Problem with fortran openmp reduction to vector

I've got the following problem, trying to parallelize my code. The simplified code looks like this:

   !$omp parallel do private(e, b0_vek) reduction(+:b_vek) schedule(static, chunk_elem)
   do e = 1, n_elem
      call AddSource(elem(e), b0_vek)
      slice=elem(e)%node(1:6) !vector with indices in the global node list
      b_vek(slice)=b_vek(slice)+b0_vek(1:6)
   end do  ! e = 1, n_elem

The summation is over all elements, each of which is defined by 6 nodes. In AddSource subroutine the vector b0_vek(1:6) is calculated for each element. It should be added to the certain part (slice) of the b_vek, however, the slices can be crossing for some elements, as elements can be neighbors and thus, can contain several same nodes.

The reduction operation doesn’t work properly here, is there any other possibility to do it with openMP?