I did tensile tests and, after the sample fractures, the pressure along the tensile axis vibrates about zero with the amplitude gradually reduces. Since the pressure works on the simulation cell, I was wondering how such stress can affect the stress in the sample which is of interest.
In some cases (case 2 below), the stress is still (negative) larger than zero upon sample fracture, which means the stress obtained does not reflect the state of the sample.
Stresses can be negative or positive and can fluctuate
dramatically, especially in a non-equilibrium calculation.
Yes, I agree. Such fluctuation can also be observed when relaxing a newly-cut slab, and the the fluctuation damps out when the slab is equilibrated. The stress fluctuation near/after the fracture probably also results from the fracture surfaces which are in a non-equilibrium state. The question, if appropriate, is how we can “correct” the stress such that it resembles the experiment behavior. In experiment the stress suddenly drops to zero upon fracture…
Well, LAMMPS is computing what its formulas for stress (on the doc page)
say it is computing. How to interpret (or “correct”) that value to
something you think the expt (at the macro scale) is telling you,
is up to you.