           + (r(2) - nucl_coord(i,2)) * (r(2) - nucl_coord(i,2)) &
           + (r(3) - nucl_coord(i,3)) * (r(3) - nucl_coord(i,3)) )
-      j1b_nucl = j1b_nucl - dexp(-a*d)
+      j1b_nucl = j1b_nucl - j1b_pen_coef(i) * dexp(-a*d)
     enddo
 
   elseif((j1b_type .eq. 5) .or. (j1b_type .eq. 105)) then
@@ -350,7 +363,7 @@ double precision function j1b_nucl_square(r)
       d = ( (r(1) - nucl_coord(i,1)) * (r(1) - nucl_coord(i,1)) &
           + (r(2) - nucl_coord(i,2)) * (r(2) - nucl_coord(i,2)) &
           + (r(3) - nucl_coord(i,3)) * (r(3) - nucl_coord(i,3)) )
-      j1b_nucl_square = j1b_nucl_square - dexp(-a*d)
+      j1b_nucl_square = j1b_nucl_square - j1b_pen_coef(i) * dexp(-a*d)
     enddo
     j1b_nucl_square = j1b_nucl_square * j1b_nucl_square
 
@@ -462,7 +475,7 @@ subroutine grad1_j1b_nucl(r, grad)
       y = r(2) - nucl_coord(i,2)
       z = r(3) - nucl_coord(i,3)
       d = x*x + y*y + z*z
-      e = a * dexp(-a*d)
+      e = a * j1b_pen_coef(i) * dexp(-a*d)
 
       fact_x += e * x
       fact_y += e * y
@@ -515,6 +528,9 @@ subroutine mu_r_val_and_grad(r1, r2, mu_val, mu_der)
   double precision              :: r(3)
   double precision              :: dm_a(1), dm_b(1), grad_dm_a(3,1), grad_dm_b(3,1)
   double precision              :: dm_tot, tmp1, tmp2, tmp3
+  double precision :: rho1, grad_rho1(3),rho2,rho_tot,inv_rho_tot
+  double precision :: f_rho1, f_rho2, d_drho_f_rho1
+  double precision :: d_dx1_f_rho1(3),d_dx_rho_f_rho(3),nume
 
   if(j1b_type .eq. 200) then
 
@@ -578,8 +594,84 @@ subroutine mu_r_val_and_grad(r1, r2, mu_val, mu_der)
     mu_der(2) = tmp3 * (grad_dm_a(2,1) + grad_dm_b(2,1))
     mu_der(3) = tmp3 * (grad_dm_a(3,1) + grad_dm_b(3,1))
 
-  else
+  elseif(j1b_type .eq. 202) then
 
+    ! mu(r1,r2) = {rho(r1) f[rho(r1)] + rho(r2) f[rho(r2)]} / RHO
+    !
+    ! RHO = rho(r1) + rho(r2)
+    !
+    ! f[rho] = alpha rho^beta + mu0 exp(-rho)
+    !
+    ! d/dx1 mu(r1,r2) = 1/RHO^2 * {RHO * d/dx1 (rho(r1) f[rho(r1)]) 
+    !                              - d/dx1 rho(r1) * [rho(r1) f[rho(r1)] + rho(r2) f[rho(r2)]] }
+    !
+    ! d/dx1 f[rho(r1)] = [0.5 alpha / sqrt(rho(r1)) - mu0 exp(-rho(r1))] (d rho(r1) / dx1)
+    !
+    ! d/dx1 (rho(r1) f[rho(r1)] = rho(r1) * d/dx1 f[rho(r1)] + f[rho(r1)] * d/dx1 rho(r1)
+     
+    !!!!!!!!! rho1,rho2,rho1+rho2
+    call get_all_rho_grad_rho(r1,r2,rho1,rho2,grad_rho1)
+    rho_tot = rho1 + rho2
+    if(rho_tot.lt.1.d-10)rho_tot = 1.d-10
+    inv_rho_tot = 1.d0/rho_tot
+    ! f(rho) = mu_r_ct * rho**beta_rho_power + mu_erf * exp(-rho)
+    call get_all_f_rho(rho1,rho2,mu_r_ct,mu_erf,beta_rho_power,f_rho1,d_drho_f_rho1,f_rho2)
+    d_dx1_f_rho1(1:3)   = d_drho_f_rho1 * grad_rho1(1:3)
+    d_dx_rho_f_rho(1:3) = rho1 * d_dx1_f_rho1(1:3) + f_rho1 * grad_rho1(1:3)
+    nume   = rho1 * f_rho1 + rho2 * f_rho2
+    mu_val = nume * inv_rho_tot
+    mu_der(1:3) = inv_rho_tot*inv_rho_tot * (rho_tot * d_dx_rho_f_rho(1:3) - grad_rho1(1:3) * nume)
+  elseif(j1b_type .eq. 203) then
+
+    ! mu(r1,r2) = {rho(r1) f[rho(r1)] + rho(r2) f[rho(r2)]} / RHO
+    !
+    ! RHO = rho(r1) + rho(r2)
+    !
+    ! f[rho] = alpha rho^beta + mu0 
+    !
+    ! d/dx1 mu(r1,r2) = 1/RHO^2 * {RHO * d/dx1 (rho(r1) f[rho(r1)]) 
+    !                              - d/dx1 rho(r1) * [rho(r1) f[rho(r1)] + rho(r2) f[rho(r2)]] }
+    !
+    ! d/dx1 f[rho(r1)] = [0.5 alpha / sqrt(rho(r1)) ] (d rho(r1) / dx1)
+    !
+    ! d/dx1 (rho(r1) f[rho(r1)] = rho(r1) * d/dx1 f[rho(r1)] + f[rho(r1)] * d/dx1 rho(r1)
+     
+    !!!!!!!!! rho1,rho2,rho1+rho2
+    call get_all_rho_grad_rho(r1,r2,rho1,rho2,grad_rho1)
+    rho_tot = rho1 + rho2
+    if(rho_tot.lt.1.d-10)rho_tot = 1.d-10
+    inv_rho_tot = 1.d0/rho_tot
+    ! f(rho) = mu_r_ct * rho**beta_rho_power + mu_erf 
+    call get_all_f_rho_simple(rho1,rho2,mu_r_ct,mu_erf,beta_rho_power,f_rho1,d_drho_f_rho1,f_rho2)
+    d_dx1_f_rho1(1:3)   = d_drho_f_rho1 * grad_rho1(1:3)
+    d_dx_rho_f_rho(1:3) = rho1 * d_dx1_f_rho1(1:3) + f_rho1 * grad_rho1(1:3)
+    nume   = rho1 * f_rho1 + rho2 * f_rho2
+    mu_val = nume * inv_rho_tot
+    mu_der(1:3) = inv_rho_tot*inv_rho_tot * (rho_tot * d_dx_rho_f_rho(1:3) - grad_rho1(1:3) * nume)
+  elseif(j1b_type .eq. 204) then
+
+    ! mu(r1,r2) = 1/2 * (f[rho(r1)] + f[rho(r2)]} 
+    !
+    ! f[rho] = alpha rho^beta + mu0 
+    !
+    ! d/dx1 mu(r1,r2) = 1/2 * d/dx1 (rho(r1) f[rho(r1)])
+    !                   
+    ! d/dx1 f[rho(r1)] = [0.5 alpha / sqrt(rho(r1)) ] (d rho(r1) / dx1)
+    !
+    ! d/dx1 (rho(r1) f[rho(r1)] = rho(r1) * d/dx1 f[rho(r1)] + f[rho(r1)] * d/dx1 rho(r1)
+     
+    !!!!!!!!! rho1,rho2,rho1+rho2
+    call get_all_rho_grad_rho(r1,r2,rho1,rho2,grad_rho1)
+    rho_tot = rho1 + rho2
+    if(rho_tot.lt.1.d-10)rho_tot = 1.d-10
+    inv_rho_tot = 1.d0/rho_tot
+    ! f(rho) = mu_r_ct * rho**beta_rho_power + mu_erf 
+    call get_all_f_rho_simple(rho1,rho2,mu_r_ct,mu_erf,beta_rho_power,f_rho1,d_drho_f_rho1,f_rho2)
+    d_dx1_f_rho1(1:3)   = d_drho_f_rho1 * grad_rho1(1:3)
+    d_dx_rho_f_rho(1:3) = rho1 * d_dx1_f_rho1(1:3) + f_rho1 * grad_rho1(1:3)
+    mu_val = 0.5d0 * ( f_rho1 + f_rho2)
+    mu_der(1:3) = d_dx_rho_f_rho(1:3) 
+  else
     print *, ' j1b_type = ', j1b_type, 'not implemented yet'
     stop
 
@@ -684,3 +776,76 @@ end function j12_mu_square
 
 ! ---
 
+subroutine f_mu_and_deriv_mu(rho,alpha,mu0,beta,f_mu,d_drho_f_mu)
+ implicit none
+ BEGIN_DOC
+! function giving mu as a function of rho
+!
+! f_mu = alpha * rho**beta + mu0 * exp(-rho)
+!
+! and its derivative with respect to rho d_drho_f_mu
+ END_DOC
+ double precision, intent(in)  :: rho,alpha,mu0,beta
+ double precision, intent(out) :: f_mu,d_drho_f_mu
+ f_mu = alpha * (rho)**beta + mu0 * dexp(-rho)
+ d_drho_f_mu = alpha * beta * rho**(beta-1.d0) - mu0 * dexp(-rho)
+
+end
+
+
+subroutine get_all_rho_grad_rho(r1,r2,rho1,rho2,grad_rho1)
+ implicit none
+ BEGIN_DOC
+! returns the density in r1,r2 and grad_rho at r1
+ END_DOC
+ double precision, intent(in) :: r1(3),r2(3)
+ double precision, intent(out):: grad_rho1(3),rho1,rho2
+ double precision              :: dm_a(1), dm_b(1), grad_dm_a(3,1), grad_dm_b(3,1)
+ call density_and_grad_alpha_beta(r1, dm_a, dm_b, grad_dm_a, grad_dm_b)
+ rho1 = dm_a(1) + dm_b(1)
+ grad_rho1(1:3) = grad_dm_a(1:3,1) + grad_dm_b(1:3,1)
+ call density_and_grad_alpha_beta(r2, dm_a, dm_b, grad_dm_a, grad_dm_b)
+ rho2 = dm_a(1) + dm_b(1)
+end
+
+subroutine get_all_f_rho(rho1,rho2,alpha,mu0,beta,f_rho1,d_drho_f_rho1,f_rho2)
+ implicit none
+ BEGIN_DOC
+! returns the values f(mu(r1)), f(mu(r2)) and d/drho(1) f(mu(r1))
+ END_DOC
+ double precision, intent(in) :: rho1,rho2,alpha,mu0,beta
+ double precision, intent(out):: f_rho1,d_drho_f_rho1,f_rho2
+ double precision :: tmp
+ call f_mu_and_deriv_mu(rho1,alpha,mu0,beta,f_rho1,d_drho_f_rho1)
+ call f_mu_and_deriv_mu(rho2,alpha,mu0,beta,f_rho2,tmp)
+end
+
+
+subroutine get_all_f_rho_simple(rho1,rho2,alpha,mu0,beta,f_rho1,d_drho_f_rho1,f_rho2)
+ implicit none
+ BEGIN_DOC
+! returns the values f(mu(r1)), f(mu(r2)) and d/drho(1) f(mu(r1))
+ END_DOC
+ double precision, intent(in) :: rho1,rho2,alpha,mu0,beta
+ double precision, intent(out):: f_rho1,d_drho_f_rho1,f_rho2
+ double precision :: tmp
+ call f_mu_and_deriv_mu_simple(rho1,alpha,mu0,beta,f_rho1,d_drho_f_rho1)
+ call f_mu_and_deriv_mu_simple(rho2,alpha,mu0,beta,f_rho2,tmp)
+end
+
+subroutine f_mu_and_deriv_mu_simple(rho,alpha,mu0,beta,f_mu,d_drho_f_mu)
+ implicit none
+ BEGIN_DOC
+! function giving mu as a function of rho
+!
+! f_mu = alpha * rho**beta + mu0 
+!
+! and its derivative with respect to rho d_drho_f_mu
+ END_DOC
+ double precision, intent(in)  :: rho,alpha,mu0,beta
+ double precision, intent(out) :: f_mu,d_drho_f_mu
+ f_mu = alpha * (rho)**beta + mu0 
+ d_drho_f_mu = alpha * beta * rho**(beta-1.d0) 
+
+end
+

src/non_h_ints_mu/new_grad_tc.irp.f

@@ -284,6 +284,7 @@ BEGIN_PROVIDER [double precision, tc_grad_and_lapl_ao, (ao_num, ao_num, ao_num,
 
   call wall_time(time1)
   print*, ' Wall time for tc_grad_and_lapl_ao = ', time1 - time0
+  call print_memory_usage()
 
 END_PROVIDER 
 

src/non_h_ints_mu/plot_mu_of_r.irp.f

@@ -0,0 +1,33 @@
+program plot_mu_of_r
+ implicit none
+ read_wf = .False.
+ touch read_wf 
+ call routine_print
+
+end
+
+
+subroutine routine_print
+ implicit none
+ character*(128) :: output
+ integer :: i_unit_output,getUnitAndOpen
+ output=trim(ezfio_filename)//'.mu_of_r'
+ i_unit_output = getUnitAndOpen(output,'w')
+ integer :: ipoint,nx
+ double precision :: xmax,xmin,r(3),dx
+ double precision :: mu_val, mu_der(3),dm_a,dm_b,grad
+ xmax =  5.D0
+ xmin = -5.D0
+ nx = 10000
+ dx = (xmax - xmin)/dble(nx)
+ r = 0.d0
+ r(1) = xmin
+ do ipoint = 1, nx
+  call mu_r_val_and_grad(r, r, mu_val, mu_der)
+  call dm_dft_alpha_beta_at_r(r,dm_a,dm_b)
+  grad = mu_der(1)**2 + mu_der(2)**2 + mu_der(3)**2 
+  grad = dsqrt(grad)
+  write(i_unit_output,'(100(F16.7,X))')r(1),mu_val,dm_a+dm_b,grad
+  r(1) += dx
+ enddo
+end

src/non_h_ints_mu/tc_integ.irp.f

@@ -100,6 +100,8 @@ BEGIN_PROVIDER [double precision, int2_grad1_u12_ao, (ao_num, ao_num, n_points_f
       !$OMP END DO
       !$OMP END PARALLEL
 
+      FREE v_ij_erf_rk_cst_mu_j1b v_ij_u_cst_mu_j1b x_v_ij_erf_rk_cst_mu_j1b
+
     elseif(j1b_type .ge. 100) then
 
       PROVIDE final_weight_at_r_vector_extra aos_in_r_array_extra
@@ -176,6 +178,7 @@ BEGIN_PROVIDER [double precision, int2_grad1_u12_ao, (ao_num, ao_num, n_points_f
 
   call wall_time(time1)
   print*, ' wall time for int2_grad1_u12_ao =', time1-time0 
+  call print_memory_usage()
 
 END_PROVIDER
 
@@ -242,6 +245,8 @@ BEGIN_PROVIDER [double precision, int2_grad1_u12_square_ao, (ao_num, ao_num, n_p
       !$OMP END DO
       !$OMP END PARALLEL
 
+      FREE u12sq_j1bsq grad12_j12
+
     else
 
       PROVIDE u12sq_j1bsq u12_grad1_u12_j1b_grad1_j1b grad12_j12
@@ -262,6 +267,8 @@ BEGIN_PROVIDER [double precision, int2_grad1_u12_square_ao, (ao_num, ao_num, n_p
       !$OMP END DO
       !$OMP END PARALLEL
 
+      FREE u12sq_j1bsq u12_grad1_u12_j1b_grad1_j1b grad12_j12
+
     endif
 
   elseif(j1b_type .ge. 100) then
@@ -324,6 +331,7 @@ BEGIN_PROVIDER [double precision, int2_grad1_u12_square_ao, (ao_num, ao_num, n_p
 
   call wall_time(time1)
   print*, ' wall time for int2_grad1_u12_square_ao =', time1-time0 
+  call print_memory_usage()
 
 END_PROVIDER
 

src/non_h_ints_mu/total_tc_int.irp.f

@@ -84,8 +84,11 @@ BEGIN_PROVIDER [double precision, ao_tc_int_chemist, (ao_num, ao_num, ao_num, ao
     enddo
   endif
 
+  FREE tc_grad_square_ao tc_grad_and_lapl_ao ao_two_e_coul
+
   call wall_time(wall1)
   print *, ' wall time for ao_tc_int_chemist ', wall1 - wall0
+  call print_memory_usage()
 
