Dear LAMMPS developers,
I have a question regarding boundary conditions for a force controlled experiment.
I would like to define a simulation box that is constrained between two parallell plates, that are separated by a constant force. This force is specified by the «fix aveforce» command for the particles at the planes. What would be the difference between having shrink-wrapped boundaries, and fixed boundaries at the plates that are updated every time step by the change_box command? How do I avoid the need for remapping of coordinates?
Sincerely,
Eivind
Dear LAMMPS developers,
I have a question regarding boundary conditions for a force controlled experiment.
I would like to define a simulation box that is constrained between two parallell plates, that are separated by a constant force. This force is specified by the «fix aveforce» command for the particles at the planes. What would be the difference between having shrink-wrapped boundaries, and fixed boundaries at the plates that are updated every time step by the change_box command? How do I avoid the need for remapping of coordinates?
there should be no need to update box boundaries for such a
simulation, provided the box is large enough. thus a simulation with
shrinkwrap or fixed boundaries should result in the same trajectory.
the "real" boundary are the parallel plates, after all. the only
properties affected are those depending directly on the length of
shrinkwrap/fixed boundary, e.g. pressure.
axel.
Thank you for the fast reply. However, if the plate separation increase by an order of magnitude towards the end of the simulation, would not all the empty space in the simulation box be wasting computational resources?
The system under investigation is a polymer under constant tensile stress, immersed in a Langevin heath bath. The size of the simulation box also depends strongly on the force on the plates.
Sincerely,
Eivind
Thank you for the fast reply. However, if the plate separation increase by an order of magnitude towards the end of the simulation, would not all the empty space in the simulation box be wasting computational resources?
you can restrict the domain decomposition to become 2d with, e.g.,
processors * * 1
and/or address load imbalances with fix balance.
axel.