Fixed bug in parallel EAM. rho_row and rho_col were scattered into the same array. Unfortunately, MPI zeros the array between scatters so half of the sum was being lost. Fixed by added a temp array for column scatter, then sum loop over nlocal.