Too large stresses

Hi! I have a question about the command

compute stress/atom

I made a simulation with Fe and Cu. No load was applied, units was set to “metal” and the ensemble was relaxed by the NVT integrator for 4000 runs with the timestep 0.005 (ps). When I plotted the results I saw that the magnitudes of the stresses were way too high (about x100 too large). To check if the error was due to that 2 elements were involved I made the same simulation again but now the ensemble only contained Fe atoms. However, once again the magnitudes of the stresses were too large after the relaxation.

I have used the following expressions to calculate and averaging the stresses:

compute NAME GRPNAME stress/atom NULL pair virial

fix NAME GRPNAME ave/atom 1 200 4000 c_stressM[1] c_stressM[2] c_stressM[3] c_stressM[4] c_stressM[5] c_stressM[6]

I used the FeCuNi.eam.alloy potential. What might be the error? Isn’t “BAR” the unit that is used when units is set to “metal”?

See the compute stress/atom doc page for units. Note there is a volume
term missing. It also tells you how to check that your per-atom stresses
are consistent with the overall system pressure. If that test passes,

your per-atom stresses are correct (assuming you also think the global
pressure is OK).