Thanks for the reply.
I allow the xy box dimension to change by using the below command.
fix 3 all box/relax x 0.0 y 0.0 xy 0.0 couple none nreset 10
As I do not want dynamic effects, I do tensile test on honeycomb at 0 K using the minimize command.
I specified three different displacement increments in separate simulations: 0.1, 0.01, 0.001 Angstroms.
For the three cases the frozen layers get separated from the sample.
Do you have any clue? or dynamics is the only recourse?
Fix box/relax only operates during a minimize command, so
I assume you are using a minimize and not a run command?
Do the xy box lengths then change during the minimize? You
can monitor this by printing the box lengths (and volume) with
the thermo output. You should also print dump files to check
that the box dimensions are changing. Is the pxx and pyy going
towards zero (it may not get there for a solid)? If all those
things are happening, then it sounds like your simulation is working
correctly. If the atoms aren’t doing what you want, that is often
not a LAMMPS isssue, but a conceptual issue with your model.
This is intuition. Hope this helps.
When you are relaxing the lateral dimensions, system is breaking the symmetry (even to a smallest degree) due to changes in x and y direction (possibly due to precision issues). It is less susceptible in second case.
You can try following. For each step of minimization, do series of iterative MD ( nve ensemble) and minimization steps. After each MD run, reset the velocity to a zero. This might take you closer to global minimum at each minimization step.