 END_PROVIDER 
 

src/nuclei/nuclei.irp.f

@@ -206,7 +206,12 @@ BEGIN_PROVIDER [ double precision, nuclear_repulsion ]
      enddo
      nuclear_repulsion *= 0.5d0
      if(point_charges)then
-      nuclear_repulsion += pt_chrg_nuclei_interaction + pt_chrg_interaction
+      print*,'bear nuclear repulsion = ',nuclear_repulsion 
+      print*,'adding the interaction between the nuclein and the point charges'
+      print*,'to the usual nuclear repulsion '
+      nuclear_repulsion += pt_chrg_nuclei_interaction 
+      print*,'new nuclear repulsion =  ',nuclear_repulsion 
+      print*,'WARNING: we do not add the interaction between the point charges themselves'
      endif
    end if
 

src/nuclei/point_charges.irp.f

@@ -205,5 +205,8 @@ BEGIN_PROVIDER [ double precision, pt_chrg_nuclei_interaction]
  enddo
  print*,'Interaction between point charges and nuclei'
  print*,'pt_chrg_nuclei_interaction = ',pt_chrg_nuclei_interaction
+ if(point_charges)then
+  provide pt_chrg_interaction
+ endif
 END_PROVIDER 
 

src/scf_utils/fock_matrix.irp.f

@@ -5,6 +5,90 @@
    ! Fock matrix on the MO basis.
    ! For open shells, the ROHF Fock Matrix is ::
    !
+   !       |  Rcc  |  F^b  |  Fcv  |
+   !       |-----------------------|
+   !       |  F^b  |  Roo  |  F^a  |
+   !       |-----------------------|
+   !       |  Fcv  |  F^a  |  Rvv  |
+   !
+   ! C: Core, O: Open, V: Virtual 
+   ! 
+   ! Rcc = Acc Fcc^a + Bcc Fcc^b
+   ! Roo = Aoo Foo^a + Boo Foo^b
+   ! Rvv = Avv Fvv^a + Bvv Fvv^b
+   ! Fcv = (F^a + F^b)/2
+   ! 
+   ! F^a: Fock matrix alpha (MO), F^b: Fock matrix beta (MO)
+   ! A,B: Coupling parameters
+   !
+   ! J. Chem. Phys. 133, 141102 (2010), https://doi.org/10.1063/1.3503173
+   ! Coupling parameters from J. Chem. Phys. 125, 204110 (2006); https://doi.org/10.1063/1.2393223.
+   !       cc   oo   vv
+   !  A  -0.5  0.5  1.5
+   !  B   1.5  0.5 -0.5
+   ! 
+   END_DOC
+   integer                        :: i,j,n
+   if (elec_alpha_num == elec_beta_num) then
+     Fock_matrix_mo = Fock_matrix_mo_alpha
+   else
+     ! Core
+     do j = 1, elec_beta_num
+       ! Core
+       do i = 1, elec_beta_num
+         fock_matrix_mo(i,j) = - 0.5d0 * fock_matrix_mo_alpha(i,j) &
+                               + 1.5d0 * fock_matrix_mo_beta(i,j)
+       enddo
+       ! Open
+       do i = elec_beta_num+1, elec_alpha_num
+         fock_matrix_mo(i,j) = fock_matrix_mo_beta(i,j)
+       enddo
+       ! Virtual
+       do i = elec_alpha_num+1, mo_num
+         fock_matrix_mo(i,j) =   0.5d0 * fock_matrix_mo_alpha(i,j) &
+                               + 0.5d0 * fock_matrix_mo_beta(i,j)
+       enddo
+     enddo
+     ! Open
+     do j = elec_beta_num+1, elec_alpha_num
+       ! Core
+       do i = 1, elec_beta_num
+         fock_matrix_mo(i,j) = fock_matrix_mo_beta(i,j)
+       enddo
+       ! Open
+       do i = elec_beta_num+1, elec_alpha_num
+         fock_matrix_mo(i,j) =   0.5d0 * fock_matrix_mo_alpha(i,j) &
+                               + 0.5d0 * fock_matrix_mo_beta(i,j)
+       enddo
+       ! Virtual
+       do i = elec_alpha_num+1, mo_num
+         fock_matrix_mo(i,j) = fock_matrix_mo_alpha(i,j)
+       enddo
+     enddo
+     ! Virtual
+     do j = elec_alpha_num+1, mo_num
+       ! Core
+       do i = 1, elec_beta_num
+         fock_matrix_mo(i,j) =   0.5d0 * fock_matrix_mo_alpha(i,j) &
+                               + 0.5d0 * fock_matrix_mo_beta(i,j)
+       enddo
+       ! Open
+       do i = elec_beta_num+1, elec_alpha_num
+         fock_matrix_mo(i,j) = fock_matrix_mo_alpha(i,j)
+       enddo
+       ! Virtual
+       do i = elec_alpha_num+1, mo_num
+         fock_matrix_mo(i,j) =   1.5d0 * fock_matrix_mo_alpha(i,j) &
+                               - 0.5d0 * fock_matrix_mo_beta(i,j)
+       enddo
+     enddo
+   endif
+
+   ! Old
+   ! BEGIN_DOC
+   ! Fock matrix on the MO basis.
+   ! For open shells, the ROHF Fock Matrix is ::
+   !
    !       |   F-K    |  F + K/2  |    F     |
    !       |---------------------------------|
    !       | F + K/2  |     F     |  F - K/2 |
@@ -16,64 +100,64 @@
    !
    ! K = Fb - Fa
    !
-   END_DOC
+   ! END_DOC
-   integer                        :: i,j,n
+   !integer                        :: i,j,n
-   if (elec_alpha_num == elec_beta_num) then
+   !if (elec_alpha_num == elec_beta_num) then
-     Fock_matrix_mo = Fock_matrix_mo_alpha
+   !  Fock_matrix_mo = Fock_matrix_mo_alpha
-   else
+   !else
 
-     do j=1,elec_beta_num
+   !  do j=1,elec_beta_num
-       ! F-K
+   !    ! F-K
-       do i=1,elec_beta_num !CC
+   !    do i=1,elec_beta_num !CC
-         Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))&
+   !      Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))&
-             - (Fock_matrix_mo_beta(i,j) - Fock_matrix_mo_alpha(i,j))
+   !          - (Fock_matrix_mo_beta(i,j) - Fock_matrix_mo_alpha(i,j))
-       enddo
+   !    enddo
-       ! F+K/2
+   !    ! F+K/2
-       do i=elec_beta_num+1,elec_alpha_num  !CA
+   !    do i=elec_beta_num+1,elec_alpha_num  !CA
-         Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))&
+   !      Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))&
-             + 0.5d0*(Fock_matrix_mo_beta(i,j) - Fock_matrix_mo_alpha(i,j))
+   !          + 0.5d0*(Fock_matrix_mo_beta(i,j) - Fock_matrix_mo_alpha(i,j))
-       enddo
+   !    enddo
-       ! F
+   !    ! F
-       do i=elec_alpha_num+1, mo_num !CV
+   !    do i=elec_alpha_num+1, mo_num !CV
-         Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))
+   !      Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))
-       enddo
+   !    enddo
-     enddo
+   !  enddo
 
-     do j=elec_beta_num+1,elec_alpha_num
+   !  do j=elec_beta_num+1,elec_alpha_num
-       ! F+K/2
+   !    ! F+K/2
-       do i=1,elec_beta_num !AC
+   !    do i=1,elec_beta_num !AC
-         Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))&
+   !      Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))&
-             + 0.5d0*(Fock_matrix_mo_beta(i,j) - Fock_matrix_mo_alpha(i,j))
+   !          + 0.5d0*(Fock_matrix_mo_beta(i,j) - Fock_matrix_mo_alpha(i,j))
-       enddo
+   !    enddo
-       ! F
+   !    ! F
-       do i=elec_beta_num+1,elec_alpha_num !AA
+   !    do i=elec_beta_num+1,elec_alpha_num !AA
-         Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))
+   !      Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))
-       enddo
+   !    enddo
-       ! F-K/2
+   !    ! F-K/2
-       do i=elec_alpha_num+1, mo_num !AV
+   !    do i=elec_alpha_num+1, mo_num !AV
-         Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))&
+   !      Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))&
-             - 0.5d0*(Fock_matrix_mo_beta(i,j) - Fock_matrix_mo_alpha(i,j))
+   !          - 0.5d0*(Fock_matrix_mo_beta(i,j) - Fock_matrix_mo_alpha(i,j))
-       enddo
+   !    enddo
-     enddo
+   !  enddo
 
-     do j=elec_alpha_num+1, mo_num
+   !  do j=elec_alpha_num+1, mo_num
-       ! F
+   !    ! F
-       do i=1,elec_beta_num !VC
+   !    do i=1,elec_beta_num !VC
-         Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))
+   !      Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))
-       enddo
+   !    enddo
-       ! F-K/2
+   !    ! F-K/2
-       do i=elec_beta_num+1,elec_alpha_num !VA
+   !    do i=elec_beta_num+1,elec_alpha_num !VA
-         Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))&
+   !      Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j))&
-             - 0.5d0*(Fock_matrix_mo_beta(i,j) - Fock_matrix_mo_alpha(i,j))
+   !          - 0.5d0*(Fock_matrix_mo_beta(i,j) - Fock_matrix_mo_alpha(i,j))
-       enddo
+   !    enddo
-       ! F+K
+   !    ! F+K
-       do i=elec_alpha_num+1,mo_num !VV
+   !    do i=elec_alpha_num+1,mo_num !VV
-         Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j)) &
+   !      Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_mo_alpha(i,j)+Fock_matrix_mo_beta(i,j)) &
-             + (Fock_matrix_mo_beta(i,j) - Fock_matrix_mo_alpha(i,j))
+   !          + (Fock_matrix_mo_beta(i,j) - Fock_matrix_mo_alpha(i,j))
-       enddo
+   !    enddo
-     enddo
+   !  enddo
 
-   endif
+   !endif
 
    do i = 1, mo_num
      Fock_matrix_diag_mo(i) = Fock_matrix_mo(i,i)
@@ -115,8 +199,6 @@
 
 END_PROVIDER
 
-
-
 BEGIN_PROVIDER [ double precision, Fock_matrix_mo_alpha, (mo_num,mo_num) ]
    implicit none
    BEGIN_DOC

src/tc_bi_ortho/31.tc_bi_ortho.bats

@@ -4,12 +4,16 @@ source $QP_ROOT/tests/bats/common.bats.sh
 source $QP_ROOT/quantum_package.rc
 
 
+function get_e() {
+  grep "eigval_right_tc_bi_orth" $1 | cut -d '=' -f 2 | xargs
+}
+
 function run_Ne() {
   qp set_file Ne_tc_scf
   qp run cisd
   qp run tc_bi_ortho | tee Ne_tc_scf.cisd_tc_bi_ortho.out
   eref=-128.77020441279302
-  energy="$(grep "eigval_right_tc_bi_orth =" Ne_tc_scf.cisd_tc_bi_ortho.out)"
+  energy=$(get_e Ne_tc_scf.cisd_tc_bi_ortho.out)
   eq $energy $eref 1e-6
 }
 
@@ -24,7 +28,7 @@ function run_C() {
   qp run cisd
   qp run tc_bi_ortho | tee C_tc_scf.cisd_tc_bi_ortho.out
   eref=-37.757536149952514
-  energy="$(grep "eigval_right_tc_bi_orth =" C_tc_scf.cisd_tc_bi_ortho.out)"
+  energy=$(get_e C_tc_scf.cisd_tc_bi_ortho.out)
   eq $energy $eref 1e-6
 }
 
@@ -38,7 +42,7 @@ function run_O() {
   qp run cisd
   qp run tc_bi_ortho | tee O_tc_scf.cisd_tc_bi_ortho.out
   eref=-74.908518517716161
-  energy="$(grep "eigval_right_tc_bi_orth =" O_tc_scf.cisd_tc_bi_ortho.out)"
+  energy=$(get_e O_tc_scf.cisd_tc_bi_ortho.out)
   eq $energy $eref 1e-6
 }
 

src/tc_bi_ortho/dressing_vectors_lr.irp.f

@@ -27,7 +27,7 @@ subroutine get_delta_bitc_right(psidet, psicoef, ndet, Nint, delta)
 
   i = 1
   j = 1
-  call htilde_mu_mat_bi_ortho(psidet(1,1,i), psidet(1,1,j), Nint, htc_mono, htc_twoe, htc_three, htc_tot)
+  call htilde_mu_mat_bi_ortho_slow(psidet(1,1,i), psidet(1,1,j), Nint, htc_mono, htc_twoe, htc_three, htc_tot)
   call hmat_bi_ortho         (psidet(1,1,i), psidet(1,1,j), Nint, h_mono, h_twoe, h_tot)
 
   delta = 0.d0
@@ -39,7 +39,7 @@ subroutine get_delta_bitc_right(psidet, psicoef, ndet, Nint, delta)
     do j = 1, ndet
 
       ! < I | Htilde | J >
-      call htilde_mu_mat_bi_ortho(psidet(1,1,i), psidet(1,1,j), Nint, htc_mono, htc_twoe, htc_three, htc_tot)
+      call htilde_mu_mat_bi_ortho_slow(psidet(1,1,i), psidet(1,1,j), Nint, htc_mono, htc_twoe, htc_three, htc_tot)
       ! < I | H | J >
       call hmat_bi_ortho(psidet(1,1,i), psidet(1,1,j), Nint, h_mono, h_twoe, h_tot)
 
@@ -78,7 +78,7 @@ subroutine get_htc_bitc_right(psidet, psicoef, ndet, Nint, delta)
 
   i = 1
   j = 1
-  call htilde_mu_mat_bi_ortho(psidet(1,1,i), psidet(1,1,j), Nint, htc_mono, htc_twoe, htc_three, htc_tot)
+  call htilde_mu_mat_bi_ortho_slow(psidet(1,1,i), psidet(1,1,j), Nint, htc_mono, htc_twoe, htc_three, htc_tot)
 
   delta = 0.d0
  !$OMP PARALLEL DO DEFAULT(NONE) SCHEDULE(dynamic,8)   &
@@ -88,7 +88,7 @@ subroutine get_htc_bitc_right(psidet, psicoef, ndet, Nint, delta)
     do j = 1, ndet
 
       ! < I | Htilde | J >
-      call htilde_mu_mat_bi_ortho(psidet(1,1,i), psidet(1,1,j), Nint, htc_mono, htc_twoe, htc_three, htc_tot)
+      call htilde_mu_mat_bi_ortho_slow(psidet(1,1,i), psidet(1,1,j), Nint, htc_mono, htc_twoe, htc_three, htc_tot)
 
       delta(i) = delta(i) + psicoef(j) * htc_tot
     enddo

src/tc_bi_ortho/e_corr_bi_ortho.irp.f

@@ -2,7 +2,7 @@
  BEGIN_PROVIDER [ double precision, e_tilde_00]
  implicit none
  double precision :: hmono,htwoe,hthree,htot
- call htilde_mu_mat_bi_ortho(HF_bitmask,HF_bitmask,N_int,hmono,htwoe,hthree,htot)
+ call htilde_mu_mat_bi_ortho_slow(HF_bitmask,HF_bitmask,N_int,hmono,htwoe,hthree,htot)
  e_tilde_00 = htot
  END_PROVIDER 
 
@@ -18,11 +18,11 @@
  do i = 1, N_det
   call get_excitation_degree(HF_bitmask,psi_det(1,1,i),degree,N_int)
   if(degree == 1 .or. degree == 2)then
-   call htilde_mu_mat_bi_ortho(psi_det(1,1,i),HF_bitmask,N_int,hmono,htwoe,hthree,htilde_ij)
+   call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,i),HF_bitmask,N_int,hmono,htwoe,hthree,htilde_ij)
-   call htilde_mu_mat_bi_ortho(psi_det(1,1,i),psi_det(1,1,i),N_int,hmono,htwoe,hthree,e_i0)
+   call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,i),psi_det(1,1,i),N_int,hmono,htwoe,hthree,e_i0)
    delta_e = e_tilde_00 - e_i0
    coef_pt1 = htilde_ij / delta_e
-   call htilde_mu_mat_bi_ortho(HF_bitmask,psi_det(1,1,i),N_int,hmono,htwoe,hthree,htilde_ij)
+   call htilde_mu_mat_bi_ortho_slow(HF_bitmask,psi_det(1,1,i),N_int,hmono,htwoe,hthree,htilde_ij)
    e_pt2_tc_bi_orth += coef_pt1 * htilde_ij
    if(degree == 1)then
     e_pt2_tc_bi_orth_single += coef_pt1 * htilde_ij
@@ -37,7 +37,7 @@
  BEGIN_PROVIDER [ double precision, e_tilde_bi_orth_00]
  implicit none
  double precision :: hmono,htwoe,hthree,htilde_ij
- call htilde_mu_mat_bi_ortho(HF_bitmask,HF_bitmask,N_int,hmono,htwoe,hthree,e_tilde_bi_orth_00)
+ call htilde_mu_mat_bi_ortho_slow(HF_bitmask,HF_bitmask,N_int,hmono,htwoe,hthree,e_tilde_bi_orth_00)
  e_tilde_bi_orth_00 += nuclear_repulsion
  END_PROVIDER 
 
