[lammps-users] which time integrating ensemble with fix deform? need help with shear

Lammps users,
I am struggling to find a good combination of fixes and commands to deform a solid while allowing non-isotropic strain response to applied macroscopic stresses/strains. I would like to apply a simple shear or coupled compression/tension to a system.

Could someone please recommend which barostat/thermostat/time integration settings they would use in combination with fix deform? I read in the forum that fix nvt may not have the correct dynamics for a deforming box, but not why. Fix npt also cannot be applied in the same axis as fix deform. This would leave fix nve as time integrator, but does it have the same unspecified limitations as fix nvt?

Here are some settings I think I can rule out:

  1. I noticed with the example file in.shear that a vacuum opened between the top set force=zero region and the deforming region for the amorphous solid I was modeling and would like to avoid this, preferably through relaxation of pressure in the z-direction. Also, the temperature in the direction of shear is not included in the thermostat, but I would like to have this.
  2. With fix deform command using shear in xz, I need some clarification as to how the periodic images should be placed and whether there might be rotation of the sample going on behind the scenes in order to report the tilt factors for only xy xz yz. For example, if I placed a periodic image +x(i) with its xlo(i) at the xhi(o) reported in the dump file, I would be opening a variable width of vacuum between these two images during tilt. I say this based on my understanding of lattice vectors- if this is not so, how should I understand the reported values please?

Joshua Askin

You can you fix nvt, or fix nve + fix langevin. If you are shearing
fast, then you probably want to use compute temp/deform as the
temperature assigned to either fix nvt or fix langevin, so that
you only thermostat the thermal components of velocity, and not
the streaming velocity due to the deformation. However if
you are using remap x with fix deform, then you may not be imparting
significant velocities to the deforming atoms, so you may be
able to just use compute temp.