Atom type-wise stress difference in multi element system

Dear experts,

Greetings. I have a question related to the per atom stress calculation by lammps using snap potential (probably true for any form of IP) for multi element system. In relaxed unit cell I am getting different stress for each atype. For AB2 system stress/A and stress/B are equal but opposite (diagonal components). Each of them should be (near) Zero at equilibrium I guess. For a ternary system I am seeing that stress for A, B, and C are very different, however their sum is zero. That probably related to the lammps, how it calculate stress/atom.
I will be appreciated if someone kindly explain me more And if there any tips to visualize stress distribution in the deformed structure in multi element system.

Best regards,

LAMMPS doesn’t pay much attention to the atom types when computing stress. The stress formalism that LAMMPS uses is “virial stress”. For that the contributions to the per-atom virial are taken from the distance vector \vec{r}_{ij} and the force vector \vec{F}_{ij} between pairs of atoms. Essentially the first sum in the definition of W_{ab} on the page compute stress/atom command — LAMMPS documentation

The other terms are for explicit bonded interactions, which are not present with SNAP.

Why should this be the case? Even if the whole system is at an overall stress-free equilibrium, why should two different atom types with two different parameterizations have the same per-atom virial?