@@ -57,7 +57,7 @@
  e_corr_double_bi_orth = 0.d0
  do i = 1, N_det
   call get_excitation_degree(HF_bitmask,psi_det(1,1,i),degree,N_int)
-  call htilde_mu_mat_bi_ortho(HF_bitmask,psi_det(1,1,i),N_int,hmono,htwoe,hthree,htilde_ij)
+  call htilde_mu_mat_bi_ortho_slow(HF_bitmask,psi_det(1,1,i),N_int,hmono,htwoe,hthree,htilde_ij)
   if(degree == 1)then
    e_corr_single_bi_orth += reigvec_tc_bi_orth(i,1) * htilde_ij/reigvec_tc_bi_orth(1,1)
    e_corr_single_bi_orth_abs += dabs(reigvec_tc_bi_orth(i,1) * htilde_ij/reigvec_tc_bi_orth(1,1))
@@ -80,7 +80,7 @@
  do i = 1, N_det
   accu += reigvec_tc_bi_orth(i,1) * leigvec_tc_bi_orth(i,1)
   do j = 1, N_det
-   call htilde_mu_mat_bi_ortho(psi_det(1,1,j),psi_det(1,1,i),N_int,hmono,htwoe,hthree,htilde_ij)
+   call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,j),psi_det(1,1,i),N_int,hmono,htwoe,hthree,htilde_ij)
    e_tc_left_right += htilde_ij * reigvec_tc_bi_orth(i,1) * leigvec_tc_bi_orth(j,1)
   enddo
  enddo
@@ -99,8 +99,8 @@ BEGIN_PROVIDER [ double precision, coef_pt1_bi_ortho, (N_det)]
   if(degree==0)then
    coef_pt1_bi_ortho(i) = 1.d0
   else
-   call htilde_mu_mat_bi_ortho(psi_det(1,1,i),HF_bitmask,N_int,hmono,htwoe,hthree,htilde_ij)
+   call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,i),HF_bitmask,N_int,hmono,htwoe,hthree,htilde_ij)
-   call htilde_mu_mat_bi_ortho(psi_det(1,1,i),psi_det(1,1,i),N_int,hmono,htwoe,hthree,e_i0)
+   call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,i),psi_det(1,1,i),N_int,hmono,htwoe,hthree,e_i0)
    delta_e = e_tilde_00 - e_i0
    coef_pt1 = htilde_ij / delta_e
    coef_pt1_bi_ortho(i)= coef_pt1

src/tc_bi_ortho/h_tc_bi_ortho_psi.irp.f

@@ -1,4 +1,4 @@
-subroutine htc_bi_ortho_calc_tdav(v, u, N_st, sze)
+subroutine htc_bi_ortho_calc_tdav_slow(v, u, N_st, sze)
 
   use bitmasks
 
@@ -27,7 +27,7 @@ subroutine htc_bi_ortho_calc_tdav(v, u, N_st, sze)
 
   i = 1
   j = 1
-  call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,i), psi_det(1,1,j), N_int, htot)
+  call htilde_mu_mat_bi_ortho_tot_slow(psi_det(1,1,i), psi_det(1,1,j), N_int, htot)
 
   v = 0.d0
  !$OMP PARALLEL DO DEFAULT(NONE) SCHEDULE(dynamic,8) &
@@ -36,7 +36,7 @@ subroutine htc_bi_ortho_calc_tdav(v, u, N_st, sze)
   do istate = 1, N_st
     do i = 1, sze
       do j = 1, sze
-        call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,i), psi_det(1,1,j), N_int, htot)
+        call htilde_mu_mat_bi_ortho_tot_slow(psi_det(1,1,i), psi_det(1,1,j), N_int, htot)
         v(i,istate) = v(i,istate) + htot * u(j,istate)
       enddo
     enddo 
@@ -45,7 +45,7 @@ subroutine htc_bi_ortho_calc_tdav(v, u, N_st, sze)
 
 end 
 
-subroutine htcdag_bi_ortho_calc_tdav(v, u, N_st, sze)
+subroutine htcdag_bi_ortho_calc_tdav_slow(v, u, N_st, sze)
 
   use bitmasks
 
@@ -71,7 +71,7 @@ subroutine htcdag_bi_ortho_calc_tdav(v, u, N_st, sze)
 
   i = 1
   j = 1
-  call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,i), psi_det(1,1,j), N_int, htot)
+  call htilde_mu_mat_bi_ortho_tot_slow(psi_det(1,1,i), psi_det(1,1,j), N_int, htot)
 
   v = 0.d0
 
@@ -81,7 +81,7 @@ subroutine htcdag_bi_ortho_calc_tdav(v, u, N_st, sze)
   do istate = 1, N_st
     do i = 1, sze
       do j = 1, sze
-        call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,j), psi_det(1,1,i), N_int, htot)
+        call htilde_mu_mat_bi_ortho_tot_slow(psi_det(1,1,j), psi_det(1,1,i), N_int, htot)
         v(i,istate) = v(i,istate) + htot * u(j,istate)
       enddo
     enddo

src/tc_bi_ortho/normal_ordered_old.irp.f

@@ -0,0 +1,392 @@
+
+! ---
+
+BEGIN_PROVIDER [ double precision, normal_two_body_bi_orth_old, (mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC 
+  ! Normal ordering of the three body interaction on the HF density
+  END_DOC 
+
+  use bitmasks ! you need to include the bitmasks_module.f90 features
+
+  implicit none
+
+  integer                        :: i, h1, p1, h2, p2
+  integer                        :: hh1, hh2, pp1, pp2
+  integer                        :: Ne(2)
+  double precision               :: hthree_aba, hthree_aaa, hthree_aab
+  double precision               :: wall0, wall1
+  integer,           allocatable :: occ(:,:)
+  integer(bit_kind), allocatable :: key_i_core(:,:)
+
+  print*,' Providing normal_two_body_bi_orth_old ...'
+  call wall_time(wall0)
+ 
+  PROVIDE N_int
+
+  if(read_tc_norm_ord) then
+
+    open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/normal_two_body_bi_orth_old', action="read")
+      read(11) normal_two_body_bi_orth_old
+    close(11)
+
+  else
+
+    PROVIDE N_int
+
+    allocate( occ(N_int*bit_kind_size,2) )
+    allocate( key_i_core(N_int,2) )
+
+    if(core_tc_op) then
+      do i = 1, N_int
+        key_i_core(i,1) = xor(ref_bitmask(i,1),core_bitmask(i,1))
+        key_i_core(i,2) = xor(ref_bitmask(i,2),core_bitmask(i,2))
+      enddo
+      call bitstring_to_list_ab(key_i_core,occ,Ne,N_int)
+    else
+      call bitstring_to_list_ab(ref_bitmask,occ,Ne,N_int)
+    endif
+
+    normal_two_body_bi_orth_old = 0.d0
+
+    !$OMP PARALLEL                                                                         &
+    !$OMP DEFAULT (NONE)                                                                   &
+    !$OMP PRIVATE (hh1, h1, hh2, h2, pp1, p1, pp2, p2, hthree_aba, hthree_aab, hthree_aaa) & 
+    !$OMP SHARED (N_int, n_act_orb, list_act, Ne, occ, normal_two_body_bi_orth_old)
+    !$OMP DO SCHEDULE (static) 
+    do hh1 = 1, n_act_orb
+      h1 = list_act(hh1) 
+      do pp1 = 1, n_act_orb
+        p1 = list_act(pp1)
+        do hh2 = 1, n_act_orb
+          h2 = list_act(hh2) 
+          do pp2 = 1, n_act_orb
+            p2 = list_act(pp2)
+            ! all contributions from the 3-e terms to the double excitations 
+            ! s1:(h1-->p1), s2:(h2-->p2) from the HF reference determinant 
+      
+
+            ! opposite spin double excitations : s1 /= s2
+            call give_aba_contraction(N_int, h1, h2, p1, p2, Ne, occ, hthree_aba)
+
+            ! same spin double excitations : s1 == s2 
+            if(h1<h2.and.p1.gt.p2)then
+             ! with opposite spin contributions 
+             call give_aab_contraction(N_int, h2, h1, p1, p2, Ne, occ, hthree_aab) ! exchange h1<->h2
+             ! same spin double excitations with same spin contributions 
+             if(Ne(2).ge.3)then
+               call give_aaa_contraction(N_int, h2, h1, p1, p2, Ne, occ, hthree_aaa) ! exchange h1<->h2
+             else
+               hthree_aaa = 0.d0
+             endif
+            else
+             ! with opposite spin contributions 
+             call give_aab_contraction(N_int, h1, h2, p1, p2, Ne, occ, hthree_aab)
+             if(Ne(2).ge.3)then
+              ! same spin double excitations with same spin contributions 
+               call give_aaa_contraction(N_int, h1, h2, p1, p2, Ne, occ, hthree_aaa)
+             else
+               hthree_aaa = 0.d0
+             endif
+            endif
+
+            normal_two_body_bi_orth_old(p2,h2,p1,h1) = 0.5d0*(hthree_aba + hthree_aab + hthree_aaa)
+          enddo
+        enddo
+      enddo
+    enddo
+    !$OMP END DO
+    !$OMP END PARALLEL
+
+    deallocate( occ )
+    deallocate( key_i_core )
+  endif
+
+  if(write_tc_norm_ord.and.mpi_master) then
+    open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/normal_two_body_bi_orth_old', action="write")
+      call ezfio_set_work_empty(.False.)
+      write(11) normal_two_body_bi_orth_old
+      close(11)
+      call ezfio_set_tc_keywords_io_tc_integ('Read')
+  endif
+
+  call wall_time(wall1)
+  print*,' Wall time for normal_two_body_bi_orth_old ', wall1-wall0
+
+END_PROVIDER 
+
+! ---
+
+subroutine give_aba_contraction(Nint, h1, h2, p1, p2, Ne, occ, hthree)
+
+  use bitmasks ! you need to include the bitmasks_module.f90 features
+
+  implicit none
+  integer, intent(in)           :: Nint, h1, h2, p1, p2
+  integer, intent(in)           :: Ne(2), occ(Nint*bit_kind_size,2)
+  double precision, intent(out) :: hthree
+  integer                       :: ii, i
+  double precision              :: int_direct, int_exc_12, int_exc_13, integral
+
+  !!!! double alpha/beta
+  hthree = 0.d0
+
+  do ii = 1, Ne(2) ! purely closed shell part 
+    i = occ(ii,2)
+
+    call give_integrals_3_body_bi_ort(i, p2, p1, i, h2, h1, integral)
+    int_direct = -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p1, p2, i, i, h2, h1, integral)
+    int_exc_13 = -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p2, i, p1, i, h2, h1, integral)
+    int_exc_12 = -1.d0 * integral
+
+    hthree += 2.d0 * int_direct - 1.d0 * (int_exc_13 + int_exc_12)
+  enddo
+
+  do ii = Ne(2) + 1, Ne(1) ! purely open-shell part 
+    i = occ(ii,1)
+
+    call give_integrals_3_body_bi_ort(i, p2, p1, i, h2, h1, integral)
+    int_direct = -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p1, p2, i, i, h2, h1, integral)
+    int_exc_13 = -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p2, i, p1, i, h2, h1, integral)
+    int_exc_12 = -1.d0 * integral
+
+    hthree += 1.d0 * int_direct - 0.5d0 * (int_exc_13 + int_exc_12)
+  enddo
+
+  return
+end
+
+! ---
+
+BEGIN_PROVIDER [ double precision, normal_two_body_bi_orth_ab, (mo_num, mo_num, mo_num, mo_num)]
+
+  BEGIN_DOC
+  ! Normal ordered two-body sector of the three-body terms for opposite spin double excitations 
+  END_DOC
+
+  use bitmasks ! you need to include the bitmasks_module.f90 features
+
+  implicit none
+  integer                        :: h1, p1, h2, p2, i
+  integer                        :: hh1, hh2, pp1, pp2
+  integer                        :: Ne(2)
+  integer,           allocatable :: occ(:,:)
+  integer(bit_kind), allocatable :: key_i_core(:,:)
+  double precision               :: hthree
+
+  PROVIDE N_int
+
+  allocate( key_i_core(N_int,2) )
+  allocate( occ(N_int*bit_kind_size,2) )
+
+  if(core_tc_op) then
+    do i = 1, N_int
+      key_i_core(i,1) = xor(ref_bitmask(i,1),core_bitmask(i,1))
+      key_i_core(i,2) = xor(ref_bitmask(i,2),core_bitmask(i,2))
+    enddo
+    call bitstring_to_list_ab(key_i_core,occ,Ne,N_int)
+  else
+    call bitstring_to_list_ab(ref_bitmask,occ,Ne,N_int)
+  endif
+
+  normal_two_body_bi_orth_ab = 0.d0
+  do hh1 = 1, n_act_orb
+    h1 = list_act(hh1) 
+    do pp1 = 1, n_act_orb
+      p1 = list_act(pp1)
+      do hh2 = 1, n_act_orb
+        h2 = list_act(hh2) 
+        do pp2 = 1, n_act_orb
+          p2 = list_act(pp2)
+          call give_aba_contraction(N_int, h1, h2, p1, p2, Ne, occ, hthree)
+
+          normal_two_body_bi_orth_ab(p2,h2,p1,h1) = hthree    
+        enddo
+      enddo
+    enddo
+  enddo
+
+  deallocate( key_i_core )
+  deallocate( occ )
+
+END_PROVIDER 
+
+! ---
+
+BEGIN_PROVIDER [ double precision, normal_two_body_bi_orth_aa_bb, (n_act_orb, n_act_orb, n_act_orb, n_act_orb)]
+
+  BEGIN_DOC
+  ! Normal ordered two-body sector of the three-body terms for same spin double excitations 
+  END_DOC
+
+  use bitmasks ! you need to include the bitmasks_module.f90 features
+
+  implicit none
+  integer                        :: i,ii,j,h1,p1,h2,p2
+  integer                        :: hh1,hh2,pp1,pp2
+  integer                        :: Ne(2)
+  integer,           allocatable :: occ(:,:)
+  integer(bit_kind), allocatable :: key_i_core(:,:)
+  double precision               :: hthree_aab, hthree_aaa
+
+  PROVIDE N_int
+
+  allocate( key_i_core(N_int,2) )
+  allocate( occ(N_int*bit_kind_size,2) )
+
+  if(core_tc_op)then
+    do i = 1, N_int
+      key_i_core(i,1) = xor(ref_bitmask(i,1),core_bitmask(i,1))
+      key_i_core(i,2) = xor(ref_bitmask(i,2),core_bitmask(i,2))
+    enddo
+    call bitstring_to_list_ab(key_i_core, occ, Ne, N_int)
+  else
+    call bitstring_to_list_ab(ref_bitmask, occ, Ne, N_int)
+  endif
+
+  normal_two_body_bi_orth_aa_bb = 0.d0
+  do hh1 = 1, n_act_orb
+    h1 = list_act(hh1) 
+    do pp1 = 1 , n_act_orb
+      p1 = list_act(pp1)
+      do hh2 = 1, n_act_orb
+        h2 = list_act(hh2) 
+        do pp2 = 1 , n_act_orb
+          p2 = list_act(pp2)
+          if(h1<h2.and.p1.gt.p2)then
+           call give_aab_contraction(N_int, h2, h1, p1, p2, Ne, occ, hthree_aab) ! exchange h1<->h2
+           if(Ne(2).ge.3)then
+             call give_aaa_contraction(N_int, h2, h1, p1, p2, Ne, occ, hthree_aaa) ! exchange h1<->h2
+           else
+             hthree_aaa = 0.d0
+           endif
+          else
+           call give_aab_contraction(N_int, h1, h2, p1, p2, Ne, occ, hthree_aab)
+           if(Ne(2).ge.3)then
+             call give_aaa_contraction(N_int, h1, h2, p1, p2, Ne, occ, hthree_aaa)
+           else
+             hthree_aaa = 0.d0
+           endif
+          endif
+          normal_two_body_bi_orth_aa_bb(p2,h2,p1,h1) = hthree_aab + hthree_aaa
+        enddo
+      enddo
+    enddo
+  enddo
+
+  deallocate( key_i_core )
+  deallocate( occ )
+
+END_PROVIDER 
+
+! ---
+
+subroutine give_aaa_contraction(Nint, h1, h2, p1, p2, Ne, occ, hthree)
+
+  BEGIN_DOC
+  ! pure same spin contribution to same spin double excitation s1=h1,p1, s2=h2,p2, with s1==s2
+  END_DOC
+
+  use bitmasks ! you need to include the bitmasks_module.f90 features
+
+  implicit none
+  integer, intent(in)           :: Nint, h1, h2, p1, p2
+  integer, intent(in)           :: Ne(2), occ(Nint*bit_kind_size,2)
+  double precision, intent(out) :: hthree
+  integer                       :: ii,i
+  double precision              :: int_direct,int_exc_12,int_exc_13,int_exc_23
+  double precision              :: integral,int_exc_l,int_exc_ll
+
+  hthree = 0.d0
+  do ii = 1, Ne(2) ! purely closed shell part 
+    i = occ(ii,2)
+
+    call give_integrals_3_body_bi_ort(i, p2, p1, i, h2, h1, integral)
+    int_direct = -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p2, p1, i, i, h2, h1, integral)
+    int_exc_l = -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p1, i, p2, i, h2, h1, integral)
+    int_exc_ll= -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p2, i, p1, i, h2, h1, integral)
+    int_exc_12= -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p1, p2, i, i, h2, h1, integral)
+    int_exc_13= -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(i, p1, p2, i, h2, h1, integral)
+    int_exc_23= -1.d0 * integral
+
+    hthree +=  1.d0 * int_direct + int_exc_l + int_exc_ll - (int_exc_12 + int_exc_13 + int_exc_23)
+  enddo
+
+  do ii = Ne(2)+1,Ne(1) ! purely open-shell part 
+    i = occ(ii,1)
+
+    call give_integrals_3_body_bi_ort(i, p2, p1, i, h2, h1, integral)
+    int_direct = -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p2, p1, i , i, h2, h1, integral)
+    int_exc_l = -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p1, i, p2, i, h2, h1, integral)
+    int_exc_ll = -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p2, i, p1, i, h2, h1, integral)
+    int_exc_12 = -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p1, p2, i, i, h2, h1, integral)
+    int_exc_13 = -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(i, p1, p2, i, h2, h1, integral)
+    int_exc_23 = -1.d0 * integral
+
+    !hthree +=  1.d0 * int_direct + 0.5d0 * (int_exc_l + int_exc_ll - (int_exc_12 + int_exc_13 + int_exc_23))
+    hthree +=  0.5d0 * int_direct + 0.5d0 * (int_exc_l + int_exc_ll - (int_exc_12 + int_exc_13 + int_exc_23))
+  enddo
+
+  return
+end
+
+! ---
+
+subroutine give_aab_contraction(Nint, h1, h2, p1, p2, Ne, occ, hthree)
+
+  use bitmasks ! you need to include the bitmasks_module.f90 features
+
+  implicit none
+  integer,          intent(in)  :: Nint, h1, h2, p1, p2
+  integer,          intent(in)  :: Ne(2), occ(Nint*bit_kind_size,2)
+  double precision, intent(out) :: hthree
+  integer                       :: ii, i
+  double precision              :: int_direct, int_exc_12, int_exc_13, int_exc_23
+  double precision              :: integral, int_exc_l, int_exc_ll
+
+  hthree = 0.d0
+  do ii = 1, Ne(2) ! purely closed shell part 
+    i = occ(ii,2)
+
+    call give_integrals_3_body_bi_ort(p2, p1, i, h2, h1, i, integral)
+    int_direct = -1.d0 * integral
+
+    call give_integrals_3_body_bi_ort(p1, p2, i, h2, h1, i, integral)
+    int_exc_23= -1.d0 * integral
+
+    hthree +=  1.d0 * int_direct - int_exc_23
+  enddo
+
+  return
+end
+
+! ---
+

src/tc_bi_ortho/print_tc_wf.irp.f

@@ -49,12 +49,12 @@ subroutine routine
  do i = 1, N_det
   call get_excitation_degree(HF_bitmask,psi_det(1,1,i),degree,N_int)
    if(degree == 1 .or. degree == 2)then
-    call htilde_mu_mat_bi_ortho(psi_det(1,1,i),HF_bitmask,N_int,hmono,htwoe,hthree,htilde_ij)
+    call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,i),HF_bitmask,N_int,hmono,htwoe,hthree,htilde_ij)
-    call htilde_mu_mat_bi_ortho(psi_det(1,1,i),psi_det(1,1,i),N_int,hmono,htwoe,hthree,e_i0)
+    call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,i),psi_det(1,1,i),N_int,hmono,htwoe,hthree,e_i0)
     delta_e = e_tilde_00 - e_i0
     coef_pt1 = htilde_ij / delta_e
  
