problem with Cu surface relaxation

Dear LAMMPS users,

I have the following problem: I make relaxation of Cu(111) surface with lammps. According to, for instance, https://www.researchgate.net/publication/229342501_Construction_of_modified_embedded_atom_method_potentials_for_Cu_Pt_and_Cu-Pt_and_modelling_surface_segregation_in_Cu3Pt_alloys the first interlayer distance for the relaxed case should be lesser than in ideal lattice but I get opposite result.

Here is the code which I use: