Different potential == different numerical material. It is typical to have different elasto-plastic response for the same “real” material modeled by different potential. There is plenty of literature on that topic. And it is not a Lammps issue, but a proper research question. NPT: probably to start the tenseil-test with a globally stress-free system. s, f and m do not generate a periodic direction. p does it. What is the unit of the pxx variable? Stress. What is the unit of the stree/atom compute? Stress*volume. All is described in the documentation of the compute, with a proper example. Julien