-    call htilde_mu_mat_bi_ortho(HF_bitmask,psi_det(1,1,i),N_int,hmono,htwoe,hthree,htilde_ij)
+    call htilde_mu_mat_bi_ortho_slow(HF_bitmask,psi_det(1,1,i),N_int,hmono,htwoe,hthree,htilde_ij)
     contrib_pt = coef_pt1 * htilde_ij
     e_pt2 += contrib_pt
  

src/tc_bi_ortho/pt2_tc_cisd.irp.f

@@ -36,11 +36,11 @@ subroutine routine
  e_corr_abs = 0.d0
  e_corr_pos = 0.d0
  e_corr_neg = 0.d0
- call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,1), psi_det(1,1,1), N_int, e00) 
+ call htilde_mu_mat_bi_ortho_tot_slow(psi_det(1,1,1), psi_det(1,1,1), N_int, e00) 
  do i = 2, N_det
-  call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,i), psi_det(1,1,1), N_int, hi0) 
+  call htilde_mu_mat_bi_ortho_tot_slow(psi_det(1,1,i), psi_det(1,1,1), N_int, hi0) 
-  call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,1), psi_det(1,1,i), N_int, h0i) 
+  call htilde_mu_mat_bi_ortho_tot_slow(psi_det(1,1,1), psi_det(1,1,i), N_int, h0i) 
-  call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,i), psi_det(1,1,i), N_int, ei) 
+  call htilde_mu_mat_bi_ortho_tot_slow(psi_det(1,1,i), psi_det(1,1,i), N_int, ei) 
   call get_excitation_degree(psi_det(1,1,1), psi_det(1,1,i),degree,N_int)
   call get_excitation(psi_det(1,1,1), psi_det(1,1,i),exc,degree,phase,N_int)
   call decode_exc(exc,degree,h1,p1,h2,p2,s1,s2)

src/tc_bi_ortho/slater_tc_3e_slow.irp.f

@@ -1,23 +1,5 @@
-subroutine provide_all_three_ints_bi_ortho
- implicit none
- BEGIN_DOC
-! routine that provides all necessary three-electron integrals 
- END_DOC
- if(three_body_h_tc)then
-  PROVIDE three_e_3_idx_direct_bi_ort three_e_3_idx_cycle_1_bi_ort three_e_3_idx_cycle_2_bi_ort
-  PROVIDE three_e_3_idx_exch23_bi_ort three_e_3_idx_exch13_bi_ort three_e_3_idx_exch12_bi_ort
-  PROVIDE three_e_4_idx_direct_bi_ort three_e_4_idx_cycle_1_bi_ort three_e_4_idx_cycle_2_bi_ort
-  PROVIDE three_e_4_idx_exch23_bi_ort three_e_4_idx_exch13_bi_ort three_e_4_idx_exch12_bi_ort
- endif
-if(.not.double_normal_ord)then
-  PROVIDE three_e_5_idx_direct_bi_ort three_e_5_idx_cycle_1_bi_ort three_e_5_idx_cycle_2_bi_ort
-  PROVIDE three_e_5_idx_exch23_bi_ort three_e_5_idx_exch13_bi_ort three_e_5_idx_exch12_bi_ort
-else
- PROVIDE normal_two_body_bi_orth
-endif
-end
 
-subroutine diag_htilde_three_body_ints_bi_ort(Nint, key_i, hthree)
+subroutine diag_htilde_three_body_ints_bi_ort_slow(Nint, key_i, hthree)
 
   BEGIN_DOC
   !  diagonal element of htilde ONLY FOR THREE-BODY TERMS WITH BI ORTHONORMAL ORBITALS
@@ -108,7 +90,7 @@ subroutine diag_htilde_three_body_ints_bi_ort(Nint, key_i, hthree)
 end
 
 
-subroutine single_htilde_three_body_ints_bi_ort(Nint, key_j, key_i, hthree)
+subroutine single_htilde_three_body_ints_bi_ort_slow(Nint, key_j, key_i, hthree)
 
   BEGIN_DOC
   ! <key_j | H_tilde | key_i> for single excitation ONLY FOR THREE-BODY TERMS WITH BI ORTHONORMAL ORBITALS
@@ -203,7 +185,7 @@ end
 
 ! ---
 
-subroutine double_htilde_three_body_ints_bi_ort(Nint, key_j, key_i, hthree)
+subroutine double_htilde_three_body_ints_bi_ort_slow(Nint, key_j, key_i, hthree)
 
   BEGIN_DOC
   ! <key_j | H_tilde | key_i> for double excitation ONLY FOR THREE-BODY TERMS  WITH BI ORTHONORMAL ORBITALS
@@ -260,7 +242,8 @@ subroutine double_htilde_three_body_ints_bi_ort(Nint, key_j, key_i, hthree)
       do m = 1, Ne(ispin) ! direct(other_spin) - exchange(s1)
        mm = occ(m,ispin)
        direct_int = three_e_5_idx_direct_bi_ort(mm,p2,h2,p1,h1)
-       exchange_int = three_e_5_idx_exch12_bi_ort(mm,p2,h2,p1,h1)
+!       exchange_int = three_e_5_idx_exch12_bi_ort(mm,p2,h2,p1,h1)
+       exchange_int = three_e_5_idx_direct_bi_ort(mm,p2,h1,p1,h2)
        hthree += direct_int - exchange_int
       enddo
       do m = 1, Ne(s1) ! pure contribution from s1

src/tc_bi_ortho/slater_tc_opt.irp.f

@@ -1,3 +1,37 @@
+! ---
+
+subroutine provide_all_three_ints_bi_ortho()
+
+  BEGIN_DOC
+  ! routine that provides all necessary three-electron integrals 
+  END_DOC
+
+  implicit none
+
+  if(three_body_h_tc) then
+
+    if(three_e_3_idx_term) then
+      PROVIDE three_e_3_idx_direct_bi_ort three_e_3_idx_cycle_1_bi_ort three_e_3_idx_cycle_2_bi_ort
+      PROVIDE three_e_3_idx_exch23_bi_ort three_e_3_idx_exch13_bi_ort three_e_3_idx_exch12_bi_ort
+    endif
+
+    if(three_e_4_idx_term) then
+      PROVIDE three_e_4_idx_direct_bi_ort three_e_4_idx_cycle_1_bi_ort three_e_4_idx_exch23_bi_ort three_e_4_idx_exch13_bi_ort 
+    endif
+
+    if(.not. double_normal_ord .and. three_e_5_idx_term) then
+      PROVIDE three_e_5_idx_direct_bi_ort 
+    elseif(double_normal_ord .and. (.not. three_e_5_idx_term)) then
+      PROVIDE normal_two_body_bi_orth
+    endif
+
+ endif
+
+ return
+end
+
+! ---
+
 subroutine htilde_mu_mat_opt_bi_ortho_tot(key_j, key_i, Nint, htot)
  implicit none
   BEGIN_DOC

src/tc_bi_ortho/slater_tc_opt_diag.irp.f

@@ -7,11 +7,11 @@
 ! Various component of the TC energy for the reference "HF" Slater determinant
  END_DOC 
  double precision :: hmono, htwoe, htot, hthree
- call diag_htilde_mu_mat_bi_ortho(N_int,HF_bitmask , hmono, htwoe, htot)
+ call diag_htilde_mu_mat_bi_ortho_slow(N_int,HF_bitmask , hmono, htwoe, htot)
  ref_tc_energy_1e = hmono
  ref_tc_energy_2e = htwoe 
  if(three_body_h_tc)then
-  call diag_htilde_three_body_ints_bi_ort(N_int, HF_bitmask, hthree)
+  call diag_htilde_three_body_ints_bi_ort_slow(N_int, HF_bitmask, hthree)
   ref_tc_energy_3e = hthree
  else
   ref_tc_energy_3e = 0.d0
@@ -156,7 +156,7 @@ subroutine ac_tc_operator(iorb,ispin,key,hmono,htwoe,hthree,Nint,na,nb)
     htwoe = htwoe + mo_bi_ortho_tc_two_e_jj(occ(i,other_spin),iorb)
   enddo
 
-  if(three_body_h_tc.and.elec_num.gt.2)then
+  if(three_body_h_tc.and.elec_num.gt.2.and.three_e_3_idx_term)then
    !!!!! 3-e part 
    !! same-spin/same-spin
    do j = 1, na
@@ -243,7 +243,7 @@ subroutine a_tc_operator(iorb,ispin,key,hmono,htwoe,hthree,Nint,na,nb)
     htwoe= htwoe- mo_bi_ortho_tc_two_e_jj(occ(i,other_spin),iorb)
   enddo
 
-  if(three_body_h_tc.and.elec_num.gt.2)then
+  if(three_body_h_tc.and.elec_num.gt.2.and.three_e_3_idx_term)then
    !!!!! 3-e part 
    !! same-spin/same-spin
    do j = 1, na

src/tc_bi_ortho/slater_tc_opt_double.irp.f

@@ -42,13 +42,13 @@ subroutine double_htilde_mu_mat_fock_bi_ortho(Nint, key_j, key_i, hmono, htwoe,
    ! opposite spin two-body
     htwoe  = mo_bi_ortho_tc_two_e(p2,p1,h2,h1)
     if(three_body_h_tc.and.elec_num.gt.2)then
-     if(.not.double_normal_ord)then
+     if(.not.double_normal_ord.and.three_e_5_idx_term)then
       if(degree_i>degree_j)then
        call three_comp_two_e_elem(key_j,h1,h2,p1,p2,s1,s2,hthree)
       else
        call three_comp_two_e_elem(key_i,h1,h2,p1,p2,s1,s2,hthree)
       endif
-     elseif(double_normal_ord.and.elec_num.gt.2)then
+     elseif(double_normal_ord)then
       htwoe += normal_two_body_bi_orth(p2,h2,p1,h1)
      endif
     endif
@@ -59,13 +59,13 @@ subroutine double_htilde_mu_mat_fock_bi_ortho(Nint, key_j, key_i, hmono, htwoe,
    ! exchange terms
    htwoe -= mo_bi_ortho_tc_two_e(p1,p2,h2,h1)
    if(three_body_h_tc.and.elec_num.gt.2)then
-    if(.not.double_normal_ord)then
+    if(.not.double_normal_ord.and.three_e_5_idx_term)then
      if(degree_i>degree_j)then
       call three_comp_two_e_elem(key_j,h1,h2,p1,p2,s1,s2,hthree)
      else
       call three_comp_two_e_elem(key_i,h1,h2,p1,p2,s1,s2,hthree)
      endif
-    elseif(double_normal_ord.and.elec_num.gt.2)then
+    elseif(double_normal_ord)then
       htwoe -= normal_two_body_bi_orth(h2,p1,h1,p2)
       htwoe += normal_two_body_bi_orth(h1,p1,h2,p2)
     endif
@@ -136,11 +136,13 @@ subroutine three_comp_two_e_elem(key_i,h1,h2,p1,p2,s1,s2,hthree)
      ! exchange between (h1,p1) and (h2,p2)
      ipart=occ_particle(i,ispin)
      direct_int  = three_e_5_idx_direct_bi_ort(ipart,p2,h2,p1,h1)
-     exchange_int = three_e_5_idx_exch12_bi_ort(ipart,p2,h2,p1,h1)
+!     exchange_int = three_e_5_idx_exch12_bi_ort(ipart,p2,h2,p1,h1)
+     exchange_int = three_e_5_idx_direct_bi_ort(ipart,p2,h1,p1,h2)
      hthree += direct_int - exchange_int
      ihole=occ_hole(i,ispin)
      direct_int  = three_e_5_idx_direct_bi_ort(ihole,p2,h2,p1,h1)
-     exchange_int = three_e_5_idx_exch12_bi_ort(ihole,p2,h2,p1,h1)
+!     exchange_int = three_e_5_idx_exch12_bi_ort(ihole,p2,h2,p1,h1)
+     exchange_int = three_e_5_idx_direct_bi_ort(ihole,p2,h1,p1,h2)
      hthree -= direct_int - exchange_int
     enddo
    !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
@@ -163,11 +165,13 @@ subroutine three_comp_two_e_elem(key_i,h1,h2,p1,p2,s1,s2,hthree)
     ! exchange between (h1,p1) and (h2,p2)
     ipart=occ_particle(i,ispin)
     direct_int  = three_e_5_idx_direct_bi_ort(ipart,p2,h2,p1,h1)
-    exchange_int = three_e_5_idx_exch12_bi_ort(ipart,p2,h2,p1,h1)
+!    exchange_int = three_e_5_idx_exch12_bi_ort(ipart,p2,h2,p1,h1)
+    exchange_int = three_e_5_idx_direct_bi_ort(ipart,p2,h1,p1,h2)
     hthree += direct_int - exchange_int
     ihole=occ_hole(i,ispin)
     direct_int  = three_e_5_idx_direct_bi_ort(ihole,p2,h2,p1,h1)
-    exchange_int = three_e_5_idx_exch12_bi_ort(ihole,p2,h2,p1,h1)
+!    exchange_int = three_e_5_idx_exch12_bi_ort(ihole,p2,h2,p1,h1)
+    exchange_int = three_e_5_idx_direct_bi_ort(ihole,p2,h1,p1,h2)
     hthree -= direct_int - exchange_int
    enddo
   else                         ! (h1,p1) == alpha/(h2,p2) == beta
@@ -342,7 +346,8 @@ subroutine give_contrib_for_aaaa(h1,h2,p1,p2,occ,Ne,contrib)
    m = occ(mm,2)
    direct_int   = three_e_5_idx_direct_bi_ort(mm,p2,h2,p1,h1)
    ! exchange between (h1,p1) and (h2,p2)
-   exchange_int = three_e_5_idx_exch12_bi_ort(mm,p2,h2,p1,h1)
+!   exchange_int = three_e_5_idx_exch12_bi_ort(mm,p2,h2,p1,h1)
+   exchange_int = three_e_5_idx_direct_bi_ort(mm,p2,h1,p1,h2)
    contrib += direct_int - exchange_int
  enddo
 end
@@ -415,7 +420,8 @@ subroutine give_contrib_for_bbbb(h1,h2,p1,p2,occ,Ne,contrib)
    m = occ(mm,1)
    direct_int   = three_e_5_idx_direct_bi_ort(mm,p2,h2,p1,h1)
    ! exchange between (h1,p1) and (h2,p2)
-   exchange_int = three_e_5_idx_exch12_bi_ort(mm,p2,h2,p1,h1)
+!   exchange_int = three_e_5_idx_exch12_bi_ort(mm,p2,h2,p1,h1)
+   exchange_int = three_e_5_idx_direct_bi_ort(mm,p2,h1,p1,h2)
    contrib += direct_int - exchange_int
  enddo
 end

src/tc_bi_ortho/slater_tc_opt_single.irp.f

@@ -106,7 +106,7 @@ subroutine get_single_excitation_from_fock_tc(key_i,key_j,h,p,spin,phase,hmono,h
   htwoe -= buffer_x(i)
  enddo
  hthree = 0.d0
- if (three_body_h_tc.and.elec_num.gt.2)then
+ if (three_body_h_tc.and.elec_num.gt.2.and.three_e_4_idx_term)then
   call three_comp_fock_elem(key_i,h,p,spin,hthree)
  endif
 
@@ -243,7 +243,9 @@ subroutine fock_ac_tc_operator(iorb,ispin,key, h_fock,p_fock, ispin_fock,hthree,
    do j = 1, nb
     jj = occ(j,other_spin) 
     direct_int = three_e_4_idx_direct_bi_ort(jj,iorb,p_fock,h_fock) ! USES 4-IDX TENSOR 
-    exchange_int = three_e_4_idx_exch12_bi_ort(jj,iorb,p_fock,h_fock) ! USES 4-IDX TENSOR 
+    ! TODO
+    ! use transpose
+    exchange_int = three_e_4_idx_exch13_bi_ort(iorb,jj,p_fock,h_fock) ! USES 4-IDX TENSOR 
     hthree += direct_int - exchange_int
    enddo
   else !! ispin NE to ispin_fock
@@ -322,7 +324,8 @@ subroutine fock_a_tc_operator(iorb,ispin,key, h_fock,p_fock, ispin_fock,hthree,N
    do j = 1, nb
     jj = occ(j,other_spin) 
     direct_int = three_e_4_idx_direct_bi_ort(jj,iorb,p_fock,h_fock) ! USES 4-IDX TENSOR 
-    exchange_int = three_e_4_idx_exch12_bi_ort(jj,iorb,p_fock,h_fock) ! USES 4-IDX TENSOR 
+    ! TODO use transpose 
+    exchange_int = three_e_4_idx_exch13_bi_ort(iorb,jj,p_fock,h_fock) ! USES 4-IDX TENSOR 
     hthree -= direct_int - exchange_int
    enddo
   else !! ispin NE to ispin_fock

src/tc_bi_ortho/slater_tc_slow.irp.f

@@ -1,7 +1,7 @@
 
 ! ---
 
-subroutine htilde_mu_mat_bi_ortho_tot(key_j, key_i, Nint, htot)
+subroutine htilde_mu_mat_bi_ortho_tot_slow(key_j, key_i, Nint, htot)
 
