[lammps-users] Can the ensemble be inconsistent in the relaxation and simulation stages?

Can the ensemble be inconsistent in the relaxation and simulation stages? For example, relaxation under NVT ensemble and deform stretching under NPT ensemble. The system can continue to maintain equilibration when the ensemble is changed?

Please do not confuse “ensemble” with the “integrator fix” that LAMMPS uses.
To have a correct statistical mechanical ensemble, several conditions must be met. The suitable time integrator fix in LAMMPS is just one of them.
That said, you can start equilibration with a fixed volume and a nose-hoover thermostat (i.e. using fix nvt) and then add a nose-hoover barostat (i.e. switching to fix npt).
However, if you are at the same time also deforming your system, you will definitely NOT be in an NPT ensemble.

How you “build” a system through multiple stages is a topic for a discussion with a person that can teach you how to do MD simulations, e.g. your adviser/supervisor or experienced colleagues. It is not a something that has anything to do with LAMMPS (you just happen to use LAMMPS, but how to “build” a simulation study is disconnected from the MD code, provided the MD code has all required features).

Sorry, I didn’t understand your answer. What I want to ask is whether the relaxation process and tensile process are not an ensemble? If so, why?

To be in an NVT or NPT ensemble you need to be in equilibrium and must not have any manipulations to the system happening except for the adjustments from the thermostat (and barostat).

Sorry, is “deform” stretching a balanced simulation or a non-equilibrium simulation?


this is really getting off-topic. this mailing list is about using LAMMPS, not to discuss statistical mechanics.

Sorry. Let me ask a clear question. In the relaxation process, NPT ensemble is used to obtain the equilibrium state of the system. Can NVT ensemble be used to maintain the equilibrium in the later simulation process?

This is not a “clear” question. Again your terminology is inadequate and we are going in circles now.
You are, again, apparently confusing an NPT ensemble with the fix npt command. You may be using fix npt, but you are not “using an NPT ensemble”. “ensemble” is a state not a functionality and, as already explained, your system cannot be in an NPT ensemble when it is not yet in equilibrium.

Whether it is meaningful to start a simulation with fix npt and then turning it off and use fix nvt instead depends on what you want to simulate and for what purpose. That certainly is technically possible, but whether that is meaningful depends crucially on the details of the simulation and the properties you want to obtain from it. Thus is more a question of the science and not really about LAMMPS. I have no time and am not interested to discuss your research with you and thus fill in for your adviser.

This is my final response on this topic.