Dear lammps users:
Now I want to strain a graphene system by lammps command ‘deform’. In my
simulation, the stress-strain cure is very similar to the other previous
work. But now I have a problem about the limit strength of the system.
When I use nve ensemble and command deform, the system will break down
suddenly at the point of fracture and all atoms will explode fully fill in
the simulation box disorderly. Could anyone give me any possible reason for
this?
what happens if you stretch a rubberband beyond the point of breakage?
it snaps back. that means the potential energy stored in the band is
suddenly released. and the snap is more intense the stronger the
rubberband is. now graphene has very strong bonds thus a *lot* of
potential energy is released.
When I use nvt ensemble, the system breaking down problem will disappear,
and the system can be strained continually. But now the problem is that the
system can bear a much larger strain than other previous work. Strangely,
the system could not be strained to break easily.
Why the nvt and nve ensemble give such different results?
first off, neither simulation is in "nve" or "nvt" ensemble. your
system does not fit the requirements for those statistical mechanical
ensemble.
back to the issue at hand:
what you see is a consequence of your choice of simulation parameters
and basically your choice is far from ideal.
basically you may be interested in two different kind of scenarios:
- determine the 0K behavior and infinitely small strain rate.
- determine the finite temperature behavior at a specific strain rate.
if you want the 0K behavior you have to use minimization instead of MD.
if you want finite temperature, you first equilibrate your system to a
given temperature, but then run without(!) a thermostat. if you take
the graphene sheet, you have an isolated system. what would it couple
to? and exchange kinetic energy with? thus you would need to run with
fix nve only. of course the straining has to heat up the system and of
course that is affected by the speed of the straining. you also have
to make sure that your time step is small enough. the higher the
strain, the shorter the time step has to be because the potential gets
steeper.
the difference you see is simply due to the fact, that you are trying
to use the thermostats in a rather unphysical way. because of its
design temp/berendsen is able to remove kinetic energy at pretty
brutal rates. basically, you have moved far from running dynamics and
are forcing the system close to where it would be at a 0K procedure.
however, the nose-hoover chains in fix nvt are meant to correctly
model a canonical ensemble with weak(!!) coupling to a heat bath. it
cannot exchange a large amount of kinetic energy quickly.
HTH,
axel.