   BEGIN_DOC
   ! <key_j | H_tilde | key_i> where |key_j> is developed on the LEFT basis and |key_i> is developed on the RIGHT basis
@@ -24,14 +24,14 @@ subroutine htilde_mu_mat_bi_ortho_tot(key_j, key_i, Nint, htot)
   if(degree.gt.2)then
     htot = 0.d0
   else
-    call htilde_mu_mat_bi_ortho(key_j, key_i, Nint, hmono, htwoe, hthree, htot)
+    call htilde_mu_mat_bi_ortho_slow(key_j, key_i, Nint, hmono, htwoe, hthree, htot)
   endif
 
-end subroutine htilde_mu_mat_bi_ortho_tot
+end subroutine htilde_mu_mat_bi_ortho_tot_slow
 
 ! --
 
-subroutine htilde_mu_mat_bi_ortho(key_j, key_i, Nint, hmono, htwoe, hthree, htot)
+subroutine htilde_mu_mat_bi_ortho_slow(key_j, key_i, Nint, hmono, htwoe, hthree, htot)
 
   BEGIN_DOC
   !
@@ -61,22 +61,22 @@ subroutine htilde_mu_mat_bi_ortho(key_j, key_i, Nint, hmono, htwoe, hthree, htot
   if(degree.gt.2) return
 
   if(degree == 0)then
-    call diag_htilde_mu_mat_bi_ortho(Nint, key_i, hmono, htwoe, htot)
+    call diag_htilde_mu_mat_bi_ortho_slow(Nint, key_i, hmono, htwoe, htot)
   else if (degree == 1)then
-    call single_htilde_mu_mat_bi_ortho(Nint, key_j, key_i, hmono, htwoe, htot)
+    call single_htilde_mu_mat_bi_ortho_slow(Nint, key_j, key_i, hmono, htwoe, htot)
   else if(degree == 2)then
-    call double_htilde_mu_mat_bi_ortho(Nint, key_j, key_i, hmono, htwoe, htot)
+    call double_htilde_mu_mat_bi_ortho_slow(Nint, key_j, key_i, hmono, htwoe, htot)
   endif
 
   if(three_body_h_tc) then
     if(degree == 2) then
-      if(.not.double_normal_ord) then
+      if(.not.double_normal_ord.and.elec_num.gt.2.and.three_e_5_idx_term) then
-        call double_htilde_three_body_ints_bi_ort(Nint, key_j, key_i, hthree)
+        call double_htilde_three_body_ints_bi_ort_slow(Nint, key_j, key_i, hthree)
       endif
-    else if(degree == 1) then
+    else if(degree == 1.and.elec_num.gt.2.and.three_e_4_idx_term) then
-      call single_htilde_three_body_ints_bi_ort(Nint, key_j, key_i, hthree)
+      call single_htilde_three_body_ints_bi_ort_slow(Nint, key_j, key_i, hthree)
-    else if(degree == 0) then
+    else if(degree == 0.and.elec_num.gt.2.and.three_e_3_idx_term) then
-      call diag_htilde_three_body_ints_bi_ort(Nint, key_i, hthree)
+      call diag_htilde_three_body_ints_bi_ort_slow(Nint, key_i, hthree)
     endif
   endif
 
@@ -89,7 +89,7 @@ end
 
 ! ---
 
-subroutine diag_htilde_mu_mat_bi_ortho(Nint, key_i, hmono, htwoe, htot)
+subroutine diag_htilde_mu_mat_bi_ortho_slow(Nint, key_i, hmono, htwoe, htot)
 
   BEGIN_DOC
   !  diagonal element of htilde ONLY FOR ONE- AND TWO-BODY TERMS 
@@ -188,7 +188,7 @@ end
 
 
 
-subroutine double_htilde_mu_mat_bi_ortho(Nint, key_j, key_i, hmono, htwoe, htot)
+subroutine double_htilde_mu_mat_bi_ortho_slow(Nint, key_j, key_i, hmono, htwoe, htot)
 
   BEGIN_DOC
   ! <key_j | H_tilde | key_i> for double excitation  ONLY FOR ONE- AND TWO-BODY TERMS 
@@ -227,18 +227,7 @@ subroutine double_htilde_mu_mat_bi_ortho(Nint, key_j, key_i, hmono, htwoe, htot)
    return
   endif
 
-!  if(core_tc_op)then
-!   print*,'core_tc_op not already taken into account for bi ortho'
-!   print*,'stopping ...'
-!   stop
-!   do i = 1, Nint
-!    key_i_core(i,1) = xor(key_i(i,1),core_bitmask(i,1))
-!    key_i_core(i,2) = xor(key_i(i,2),core_bitmask(i,2))
-!   enddo
-!   call bitstring_to_list_ab(key_i_core, occ, Ne, Nint)
-!  else
    call bitstring_to_list_ab(key_i, occ, Ne, Nint)
-!  endif
   call get_double_excitation(key_i, key_j, exc, phase, Nint)
   call decode_exc(exc, 2, h1, p1, h2, p2, s1, s2)
 
@@ -246,7 +235,7 @@ subroutine double_htilde_mu_mat_bi_ortho(Nint, key_j, key_i, hmono, htwoe, htot)
    ! opposite spin two-body 
 !   key_j, key_i
     htwoe  = mo_bi_ortho_tc_two_e(p2,p1,h2,h1) 
-    if(double_normal_ord.and.+Ne(1).gt.2)then
+    if(three_body_h_tc.and.double_normal_ord.and.+Ne(1).gt.2)then
      htwoe += normal_two_body_bi_orth(p2,h2,p1,h1)!!! WTF ???
     endif
   else
@@ -255,7 +244,7 @@ subroutine double_htilde_mu_mat_bi_ortho(Nint, key_j, key_i, hmono, htwoe, htot)
    htwoe  = mo_bi_ortho_tc_two_e(p2,p1,h2,h1)  
    ! exchange terms 
    htwoe -= mo_bi_ortho_tc_two_e(p1,p2,h2,h1) 
-   if(double_normal_ord.and.+Ne(1).gt.2)then
+   if(three_body_h_tc.and.double_normal_ord.and.+Ne(1).gt.2)then
     htwoe -= normal_two_body_bi_orth(h2,p1,h1,p2)!!! WTF ???
     htwoe += normal_two_body_bi_orth(h1,p1,h2,p2)!!! WTF ???
    endif
@@ -266,7 +255,7 @@ subroutine double_htilde_mu_mat_bi_ortho(Nint, key_j, key_i, hmono, htwoe, htot)
 end
 
 
-subroutine single_htilde_mu_mat_bi_ortho(Nint, key_j, key_i, hmono, htwoe, htot)
+subroutine single_htilde_mu_mat_bi_ortho_slow(Nint, key_j, key_i, hmono, htwoe, htot)
 
   BEGIN_DOC
   ! <key_j | H_tilde | key_i> for single excitation ONLY FOR ONE- AND TWO-BODY TERMS 

src/tc_bi_ortho/symmetrized_3_e_int.irp.f

@@ -96,9 +96,11 @@ double precision function three_e_single_parrallel_spin(m,j,k,i)
  implicit none
  integer, intent(in) :: i,k,j,m
   three_e_single_parrallel_spin = three_e_4_idx_direct_bi_ort(m,j,k,i)  ! direct
-  three_e_single_parrallel_spin += three_e_4_idx_cycle_1_bi_ort(m,j,k,i) + three_e_4_idx_cycle_2_bi_ort(m,j,k,i) & ! two cyclic permutations 
+  three_e_single_parrallel_spin += three_e_4_idx_cycle_1_bi_ort(m,j,k,i) + three_e_4_idx_cycle_1_bi_ort(j,m,k,i) & ! two cyclic permutations 
   - three_e_4_idx_exch23_bi_ort(m,j,k,i) - three_e_4_idx_exch13_bi_ort(m,j,k,i)  & ! two first exchange 
-  - three_e_4_idx_exch12_bi_ort(m,j,k,i) ! last exchange 
+  - three_e_4_idx_exch13_bi_ort(j,m,k,i) ! last exchange 
+  ! TODO
+  ! use transpose
 end
 
 double precision function three_e_double_parrallel_spin(m,l,j,k,i)
@@ -107,5 +109,6 @@ double precision function three_e_double_parrallel_spin(m,l,j,k,i)
   three_e_double_parrallel_spin = three_e_5_idx_direct_bi_ort(m,l,j,k,i)  ! direct
   three_e_double_parrallel_spin += three_e_5_idx_cycle_1_bi_ort(m,l,j,k,i) + three_e_5_idx_cycle_2_bi_ort(m,l,j,k,i) & ! two cyclic permutations 
   - three_e_5_idx_exch23_bi_ort(m,l,j,k,i) - three_e_5_idx_exch13_bi_ort(m,l,j,k,i)  & ! two first exchange 
-  - three_e_5_idx_exch12_bi_ort(m,l,j,k,i) ! last exchange 
+!  - three_e_5_idx_exch12_bi_ort(m,l,j,k,i) ! last exchange 
+  - three_e_5_idx_direct_bi_ort(m,l,i,k,j) ! last exchange 
 end

src/tc_bi_ortho/tc_cisd_sc2_utils.irp.f

@@ -11,10 +11,10 @@
  allocate(H_jj(N_det),vec_tmp(N_det,n_states_diag),eigval_tmp(N_states))
  dressing_dets = 0.d0
  do i = 1, N_det
-  call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,i), psi_det(1,1,i), N_int, H_jj(i))
+  call htilde_mu_mat_bi_ortho_tot_slow(psi_det(1,1,i), psi_det(1,1,i), N_int, H_jj(i))
   call get_excitation_degree(HF_bitmask,psi_det(1,1,i),degree,N_int)
   if(degree == 1 .or. degree == 2)then
-   call htilde_mu_mat_bi_ortho(HF_bitmask,psi_det(1,1,i),N_int,hmono,htwoe,hthree,h0j(i))
+   call htilde_mu_mat_bi_ortho_slow(HF_bitmask,psi_det(1,1,i),N_int,hmono,htwoe,hthree,h0j(i))
   endif
  enddo
  reigvec_tc_bi_orth_tmp = 0.d0
@@ -29,7 +29,7 @@
   vec_tmp(istate,istate) = 1.d0
  enddo
  print*,'Diagonalizing the TC CISD '
- call davidson_general_diag_dressed_ext_rout_nonsym_b1space(vec_tmp, H_jj, dressing_dets,eigval_tmp, N_det, n_states, n_states_diag, converged, htc_bi_ortho_calc_tdav)
+ call davidson_general_diag_dressed_ext_rout_nonsym_b1space(vec_tmp, H_jj, dressing_dets,eigval_tmp, N_det, n_states, n_states_diag, converged, htc_bi_ortho_calc_tdav_slow)
  do i = 1, N_det 
   e_corr_dets(i) = reigvec_tc_bi_orth_tmp(i,1) * h0j(i)/reigvec_tc_bi_orth_tmp(1,1)
  enddo
@@ -41,8 +41,8 @@
  it = 0
  dressing_dets = 0.d0
   double precision, allocatable :: H_jj(:),vec_tmp(:,:),eigval_tmp(:)
-  external                         htc_bi_ortho_calc_tdav
+  external                         htc_bi_ortho_calc_tdav_slow
-  external                         htcdag_bi_ortho_calc_tdav
+  external                         htcdag_bi_ortho_calc_tdav_slow
   logical                       :: converged
  do while (dabs(E_before-E_current).gt.thr)
   it += 1
@@ -66,7 +66,7 @@
   do istate = N_states+1, n_states_diag
    vec_tmp(istate,istate) = 1.d0
   enddo
-  call davidson_general_diag_dressed_ext_rout_nonsym_b1space(vec_tmp, H_jj, dressing_dets,eigval_tmp, N_det, n_states, n_states_diag, converged, htc_bi_ortho_calc_tdav)
+  call davidson_general_diag_dressed_ext_rout_nonsym_b1space(vec_tmp, H_jj, dressing_dets,eigval_tmp, N_det, n_states, n_states_diag, converged, htc_bi_ortho_calc_tdav_slow)
   print*,'outside Davidson'
   print*,'eigval_tmp(1) = ',eigval_tmp(1)
   do i = 1, N_det 

src/tc_bi_ortho/tc_h_eigvectors.irp.f

@@ -207,8 +207,6 @@ end
   else ! n_det > N_det_max_full
 
     double precision, allocatable :: H_jj(:),vec_tmp(:,:)
-    external                         htc_bi_ortho_calc_tdav
-    external                         htcdag_bi_ortho_calc_tdav
     external                         H_tc_u_0_opt
     external                         H_tc_dagger_u_0_opt
     external                         H_tc_s2_dagger_u_0_opt
@@ -217,7 +215,7 @@ end
     allocate(H_jj(N_det),vec_tmp(N_det,n_states_diag))
 
     do i = 1, N_det
-      call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,i), psi_det(1,1,i), N_int, H_jj(i))
+      call htilde_mu_mat_opt_bi_ortho_tot(psi_det(1,1,i), psi_det(1,1,i), N_int, H_jj(i))
     enddo
 
     print*,'---------------------------------'
@@ -259,7 +257,6 @@ end
     do istate = N_states+1, n_states_diag
       vec_tmp(istate,istate) = 1.d0
     enddo
-    !call davidson_general_ext_rout_nonsym_b1space(vec_tmp, H_jj, eigval_right_tc_bi_orth, N_det, n_states, n_states_diag, converged, htc_bi_ortho_calc_tdav)
     !call davidson_general_ext_rout_nonsym_b1space(vec_tmp, H_jj, eigval_right_tc_bi_orth, N_det, n_states, n_states_diag, converged, H_tc_u_0_opt)
     converged = .False.
     i_it = 0

src/tc_bi_ortho/tc_hmat.irp.f

@@ -9,33 +9,25 @@
  
   implicit none
   integer          :: i, j
-  double precision :: hmono,htwoe,hthree,htot
+  double precision :: htot
 
   PROVIDE N_int
 
   i = 1
   j = 1
-  call htilde_mu_mat_bi_ortho(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
+  call htilde_mu_mat_opt_bi_ortho_tot(psi_det(1,1,j), psi_det(1,1,i), N_int, htot)
 
- !$OMP PARALLEL DO SCHEDULE(GUIDED) DEFAULT(NONE) PRIVATE(i,j,hmono, htwoe, hthree, htot) &
+ !$OMP PARALLEL DO SCHEDULE(GUIDED) DEFAULT(NONE) PRIVATE(i,j, htot) &
  !$OMP SHARED (N_det, psi_det, N_int,htilde_matrix_elmt_bi_ortho)
     do i = 1, N_det
       do j = 1, N_det
         ! < J | Htilde | I >
-        call htilde_mu_mat_bi_ortho(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
+        call htilde_mu_mat_opt_bi_ortho_tot(psi_det(1,1,j), psi_det(1,1,i), N_int, htot)
 
-        !print *, ' hmono  = ', hmono
-        !print *, ' htwoe  = ', htwoe
-        !print *, ' hthree = ', hthree
         htilde_matrix_elmt_bi_ortho(j,i) = htot
       enddo
     enddo
  !$OMP END PARALLEL DO
-! print*,'htilde_matrix_elmt_bi_ortho = '
-! do i = 1, min(100,N_det)
-!  write(*,'(100(F16.10,X))')htilde_matrix_elmt_bi_ortho(1:min(100,N_det),i)
-! enddo
-
 
 END_PROVIDER 
 

src/tc_bi_ortho/tc_som.irp.f

@@ -56,8 +56,8 @@ subroutine main()
   U_SOM = 0.d0 
   do i = 1, N_det
     if(i == i_HF) cycle
-    call htilde_mu_mat_bi_ortho(psi_det(1,1,i_HF), psi_det(1,1,i), N_int, hmono_1, htwoe_1, hthree_1, htot_1)
+    call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,i_HF), psi_det(1,1,i), N_int, hmono_1, htwoe_1, hthree_1, htot_1)
-    call htilde_mu_mat_bi_ortho(psi_det(1,1,i), psi_det(1,1,i_HF), N_int, hmono_2, htwoe_2, hthree_2, htot_2)
+    call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,i), psi_det(1,1,i_HF), N_int, hmono_2, htwoe_2, hthree_2, htot_2)
     U_SOM += htot_1 * htot_2
   enddo
   U_SOM = 0.5d0 * U_SOM

src/tc_bi_ortho/tc_utils.irp.f

@@ -12,7 +12,7 @@ subroutine write_tc_energy()
     do i = 1, N_det
       do j = 1, N_det
         !htot = htilde_matrix_elmt_bi_ortho(i,j)
-        call htilde_mu_mat_bi_ortho(psi_det(1,1,i), psi_det(1,1,j), N_int, hmono, htwoe, hthree, htot)
+        call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,i), psi_det(1,1,j), N_int, hmono, htwoe, hthree, htot)
         E_TC = E_TC + psi_l_coef_bi_ortho(i,k) * psi_r_coef_bi_ortho(j,k) * htot
         !E_TC = E_TC + leigvec_tc_bi_orth(i,k) * reigvec_tc_bi_orth(j,k) * htot
       enddo
@@ -38,15 +38,16 @@ subroutine write_tc_var()
 
   implicit none
   integer          :: i, j, k
-  double precision :: hmono, htwoe, hthree, htot
+  double precision :: hmono, htwoe, hthree, htot_1j, htot_j1
   double precision :: SIGMA_TC
 
   do k = 1, n_states
 
     SIGMA_TC = 0.d0
     do j = 2, N_det
-      call htilde_mu_mat_bi_ortho(psi_det(1,1,1), psi_det(1,1,j), N_int, hmono, htwoe, hthree, htot)
+      call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,1), psi_det(1,1,j), N_int, hmono, htwoe, hthree, htot_1j)
-      SIGMA_TC = SIGMA_TC + htot * htot
+      call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,j), psi_det(1,1,1), N_int, hmono, htwoe, hthree, htot_j1)
+      SIGMA_TC = SIGMA_TC + htot_1j * htot_j1
     enddo
 
     print *, " state    : ", k

src/tc_bi_ortho/test_normal_order.irp.f

@@ -35,7 +35,7 @@ subroutine test
      det_i = ref_bitmask
      call do_single_excitation(det_i,h1,p1,s1,i_ok)
      call do_single_excitation(det_i,h2,p2,s2,i_ok)
-     call htilde_mu_mat_bi_ortho(det_i,HF_bitmask,N_int,hmono,htwoe,hthree,htilde_ij)
+     call htilde_mu_mat_bi_ortho_slow(det_i,HF_bitmask,N_int,hmono,htwoe,hthree,htilde_ij)
      call get_excitation_degree(ref_bitmask,det_i,degree,N_int)
      call get_excitation(ref_bitmask,det_i,exc,degree,phase,N_int)
      hthree *= phase
@@ -67,7 +67,7 @@ do h1 = 1, elec_alpha_num
     if(i_ok.ne.1)cycle
     call do_single_excitation(det_i,h2,p2,s2,i_ok)
     if(i_ok.ne.1)cycle
-    call htilde_mu_mat_bi_ortho(det_i,ref_bitmask,N_int,hmono,htwoe,hthree,htilde_ij)
+    call htilde_mu_mat_bi_ortho_slow(det_i,ref_bitmask,N_int,hmono,htwoe,hthree,htilde_ij)
     call get_excitation_degree(ref_bitmask,det_i,degree,N_int)
     call get_excitation(ref_bitmask,det_i,exc,degree,phase,N_int)
     integer :: hh1, pp1, hh2, pp2, ss1, ss2
@@ -103,7 +103,7 @@ do h1 = 1, elec_beta_num
     if(i_ok.ne.1)cycle
     call do_single_excitation(det_i,h2,p2,s2,i_ok)
     if(i_ok.ne.1)cycle
-    call htilde_mu_mat_bi_ortho(det_i,ref_bitmask,N_int,hmono,htwoe,hthree,htilde_ij)
+    call htilde_mu_mat_bi_ortho_slow(det_i,ref_bitmask,N_int,hmono,htwoe,hthree,htilde_ij)
     call get_excitation_degree(ref_bitmask,det_i,degree,N_int)
     call get_excitation(ref_bitmask,det_i,exc,degree,phase,N_int)
     call decode_exc(exc, 2, hh1, pp1, hh2, pp2, ss1, ss2)

src/tc_bi_ortho/test_s2_tc.irp.f

@@ -91,7 +91,7 @@ subroutine routine_test_s2_davidson
  external H_tc_s2_u_0_opt
  allocate(H_jj(N_det),vec_tmp(N_det,n_states_diag),energies(n_states_diag), s2(n_states_diag))
  do i = 1, N_det
-   call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,i), psi_det(1,1,i), N_int, H_jj(i))
+   call htilde_mu_mat_bi_ortho_tot_slow(psi_det(1,1,i), psi_det(1,1,i), N_int, H_jj(i))
  enddo
  ! Preparing the left-eigenvector
  print*,'Computing the left-eigenvector '

src/tc_bi_ortho/test_tc_bi_ortho.irp.f

@@ -11,12 +11,17 @@ program tc_bi_ortho
   touch read_wf
   touch  my_grid_becke my_n_pt_r_grid my_n_pt_a_grid
 
- call test_h_u0
+! call test_h_u0
 ! call test_slater_tc_opt
 ! call timing_tot
 ! call timing_diag
 ! call timing_single
 ! call timing_double
+
+  call test_no()
+  !call test_no_aba()
+  !call test_no_aab()
+  !call test_no_aaa()
 end
 
 subroutine test_h_u0
@@ -31,7 +36,7 @@ subroutine test_h_u0
   u_0(i) = psi_r_coef_bi_ortho(i,1)
  enddo
  call H_tc_u_0_nstates_openmp(v_0_new,u_0,N_states,N_det, do_right)
- call htc_bi_ortho_calc_tdav (v_0_ref,u_0,N_states,N_det)
+ call htc_bi_ortho_calc_tdav_slow (v_0_ref,u_0,N_states,N_det)
  print*,'difference right '
  accu = 0.d0
  do i = 1, N_det
@@ -42,7 +47,7 @@ subroutine test_h_u0
  do_right = .False.
  v_0_new = 0.d0
  call H_tc_u_0_nstates_openmp(v_0_new,u_0,N_states,N_det, do_right)
- call htcdag_bi_ortho_calc_tdav(v_0_ref_dagger,u_0,N_states,N_det, do_right)
+ call htcdag_bi_ortho_calc_tdav_slow(v_0_ref_dagger,u_0,N_states,N_det, do_right)
  print*,'difference left'
  accu = 0.d0
  do i = 1, N_det
@@ -63,7 +68,7 @@ subroutine test_slater_tc_opt
  i_count = 0.d0
  do i = 1, N_det
   do j = 1,N_det
-   call htilde_mu_mat_bi_ortho(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
+   call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
    call htilde_mu_mat_opt_bi_ortho(psi_det(1,1,j), psi_det(1,1,i), N_int, hnewmono, hnewtwoe, hnewthree, hnewtot)
    if(dabs(htot).gt.1.d-15)then
      i_count += 1.D0
@@ -99,7 +104,7 @@ subroutine timing_tot
   do j = 1, N_det
 !   call get_excitation_degree(psi_det(1,1,j), psi_det(1,1,i),degree,N_int)
    i_count += 1.d0
-   call htilde_mu_mat_bi_ortho(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
+   call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
   enddo
  enddo
  call wall_time(wall1)
@@ -146,7 +151,7 @@ subroutine timing_diag
  do i = 1, N_det
   do j = i,i 
    i_count += 1.d0
-   call htilde_mu_mat_bi_ortho(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
+   call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
   enddo
  enddo
  call wall_time(wall1)
@@ -183,7 +188,7 @@ subroutine timing_single
    if(degree.ne.1)cycle
    i_count += 1.d0
    call wall_time(wall0)
-   call htilde_mu_mat_bi_ortho(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
+   call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
    call wall_time(wall1)
    accu += wall1 - wall0
   enddo
@@ -225,7 +230,7 @@ subroutine timing_double
    if(degree.ne.2)cycle
    i_count += 1.d0
    call wall_time(wall0)
-   call htilde_mu_mat_bi_ortho(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
+   call htilde_mu_mat_bi_ortho_slow(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
    call wall_time(wall1)
    accu += wall1 - wall0
   enddo
@@ -252,3 +257,169 @@ subroutine timing_double
 
 end
 
+! ---
+
+subroutine test_no()
+
+  implicit none
+  integer          :: i, j, k, l
+  double precision :: accu, contrib, new, ref, thr
+
+  print*, ' testing normal_two_body_bi_orth ...'
+
+  thr = 1d-8
+
+  PROVIDE normal_two_body_bi_orth_old
+  PROVIDE normal_two_body_bi_orth
+
+  accu = 0.d0
+  do i = 1, mo_num
+    do j = 1, mo_num
+      do k = 1, mo_num
+        do l = 1, mo_num
+
+          new = normal_two_body_bi_orth    (l,k,j,i)
+          ref = normal_two_body_bi_orth_old(l,k,j,i)
+          contrib = dabs(new - ref)
+          accu += contrib
+          if(contrib .gt. thr) then
+            print*, ' problem on normal_two_body_bi_orth'
+            print*, l, k, j, i
+            print*, ref, new, contrib
+            stop
+          endif
+
+        enddo
+      enddo
+    enddo
+  enddo
+  print*, ' accu on normal_two_body_bi_orth = ', accu / dble(mo_num)**4
+
+ return
+end
+
+! ---
+
+subroutine test_no_aba()
+
+  implicit none
+  integer          :: i, j, k, l
+  double precision :: accu, contrib, new, ref, thr
+
+  print*, ' testing no_aba_contraction ...'
+
+  thr = 1d-8
+
+  PROVIDE no_aba_contraction_v0
+  PROVIDE no_aba_contraction
+
+  accu = 0.d0
+  do i = 1, mo_num
+    do j = 1, mo_num
+      do k = 1, mo_num
+        do l = 1, mo_num
+
+          new = no_aba_contraction   (l,k,j,i)
+          ref = no_aba_contraction_v0(l,k,j,i)
+          contrib = dabs(new - ref)
+          accu += contrib
+          if(contrib .gt. thr) then
+            print*, ' problem on no_aba_contraction'
+            print*, l, k, j, i
+            print*, ref, new, contrib
+            stop
+          endif
+
+        enddo
+      enddo
+    enddo
+  enddo
+  print*, ' accu on no_aba_contraction = ', accu / dble(mo_num)**4
+
+ return
+end
+
+! ---
+
+
+subroutine test_no_aab()
+
+  implicit none
+  integer          :: i, j, k, l
+  double precision :: accu, contrib, new, ref, thr
+
+  print*, ' testing no_aab_contraction ...'
+
+  thr = 1d-8
+
+  PROVIDE no_aab_contraction_v0
+  PROVIDE no_aab_contraction
+
+  accu = 0.d0
+  do i = 1, mo_num
+    do j = 1, mo_num
+      do k = 1, mo_num
+        do l = 1, mo_num
+
+          new = no_aab_contraction   (l,k,j,i)
+          ref = no_aab_contraction_v0(l,k,j,i)
+          contrib = dabs(new - ref)
+          accu += contrib
+          if(contrib .gt. thr) then
+            print*, ' problem on no_aab_contraction'
+            print*, l, k, j, i
+            print*, ref, new, contrib
+            stop
+          endif
+
+        enddo
+      enddo
+    enddo
+  enddo
+  print*, ' accu on no_aab_contraction = ', accu / dble(mo_num)**4
+
+ return
+end
+
+! ---
+
+subroutine test_no_aaa()
+
+  implicit none
+  integer          :: i, j, k, l
+  double precision :: accu, contrib, new, ref, thr
+
+  print*, ' testing no_aaa_contraction ...'
+
+  thr = 1d-8
+
+  PROVIDE no_aaa_contraction_v0
+  PROVIDE no_aaa_contraction
+
+  accu = 0.d0
+  do i = 1, mo_num
+    do j = 1, mo_num
+      do k = 1, mo_num
+        do l = 1, mo_num
+
+          new = no_aaa_contraction   (l,k,j,i)
+          ref = no_aaa_contraction_v0(l,k,j,i)
+          contrib = dabs(new - ref)
+          accu += contrib
+          if(contrib .gt. thr) then
+            print*, ' problem on no_aaa_contraction'
+            print*, l, k, j, i
+            print*, ref, new, contrib
+            stop
+          endif
+
+        enddo
+      enddo
+    enddo
+  enddo
+  print*, ' accu on no_aaa_contraction = ', accu / dble(mo_num)**4
+
+ return
+end
+
+! ---

src/tc_bi_ortho/test_tc_fock.irp.f

@@ -25,8 +25,7 @@ subroutine test_3e
  implicit none
  double precision :: integral_aaa,integral_aab,integral_abb,integral_bbb,accu
  double precision ::  hmono, htwoe, hthree, htot
- call htilde_mu_mat_bi_ortho(ref_bitmask, ref_bitmask, N_int, hmono, htwoe, hthree, htot)
+ call htilde_mu_mat_bi_ortho_slow(ref_bitmask, ref_bitmask, N_int, hmono, htwoe, hthree, htot)
-! call diag_htilde_three_body_ints_bi_ort(N_int, ref_bitmask, hthree)
  print*,'hmono = ',hmono
  print*,'htwoe = ',htwoe
  print*,'hthree= ',hthree
@@ -88,7 +87,7 @@ subroutine routine_3()
        print*, ' excited det'
        call debug_det(det_i, N_int)
  
-       call htilde_mu_mat_bi_ortho(det_i, ref_bitmask, N_int, hmono, htwoe, hthree, htilde_ij)
+       call htilde_mu_mat_bi_ortho_slow(det_i, ref_bitmask, N_int, hmono, htwoe, hthree, htilde_ij)
        if(dabs(hthree).lt.1.d-10)cycle
        ref = hthree 
        if(s1 == 1)then
@@ -156,7 +155,7 @@ subroutine routine_tot()
         stop
        endif
  
-       call htilde_mu_mat_bi_ortho(det_i, ref_bitmask, N_int, hmono, htwoe, hthree, htilde_ij)
+       call htilde_mu_mat_bi_ortho_slow(det_i, ref_bitmask, N_int, hmono, htwoe, hthree, htilde_ij)
        print*,htilde_ij
        if(dabs(htilde_ij).lt.1.d-10)cycle
        print*, ' excited det'

src/tc_keywords/EZFIO.cfg

@@ -16,6 +16,24 @@ doc: If |true|, three-body terms are included
 interface: ezfio,provider,ocaml
 default: True
 
+[three_e_3_idx_term]
+type: logical
+doc: If |true|, the diagonal 3-idx terms of the 3-e interaction are taken
+interface: ezfio,provider,ocaml
+default: True
+
+[three_e_4_idx_term]
+type: logical
+doc: If |true|, the off-diagonal 4-idx terms of the 3-e interaction are taken
+interface: ezfio,provider,ocaml
+default: True
+
+[three_e_5_idx_term]
+type: logical
+doc: If |true|, the off-diagonal 5-idx terms of the 3-e interaction are taken
+interface: ezfio,provider,ocaml
+default: True
+
 [pure_three_body_h_tc]
 type: logical
 doc: If |true|, pure triple excitation three-body terms are included
@@ -112,6 +130,12 @@ doc: exponents of the 1-body Jastrow
 interface: ezfio
 size: (nuclei.nucl_num)
 
+[j1b_pen_coef]
+type: double precision
+doc: coefficients of the 1-body Jastrow
+interface: ezfio
+size: (nuclei.nucl_num)
+
 [j1b_coeff]
 type: double precision
 doc: coeff of the 1-body Jastrow
@@ -130,6 +154,12 @@ doc: a parameter used to define mu(r)
 interface: ezfio, provider, ocaml
 default: 6.203504908994001e-1
 
+[beta_rho_power]
+type: double precision
+doc: a parameter used to define mu(r)
+interface: ezfio, provider, ocaml
+default: 0.5
+
 [thr_degen_tc]
 type: Threshold
 doc: Threshold to determine if two orbitals are degenerate in TCSCF in order to avoid random quasi orthogonality between the right- and left-eigenvector for the same eigenvalue

src/tc_keywords/j1b_pen.irp.f

@@ -1,17 +1,22 @@
 
 ! ---
 
-BEGIN_PROVIDER [ double precision, j1b_pen, (nucl_num) ]
+ BEGIN_PROVIDER [ double precision, j1b_pen     , (nucl_num) ]
+&BEGIN_PROVIDER [ double precision, j1b_pen_coef, (nucl_num) ]
 
   BEGIN_DOC
-  ! exponents of the 1-body Jastrow
+  ! parameters of the 1-body Jastrow
   END_DOC
 
   implicit none
   logical :: exists
+  integer :: i
+  integer :: ierr
 
   PROVIDE ezfio_filename
 
+  ! ---
+
   if (mpi_master) then
     call ezfio_has_tc_keywords_j1b_pen(exists)
   endif
@@ -23,7 +28,6 @@ BEGIN_PROVIDER [ double precision, j1b_pen, (nucl_num) ]
 
   IRP_IF MPI
     include 'mpif.h'
-    integer :: ierr
     call MPI_BCAST(j1b_pen, (nucl_num), MPI_DOUBLE_PRECISION, 0, MPI_COMM_WORLD, ierr)
     if (ierr /= MPI_SUCCESS) then
       stop 'Unable to read j1b_pen with MPI'
@@ -31,7 +35,6 @@ BEGIN_PROVIDER [ double precision, j1b_pen, (nucl_num) ]
   IRP_ENDIF
 
   if (exists) then
-
     if (mpi_master) then
       write(6,'(A)') '.. >>>>> [ IO READ: j1b_pen ] <<<<< ..'
       call ezfio_get_tc_keywords_j1b_pen(j1b_pen)
@@ -42,19 +45,54 @@ BEGIN_PROVIDER [ double precision, j1b_pen, (nucl_num) ]
         endif
       IRP_ENDIF
     endif
-
   else
- 
-    integer :: i
     do i = 1, nucl_num
       j1b_pen(i) = 1d5
     enddo
-
   endif
- print*,'parameters for nuclei jastrow'
+
- do i = 1, nucl_num
+  ! ---
-  print*,'i,Z,j1b_pen(i)',i,nucl_charge(i),j1b_pen(i)
+
- enddo
+  if (mpi_master) then
+    call ezfio_has_tc_keywords_j1b_pen_coef(exists)
+  endif
+
+  IRP_IF MPI_DEBUG
+    print *,  irp_here, mpi_rank
+    call MPI_BARRIER(MPI_COMM_WORLD, ierr)
+  IRP_ENDIF
+
+  IRP_IF MPI
+    call MPI_BCAST(j1b_pen_coef, (nucl_num), MPI_DOUBLE_PRECISION, 0, MPI_COMM_WORLD, ierr)
+    if (ierr /= MPI_SUCCESS) then
+      stop 'Unable to read j1b_pen_coef with MPI'
+    endif
+  IRP_ENDIF
+
+  if (exists) then
+    if (mpi_master) then
+      write(6,'(A)') '.. >>>>> [ IO READ: j1b_pen_coef ] <<<<< ..'
+      call ezfio_get_tc_keywords_j1b_pen_coef(j1b_pen_coef)
+      IRP_IF MPI
+        call MPI_BCAST(j1b_pen_coef, (nucl_num), MPI_DOUBLE_PRECISION, 0, MPI_COMM_WORLD, ierr)
+        if (ierr /= MPI_SUCCESS) then
+          stop 'Unable to read j1b_pen_coef with MPI'
+        endif
+      IRP_ENDIF
+    endif
+  else
+    do i = 1, nucl_num
+      j1b_pen_coef(i) = 1d0
+    enddo
+  endif
+
+  ! ---
+
+  print *, ' parameters for nuclei jastrow'
+  print *, ' i, Z, j1b_pen, j1b_pen_coef'
+  do i = 1, nucl_num
+    print *, i, nucl_charge(i), j1b_pen(i), j1b_pen_coef(i)
+  enddo
 
 END_PROVIDER
 
@@ -114,3 +152,4 @@ BEGIN_PROVIDER [ double precision, j1b_coeff, (nucl_num) ]
 END_PROVIDER
 
 ! ---
+

src/tc_scf/rh_tcscf_diis.irp.f

@@ -11,6 +11,7 @@ subroutine rh_tcscf_diis()
 
   integer                       :: i, j, it
   integer                       :: dim_DIIS, index_dim_DIIS
+  logical                       :: converged
   double precision              :: etc_tot, etc_1e, etc_2e, etc_3e, e_save, e_delta
   double precision              :: tc_grad, g_save, g_delta, g_delta_th
   double precision              :: level_shift_save, rate_th
@@ -92,8 +93,9 @@ subroutine rh_tcscf_diis()
 
   PROVIDE FQS_SQF_ao Fock_matrix_tc_ao_tot
 
+  converged = .false.
   !do while((tc_grad .gt. dsqrt(thresh_tcscf)) .and. (er_DIIS .gt. dsqrt(thresh_tcscf)))
-  do while(er_DIIS .gt. dsqrt(thresh_tcscf))
+  do while(.not. converged)
 
     call wall_time(t0)
 
@@ -218,21 +220,56 @@ subroutine rh_tcscf_diis()
     !g_delta_th  = dabs(tc_grad) ! g_delta)
     er_delta_th = dabs(er_DIIS) !er_delta)
 
+    converged = er_DIIS .lt. dsqrt(thresh_tcscf)
+
     call wall_time(t1)
     !write(6, '(I4,1X, F16.10,1X, F16.10,1X, F16.10,1X, F16.10,1X, F16.10,1X, F16.10,1X, F16.10,1X, F16.10,1X, I4,1X, F8.2)')  &
     !  it, etc_tot, etc_1e, etc_2e, etc_3e, e_delta, tc_grad, er_DIIS, level_shift_tcscf, dim_DIIS, (t1-t0)/60.d0
     write(6, '(I4,1X, F16.10,1X, F16.10,1X, F16.10,1X, F16.10,1X, F16.10,1X, F16.10,1X, F16.10,1X, I4,1X, F8.2)')  &
       it, etc_tot, etc_1e, etc_2e, etc_3e, e_delta, er_DIIS, level_shift_tcscf, dim_DIIS, (t1-t0)/60.d0
 
+
+!   Write data in JSON file
+
+    call lock_io
+    if (it == 1) then
+      write(json_unit, json_dict_uopen_fmt)
+    else
+      write(json_unit, json_dict_close_uopen_fmt)
+    endif
+    write(json_unit, json_int_fmt)  ' iteration      ', it
+    write(json_unit, json_real_fmt) ' SCF TC Energy  ', etc_tot
+    write(json_unit, json_real_fmt) ' E(1e)          ', etc_1e
+    write(json_unit, json_real_fmt) ' E(2e)          ', etc_2e
+    write(json_unit, json_real_fmt) ' E(3e)          ', etc_3e
+    write(json_unit, json_real_fmt) ' delta Energy   ', e_delta
+    write(json_unit, json_real_fmt) ' DIIS error     ', er_DIIS
+    write(json_unit, json_real_fmt) ' level_shift    ', level_shift_tcscf
+    write(json_unit, json_real_fmt) ' DIIS           ', dim_DIIS
+    write(json_unit, json_real_fmt) ' Wall time (min)', (t1-t0)/60.d0
+    call unlock_io
+
     if(er_delta .lt. 0.d0) then
       call ezfio_set_tc_scf_bitc_energy(etc_tot)
       call ezfio_set_bi_ortho_mos_mo_l_coef(mo_l_coef)
       call ezfio_set_bi_ortho_mos_mo_r_coef(mo_r_coef)
+      write(json_unit, json_true_fmt) 'saved'
+    else
+      write(json_unit, json_false_fmt) 'saved'
     endif
+    call lock_io
 
+    if (converged) then
+      write(json_unit, json_true_fmtx) 'converged'
+    else
+      write(json_unit, json_false_fmtx) 'converged'
+    endif
+    call unlock_io
     if(qp_stop()) exit
   enddo
 
+  write(json_unit, json_dict_close_fmtx)
+
   ! ---
 
   print *, ' TCSCF DIIS converged !'

src/tc_scf/tc_scf.irp.f

@@ -8,6 +8,8 @@ program tc_scf
 
   implicit none
 
+  write(json_unit,json_array_open_fmt) 'tc-scf'
+
   print *, ' starting ...'
 
   my_grid_becke  = .True.
@@ -57,6 +59,8 @@ program tc_scf
 
   endif
 
+  write(json_unit,json_array_close_fmtx)
+  call json_close
 
 end
 

src/tools/print_wf.irp.f

@@ -47,6 +47,7 @@ subroutine routine
  do i = 1, min(N_det_print_wf,N_det)
   print*,''
   print*,'i = ',i
+  print *,psi_det_sorted(1,1,i)
   call debug_det(psi_det_sorted(1,1,i),N_int)
   call get_excitation_degree(psi_det_sorted(1,1,i),psi_det_sorted(1,1,1),degree,N_int)
   print*,'degree = ',degree

src/trexio/EZFIO.cfg

@@ -10,11 +10,17 @@ doc: Name of the exported TREXIO file
 interface: ezfio, ocaml, provider
 default: None
 
-[export_rdm]
+[export_basis]
 type: logical
-doc: If True, export two-body reduced density matrix
+doc: If True, export basis set and AOs
 interface: ezfio, ocaml, provider
-default: False
+default: True
+
+[export_mos]
+type: logical
+doc: If True, export basis set and AOs
+interface: ezfio, ocaml, provider
+default: True
 
 [export_ao_one_e_ints]
 type: logical
@@ -22,12 +28,6 @@ doc: If True, export one-electron integrals in AO basis
 interface: ezfio, ocaml, provider
 default: False
 
-[export_mo_one_e_ints]
-type: logical
-doc: If True, export one-electron integrals in MO basis
-interface: ezfio, ocaml, provider
-default: False
-
 [export_ao_two_e_ints]
 type: logical
 doc: If True, export two-electron integrals in AO basis
@@ -40,6 +40,12 @@ doc: If True, export Cholesky-decomposed two-electron integrals in AO basis
 interface: ezfio, ocaml, provider
 default: False
 
+[export_mo_one_e_ints]
+type: logical
+doc: If True, export one-electron integrals in MO basis
+interface: ezfio, ocaml, provider
+default: False
+
 [export_mo_two_e_ints]
 type: logical
 doc: If True, export two-electron integrals in MO basis
@@ -52,3 +58,9 @@ doc: If True, export Cholesky-decomposed two-electron integrals in MO basis
 interface: ezfio, ocaml, provider
 default: False
 
+[export_rdm]
+type: logical
+doc: If True, export two-body reduced density matrix
+interface: ezfio, ocaml, provider
+default: False
+

src/trexio/export_trexio.irp.f

@@ -2,6 +2,6 @@ program export_trexio_prog
   implicit none
   read_wf = .True.
   SOFT_TOUCH read_wf
-  call export_trexio
+  call export_trexio(.False.)
 end
 

src/trexio/export_trexio_routines.irp.f

@@ -1,15 +1,17 @@
-subroutine export_trexio
+subroutine export_trexio(update)
   use trexio
   implicit none
   BEGIN_DOC
   !     Exports the wave function in TREXIO format
   END_DOC
 
+  logical, intent(in)            :: update
   integer(trexio_t)              :: f(N_states) ! TREXIO file handle
   integer(trexio_exit_code)      :: rc
   integer                        :: k
   double precision, allocatable  :: factor(:)
   character*(256)  :: filenames(N_states)
+  character :: rw
 
   filenames(1) = trexio_filename
   do k=2,N_states
@@ -18,15 +20,26 @@ subroutine export_trexio
 
   do k=1,N_states
     print *, 'TREXIO file : ', trim(filenames(k))
-    call system('test -f '//trim(filenames(k))//' && mv '//trim(filenames(k))//' '//trim(filenames(k))//'.bak')
+    if (update) then
+      call system('test -f '//trim(filenames(k))//' && cp -r '//trim(filenames(k))//' '//trim(filenames(k))//'.bak')
+    else
+      call system('test -f '//trim(filenames(k))//' && mv '//trim(filenames(k))//' '//trim(filenames(k))//'.bak')
+    endif
   enddo
   print *, ''
 
+  if (update) then
+     rw = 'u'
+  else
+     rw = 'w'
+  endif
+
+
   do k=1,N_states
     if (backend == 0) then
-      f(k) = trexio_open(filenames(k), 'u', TREXIO_HDF5, rc)
+      f(k) = trexio_open(filenames(k), rw, TREXIO_HDF5, rc)
     else if (backend == 1) then
-      f(k) = trexio_open(filenames(k), 'u', TREXIO_TEXT, rc)
+      f(k) = trexio_open(filenames(k), rw, TREXIO_TEXT, rc)
     endif
     if (f(k) == 0_8) then
       print *, 'Unable to open TREXIO file for writing'
@@ -171,92 +184,95 @@ subroutine export_trexio
   endif
 
 
+  if (export_basis) then
+
 ! Basis
 ! -----
 
-  print *, 'Basis'
+    print *, 'Basis'
 
+    rc = trexio_write_basis_type(f(1), 'Gaussian', len('Gaussian'))
+    call trexio_assert(rc, TREXIO_SUCCESS)
 
-  rc = trexio_write_basis_type(f(1), 'Gaussian', len('Gaussian'))
+    rc = trexio_write_basis_prim_num(f(1), prim_num)
-  call trexio_assert(rc, TREXIO_SUCCESS)
+    call trexio_assert(rc, TREXIO_SUCCESS)
 
-  rc = trexio_write_basis_prim_num(f(1), prim_num)
+     rc = trexio_write_basis_shell_num(f(1), shell_num)
-  call trexio_assert(rc, TREXIO_SUCCESS)
+     call trexio_assert(rc, TREXIO_SUCCESS)
 
-   rc = trexio_write_basis_shell_num(f(1), shell_num)
+     rc = trexio_write_basis_nucleus_index(f(1), basis_nucleus_index)
-   call trexio_assert(rc, TREXIO_SUCCESS)
+     call trexio_assert(rc, TREXIO_SUCCESS)
 
-   rc = trexio_write_basis_nucleus_index(f(1), basis_nucleus_index)
+     rc = trexio_write_basis_shell_ang_mom(f(1), shell_ang_mom)
-   call trexio_assert(rc, TREXIO_SUCCESS)
+     call trexio_assert(rc, TREXIO_SUCCESS)
 
-   rc = trexio_write_basis_shell_ang_mom(f(1), shell_ang_mom)
+     allocate(factor(shell_num))
-   call trexio_assert(rc, TREXIO_SUCCESS)
+!     if (ao_normalized) then
+!       factor(1:shell_num) = shell_normalization_factor(1:shell_num)
+!     else
+       factor(1:shell_num) = 1.d0
+!     endif
+     rc = trexio_write_basis_shell_factor(f(1), factor)
+     call trexio_assert(rc, TREXIO_SUCCESS)
 
-   allocate(factor(shell_num))
+     deallocate(factor)
-   if (ao_normalized) then
-     factor(1:shell_num) = shell_normalization_factor(1:shell_num)
-   else
-     factor(1:shell_num) = 1.d0
-   endif
-   rc = trexio_write_basis_shell_factor(f(1), factor)
-   call trexio_assert(rc, TREXIO_SUCCESS)
 
-   deallocate(factor)
+    rc = trexio_write_basis_shell_index(f(1), shell_index)
+    call trexio_assert(rc, TREXIO_SUCCESS)
 
-  rc = trexio_write_basis_shell_index(f(1), shell_index)
+    rc = trexio_write_basis_exponent(f(1), prim_expo)
-  call trexio_assert(rc, TREXIO_SUCCESS)
+    call trexio_assert(rc, TREXIO_SUCCESS)
 
-  rc = trexio_write_basis_exponent(f(1), prim_expo)
+    rc = trexio_write_basis_coefficient(f(1), prim_coef)
-  call trexio_assert(rc, TREXIO_SUCCESS)
+    call trexio_assert(rc, TREXIO_SUCCESS)
 
-  rc = trexio_write_basis_coefficient(f(1), prim_coef)
+    allocate(factor(prim_num))
-  call trexio_assert(rc, TREXIO_SUCCESS)
+    if (primitives_normalized) then
-
+      factor(1:prim_num) = prim_normalization_factor(1:prim_num)
-  allocate(factor(prim_num))
+    else
-  if (primitives_normalized) then
+      factor(1:prim_num) = 1.d0
-    factor(1:prim_num) = prim_normalization_factor(1:prim_num)
+    endif
-  else
+    rc = trexio_write_basis_prim_factor(f(1), factor)
-    factor(1:prim_num) = 1.d0
+    call trexio_assert(rc, TREXIO_SUCCESS)
-  endif
+    deallocate(factor)
-  rc = trexio_write_basis_prim_factor(f(1), factor)
-  call trexio_assert(rc, TREXIO_SUCCESS)
-  deallocate(factor)
 
 
 ! Atomic orbitals
 ! ---------------
 
-  print *, 'AOs'
+    print *, 'AOs'
 
-  rc = trexio_write_ao_num(f(1), ao_num)
+    rc = trexio_write_ao_num(f(1), ao_num)
-  call trexio_assert(rc, TREXIO_SUCCESS)
+    call trexio_assert(rc, TREXIO_SUCCESS)
 
-  rc = trexio_write_ao_cartesian(f(1), 1)
+    rc = trexio_write_ao_cartesian(f(1), 1)
-  call trexio_assert(rc, TREXIO_SUCCESS)
+    call trexio_assert(rc, TREXIO_SUCCESS)
 
-  rc = trexio_write_ao_shell(f(1), ao_shell)
+    rc = trexio_write_ao_shell(f(1), ao_shell)
-  call trexio_assert(rc, TREXIO_SUCCESS)
+    call trexio_assert(rc, TREXIO_SUCCESS)
 
-  integer :: i, pow0(3), powA(3), j, l, nz
+    integer :: i, pow0(3), powA(3), j, l, nz
-  double precision :: normA, norm0, C_A(3), overlap_x, overlap_z, overlap_y, c
+    double precision :: normA, norm0, C_A(3), overlap_x, overlap_z, overlap_y, c
-  nz=100
+    nz=100
 
-  C_A(1) = 0.d0
+    C_A(1) = 0.d0
-  C_A(2) = 0.d0
+    C_A(2) = 0.d0
-  C_A(3) = 0.d0
+    C_A(3) = 0.d0
+
+    allocate(factor(ao_num))
+    if (ao_normalized) then
+      do i=1,ao_num
+        l = ao_first_of_shell(ao_shell(i))
+        factor(i) = (ao_coef_normalized(i,1)+tiny(1.d0))/(ao_coef_normalized(l,1)+tiny(1.d0))
+      enddo
+    else
+      factor(:) = 1.d0
+    endif
+    rc = trexio_write_ao_normalization(f(1), factor)
+    call trexio_assert(rc, TREXIO_SUCCESS)
+    deallocate(factor)
 
-  allocate(factor(ao_num))
-  if (ao_normalized) then
-    do i=1,ao_num
-      l = ao_first_of_shell(ao_shell(i))
-      factor(i) = (ao_coef_normalized(i,1)+tiny(1.d0))/(ao_coef_normalized(l,1)+tiny(1.d0))
-    enddo
-  else
-    factor(:) = 1.d0
   endif
-  rc = trexio_write_ao_normalization(f(1), factor)
-  call trexio_assert(rc, TREXIO_SUCCESS)
-  deallocate(factor)
 
 ! One-e AO integrals
 ! ------------------
@@ -375,28 +391,30 @@ subroutine export_trexio
 ! Molecular orbitals
 ! ------------------
 
-  print *, 'MOs'
+  if (export_mos) then
+    print *, 'MOs'
 
-  rc = trexio_write_mo_type(f(1), mo_label, len(trim(mo_label)))
+    rc = trexio_write_mo_type(f(1), mo_label, len(trim(mo_label)))
-  call trexio_assert(rc, TREXIO_SUCCESS)
-
-  do k=1,N_states
-    rc = trexio_write_mo_num(f(k), mo_num)
     call trexio_assert(rc, TREXIO_SUCCESS)
-  enddo
 
-  rc = trexio_write_mo_coefficient(f(1), mo_coef)
+    do k=1,N_states
-  call trexio_assert(rc, TREXIO_SUCCESS)
+      rc = trexio_write_mo_num(f(k), mo_num)
+      call trexio_assert(rc, TREXIO_SUCCESS)
+    enddo
 
-  if ( (trim(mo_label) == 'Canonical').and. &
+    rc = trexio_write_mo_coefficient(f(1), mo_coef)
-       (export_mo_two_e_ints_cholesky.or.export_mo_two_e_ints) ) then
+    call trexio_assert(rc, TREXIO_SUCCESS)
-    rc = trexio_write_mo_energy(f(1), fock_matrix_diag_mo)
+
+    if ( (trim(mo_label) == 'Canonical').and. &
+         (export_mo_two_e_ints_cholesky.or.export_mo_two_e_ints) ) then
+      rc = trexio_write_mo_energy(f(1), fock_matrix_diag_mo)
+      call trexio_assert(rc, TREXIO_SUCCESS)
+    endif
+
+    rc = trexio_write_mo_class(f(1), mo_class, len(mo_class(1)))
     call trexio_assert(rc, TREXIO_SUCCESS)
   endif
 
-  rc = trexio_write_mo_class(f(1), mo_class, len(mo_class(1)))
-  call trexio_assert(rc, TREXIO_SUCCESS)
-
 ! One-e MO integrals
 ! ------------------
 

src/trexio/import_trexio_integrals.irp.f

@@ -3,6 +3,7 @@ program import_integrals_ao
   implicit none
   integer(trexio_t)              :: f ! TREXIO file handle
   integer(trexio_exit_code)      :: rc
+  PROVIDE mo_num
 
   f = trexio_open(trexio_filename, 'r', TREXIO_AUTO, rc)
   if (f == 0_8) then
@@ -42,10 +43,10 @@ subroutine run(f)
 
   if (trexio_has_nucleus_repulsion(f) == TREXIO_SUCCESS) then
     rc = trexio_read_nucleus_repulsion(f, s)
-    call trexio_assert(rc, TREXIO_SUCCESS)
     if (rc /= TREXIO_SUCCESS) then
       print *, irp_here, rc
       print *, 'Error reading nuclear repulsion'
+      call trexio_assert(rc, TREXIO_SUCCESS)
       stop -1
     endif
     call ezfio_set_nuclei_nuclear_repulsion(s)
@@ -63,6 +64,7 @@ subroutine run(f)
     if (rc /= TREXIO_SUCCESS) then
       print *, irp_here
       print *, 'Error reading AO overlap'
+      call trexio_assert(rc, TREXIO_SUCCESS)
       stop -1
     endif
     call ezfio_set_ao_one_e_ints_ao_integrals_overlap(A)
@@ -74,6 +76,7 @@ subroutine run(f)
     if (rc /= TREXIO_SUCCESS) then
       print *, irp_here
       print *, 'Error reading AO kinetic integrals'
+      call trexio_assert(rc, TREXIO_SUCCESS)
       stop -1
     endif
     call ezfio_set_ao_one_e_ints_ao_integrals_kinetic(A)
@@ -85,6 +88,7 @@ subroutine run(f)
 !    if (rc /= TREXIO_SUCCESS) then
 !      print *, irp_here
 !      print *, 'Error reading AO ECP local integrals'
+!      call trexio_assert(rc, TREXIO_SUCCESS)
 !      stop -1
 !    endif
 !    call ezfio_set_ao_one_e_ints_ao_integrals_pseudo(A)
@@ -96,6 +100,7 @@ subroutine run(f)
     if (rc /= TREXIO_SUCCESS) then
       print *, irp_here
       print *, 'Error reading AO potential N-e integrals'
+      call trexio_assert(rc, TREXIO_SUCCESS)
       stop -1
     endif
     call ezfio_set_ao_one_e_ints_ao_integrals_n_e(A)
@@ -106,41 +111,112 @@ subroutine run(f)
 
   ! AO 2e integrals
   ! ---------------
-  PROVIDE ao_integrals_map
 
-  integer*4 :: BUFSIZE
+  rc = trexio_has_ao_2e_int(f)
-  BUFSIZE=ao_num**2
+  PROVIDE ao_num
-  allocate(buffer_i(BUFSIZE), buffer_values(BUFSIZE))
+  if (rc /= TREXIO_HAS_NOT) then
-  allocate(Vi(4,BUFSIZE), V(BUFSIZE))
+      PROVIDE ao_integrals_map
 
-  integer*8 :: offset, icount
+      integer*4 :: BUFSIZE
+      BUFSIZE=ao_num**2
+      allocate(buffer_i(BUFSIZE), buffer_values(BUFSIZE))
+      allocate(Vi(4,BUFSIZE), V(BUFSIZE))
 
-  offset = 0_8
+      integer*8 :: offset, icount
-  icount = BUFSIZE
+
-  rc = TREXIO_SUCCESS
+      offset = 0_8
-  do while (icount == size(V))
+      icount = BUFSIZE
-    rc = trexio_read_ao_2e_int_eri(f, offset, icount, Vi, V)
+      rc = TREXIO_SUCCESS
-    do m=1,icount
+      do while (icount == size(V))
-      i = Vi(1,m)
+        rc = trexio_read_ao_2e_int_eri(f, offset, icount, Vi, V)
-      j = Vi(2,m)
+        do m=1,icount
-      k = Vi(3,m)
+          i = Vi(1,m)
-      l = Vi(4,m)
+          j = Vi(2,m)
-      integral = V(m)
+          k = Vi(3,m)
-      call two_e_integrals_index(i, j, k, l, buffer_i(m) )
+          l = Vi(4,m)
-      buffer_values(m) = integral
+          integral = V(m)
-    enddo
+          call two_e_integrals_index(i, j, k, l, buffer_i(m) )
-    call insert_into_ao_integrals_map(int(icount,4),buffer_i,buffer_values)
+          buffer_values(m) = integral
-    offset = offset + icount
+        enddo
+        call insert_into_ao_integrals_map(int(icount,4),buffer_i,buffer_values)
+        offset = offset + icount
+        if (rc /= TREXIO_SUCCESS) then
+            exit
+        endif
+      end do
+      n_integrals = offset
+
+      call map_sort(ao_integrals_map)
+      call map_unique(ao_integrals_map)
+
+      call map_save_to_disk(trim(ezfio_filename)//'/work/ao_ints',ao_integrals_map)
+      call ezfio_set_ao_two_e_ints_io_ao_two_e_integrals('Read')
+
+      deallocate(buffer_i, buffer_values, Vi, V)
+      print *, 'AO integrals read from TREXIO file'
+  else
+      print *, 'AO integrals not found in TREXIO file'
+  endif
+
+  ! MO integrals
+  ! ------------
+
+  allocate(A(mo_num, mo_num))
+  if (trexio_has_mo_1e_int_core_hamiltonian(f) == TREXIO_SUCCESS) then
+    rc = trexio_read_mo_1e_int_core_hamiltonian(f, A)
     if (rc /= TREXIO_SUCCESS) then
-        exit
+      print *, irp_here
+      print *, 'Error reading MO 1e integrals'
+      call trexio_assert(rc, TREXIO_SUCCESS)
+      stop -1
     endif
-  end do
+    call ezfio_set_mo_one_e_ints_mo_one_e_integrals(A)
-  n_integrals = offset
+    call ezfio_set_mo_one_e_ints_io_mo_one_e_integrals('Read')
+  endif
+  deallocate(A)
 
-  call map_sort(ao_integrals_map)
+  ! MO 2e integrals
-  call map_unique(ao_integrals_map)
+  ! ---------------
 
-  call map_save_to_disk(trim(ezfio_filename)//'/work/ao_ints',ao_integrals_map)
+  rc = trexio_has_mo_2e_int(f)
-  call ezfio_set_ao_two_e_ints_io_ao_two_e_integrals('Read')
+  if (rc /= TREXIO_HAS_NOT) then
+
+      BUFSIZE=mo_num**2
+      allocate(buffer_i(BUFSIZE), buffer_values(BUFSIZE))
+      allocate(Vi(4,BUFSIZE), V(BUFSIZE))
+
+
+      offset = 0_8
+      icount = BUFSIZE
+      rc = TREXIO_SUCCESS
+      do while (icount == size(V))
+        rc = trexio_read_mo_2e_int_eri(f, offset, icount, Vi, V)
+        do m=1,icount
+          i = Vi(1,m)
+          j = Vi(2,m)
+          k = Vi(3,m)
+          l = Vi(4,m)
+          integral = V(m)
+          call two_e_integrals_index(i, j, k, l, buffer_i(m) )
+          buffer_values(m) = integral
+        enddo
+        call map_append(mo_integrals_map, buffer_i, buffer_values, int(icount,4))
+        offset = offset + icount
+        if (rc /= TREXIO_SUCCESS) then
+            exit
+        endif
+      end do
+      n_integrals = offset
+
+      call map_sort(mo_integrals_map)
+      call map_unique(mo_integrals_map)
+
+      call map_save_to_disk(trim(ezfio_filename)//'/work/mo_ints',mo_integrals_map)
+      call ezfio_set_mo_two_e_ints_io_mo_two_e_integrals('Read')
+      deallocate(buffer_i, buffer_values, Vi, V)
+      print *, 'MO integrals read from TREXIO file'
+  else
+      print *, 'MO integrals not found in TREXIO file'
+  endif
 
 end

src/utils/integration.irp.f

@@ -468,8 +468,6 @@ end subroutine
 
 
 
-
-
 subroutine multiply_poly(b,nb,c,nc,d,nd)
   implicit none
   BEGIN_DOC
@@ -484,33 +482,292 @@ subroutine multiply_poly(b,nb,c,nc,d,nd)
 
   integer                        :: ndtmp
   integer                        :: ib, ic, id, k
-  if(ior(nc,nb) >= 0) then ! True if nc>=0 and nb>=0
+  if(ior(nc,nb) < 0) return !False if nc>=0 and nb>=0
-    continue
+
-  else
+  select case (nb)
-    return
+    case (0)
-  endif
+      call multiply_poly_b0(b,c,nc,d,nd)
-  ndtmp = nb+nc
+      return
+    case (1)
+      call multiply_poly_b1(b,c,nc,d,nd)
+      return
+    case (2)
+      call multiply_poly_b2(b,c,nc,d,nd)
+      return
+  end select
+
+  select case (nc)
+    case (0)
+      call multiply_poly_c0(b,nb,c,d,nd)
+      return
+    case (1)
+      call multiply_poly_c1(b,nb,c,d,nd)
+      return
+    case (2)
+      call multiply_poly_c2(b,nb,c,d,nd)
+      return
+  end select
+
+  do ib=0,nb
+    do ic = 0,nc
+      d(ib+ic) = d(ib+ic) + c(ic) * b(ib)
+    enddo
+  enddo
+
+  do nd = nb+nc,0,-1
+    if (d(nd) /= 0.d0) exit
+  enddo
+
+end
+
+
+subroutine multiply_poly_b0(b,c,nc,d,nd)
+  implicit none
+  BEGIN_DOC
+  ! Multiply two polynomials
+  ! D(t) += B(t)*C(t)
+  END_DOC
+
+  integer, intent(in)            :: nc
+  integer, intent(out)           :: nd
+  double precision, intent(in)   :: b(0:0), c(0:nc)
+  double precision, intent(inout) :: d(0:nc)
+
+  integer                        :: ndtmp
+  integer                        :: ic, id, k
+  if(nc < 0) return !False if nc>=0
 
   do ic = 0,nc
     d(ic) = d(ic) + c(ic) * b(0)
   enddo
 
-  do ib=1,nb
+  do nd = nc,0,-1
-    d(ib) = d(ib) + c(0) * b(ib)
+    if (d(nd) /= 0.d0) exit
-    do ic = 1,nc
-      d(ib+ic) = d(ib+ic) + c(ic) * b(ib)
-    enddo
-  enddo
-
-  do nd = ndtmp,0,-1
-    if (d(nd) == 0.d0) then
-      cycle
-    endif
-    exit
   enddo
 
 end
 
+subroutine multiply_poly_b1(b,c,nc,d,nd)
+  implicit none
+  BEGIN_DOC
+  ! Multiply two polynomials
+  ! D(t) += B(t)*C(t)
+  END_DOC
+
+  integer, intent(in)            :: nc
+  integer, intent(out)           :: nd
+  double precision, intent(in)   :: b(0:1), c(0:nc)
+  double precision, intent(inout) :: d(0:1+nc)
+
+  integer                        :: ndtmp
+  integer                        :: ib, ic, id, k
+  if(nc < 0) return !False if nc>=0
+
+
+  select case (nc)
+    case (0)
+      d(0) = d(0) + c(0) * b(0)
+      d(1) = d(1) + c(0) * b(1)
+
+    case (1)
+      d(0) = d(0) + c(0) * b(0)
+      d(1) = d(1) + c(0) * b(1) + c(1) * b(0)
+      d(2) = d(2) + c(1) * b(1)
+
+    case default
+      d(0) = d(0) + c(0) * b(0)
+      do ic = 1,nc
+        d(ic) = d(ic) + c(ic) * b(0) + c(ic-1) * b(1)
+      enddo
+      d(nc+1) = d(nc+1) + c(nc) * b(1)
+
+  end select
+
+  do nd = 1+nc,0,-1
+    if (d(nd) /= 0.d0) exit
+  enddo
+
+end
+
+
+subroutine multiply_poly_b2(b,c,nc,d,nd)
+  implicit none
+  BEGIN_DOC
+  ! Multiply two polynomials
+  ! D(t) += B(t)*C(t)
+  END_DOC
+
+  integer, intent(in)            :: nc
+  integer, intent(out)           :: nd
+  double precision, intent(in)   :: b(0:2), c(0:nc)
+  double precision, intent(inout) :: d(0:2+nc)
+
+  integer                        :: ndtmp
+  integer                        :: ib, ic, id, k
+  if(nc < 0) return !False if nc>=0
+
+  select case (nc)
+    case (0)
+      d(0) = d(0) + c(0) * b(0)
+      d(1) = d(1) + c(0) * b(1)
+      d(2) = d(2) + c(0) * b(2)
+
+    case (1)
+      d(0) = d(0) + c(0) * b(0)
+      d(1) = d(1) + c(0) * b(1) + c(1) * b(0)
+      d(2) = d(2) + c(0) * b(2) + c(1) * b(1)
+      d(3) = d(3) + c(1) * b(2)
+
+    case (2)
+      d(0) = d(0) + c(0) * b(0)
+      d(1) = d(1) + c(0) * b(1) + c(1) * b(0)
+      d(2) = d(2) + c(0) * b(2) + c(1) * b(1) + c(2) * b(0)
+      d(3) = d(3) + c(2) * b(1) + c(1) * b(2)
+      d(4) = d(4) + c(2) * b(2)
+
+    case default
+
+      d(0) = d(0) + c(0) * b(0)
+      d(1) = d(1) + c(0) * b(1) + c(1) * b(0)
+      do ic = 2,nc
+        d(ic) = d(ic) + c(ic) * b(0) + c(ic-1) * b(1) + c(ic-2) * b(2)
+      enddo
+      d(nc+1) = d(nc+1) + c(nc) * b(1) + c(nc-1) * b(2)
+      d(nc+2) = d(nc+2) + c(nc) * b(2)
+
+  end select
+
+  do nd = 2+nc,0,-1
+    if (d(nd) /= 0.d0) exit
+  enddo
+
+end
+
+
+subroutine multiply_poly_c0(b,nb,c,d,nd)
+  implicit none
+  BEGIN_DOC
+  ! Multiply two polynomials
+  ! D(t) += B(t)*C(t)
+  END_DOC
+
+  integer, intent(in)            :: nb
+  integer, intent(out)           :: nd
+  double precision, intent(in)   :: b(0:nb), c(0:0)
+  double precision, intent(inout) :: d(0:nb)
+
+  integer                        :: ndtmp
+  integer                        :: ib, ic, id, k
+  if(nb < 0) return !False if nb>=0
+
+  do ib=0,nb
+      d(ib) = d(ib) + c(0) * b(ib)
+  enddo
+
+  do nd = nb,0,-1
+    if (d(nd) /= 0.d0) exit
+  enddo
+
+end
+
+
+subroutine multiply_poly_c1(b,nb,c,d,nd)
+  implicit none
+  BEGIN_DOC
+  ! Multiply two polynomials
+  ! D(t) += B(t)*C(t)
+  END_DOC
+
+  integer, intent(in)            :: nb
+  integer, intent(out)           :: nd
+  double precision, intent(in)   :: b(0:nb), c(0:1)
+  double precision, intent(inout) :: d(0:nb+1)
+
+  integer                        :: ndtmp
+  integer                        :: ib, ic, id, k
+  if(nb < 0) return !False if nb>=0
+
+  select case (nb)
+    case (0)
+      d(0) = d(0) + c(0) * b(0)
+      d(1) = d(1) + c(1) * b(0)
+
+    case (1)
+      d(0) = d(0) + c(0) * b(0)
+      d(1) = d(1) + c(0) * b(1) + c(1) * b(0)
+      d(2) = d(2) + c(1) * b(1)
+
+    case default
+      d(0) = d(0) + c(0) * b(0)
+      do ib=1,nb
+        d(ib) = d(ib) + c(0) * b(ib) + c(1) * b(ib-1)
+      enddo
+      d(nb+1) = d(nb+1) + c(1) * b(nb)
+
+  end select
+
+  do nd = nb+1,0,-1
+    if (d(nd) /= 0.d0) exit
+  enddo
+
+end
+
+
+subroutine multiply_poly_c2(b,nb,c,d,nd)
+  implicit none
+  BEGIN_DOC
+  ! Multiply two polynomials
+  ! D(t) += B(t)*C(t)
+  END_DOC
+
+  integer, intent(in)            :: nb
+  integer, intent(out)           :: nd
+  double precision, intent(in)   :: b(0:nb), c(0:2)
+  double precision, intent(inout) :: d(0:nb+2)
+
+  integer                        :: ndtmp
+  integer                        :: ib, ic, id, k
+  if(nb < 0) return !False if nb>=0
+
+  select case (nb)
+    case (0)
+      d(0) = d(0) + c(0) * b(0)
+      d(1) = d(1) + c(1) * b(0)
+      d(2) = d(2) + c(2) * b(0)
+
+    case (1)
+      d(0) = d(0) + c(0) * b(0)
+      d(1) = d(1) + c(0) * b(1) + c(1) * b(0)
+      d(2) = d(2) + c(1) * b(1) + c(2) * b(0)
+      d(3) = d(3) + c(2) * b(1)
+
+    case (2)
+      d(0) = d(0) + c(0) * b(0)
+      d(1) = d(1) + c(0) * b(1) + c(1) * b(0)
+      d(2) = d(2) + c(0) * b(2) + c(1) * b(1) + c(2) * b(0)
+      d(3) = d(3) + c(1) * b(2) + c(2) * b(1)
+      d(4) = d(4) + c(2) * b(2)
+
+    case default
+      d(0) = d(0) + c(0) * b(0)
+      d(1) = d(1) + c(0) * b(1) + c(1) * b(0)
+      do ib=2,nb
+        d(ib) = d(ib) + c(0) * b(ib) + c(1) * b(ib-1) + c(2) * b(ib-2)
+      enddo
+      d(nb+1) = d(nb+1) + c(1) * b(nb) + c(2) * b(nb-1)
+      d(nb+2) = d(nb+2) + c(2) * b(nb)
+
+  end select
+
+  do nd = nb+2,0,-1
+    if (d(nd) /= 0.d0) exit
+  enddo
+
+end
+
+
+
+
 subroutine multiply_poly_v(b,nb,c,nc,d,nd,n_points)
   implicit none
   BEGIN_DOC