Is it possible to use move rotate on a group of atoms and simultaneously use an ensemble in that group?

Hello everyone,

I have a structure consisting of a cluster of tubes interacting by VdW. The correct name is tube bundle I suppose.

What I want:
Simulate the torsional resistance of this structure, using a reactive force field.

What I have done:

  • I created this tube bundle (it’s four tubes side by side) and made in that bundle three groups of atoms, which I called left, flake, and right, left and right being the edges of the structure.
  • The left group I made set force 0.0 0.0 0.0, i.e., they will not be integrated in time.
  • The flake I set with an NVT ensemble at room temperature.
  • The right group I did a fix move rotate with a period of 300 ps, and as the structure has its center of mass at the origin, I used (Px,Py,Pz) = (0,0,0). I placed the tubes with the longitudinal axis parallel to the Z axis, so (Rx,Ry,Rz) = (0,0,1).

My problem:

  • The tubes have a high diameter, and as the simulation progresses, the diameter at the flake decreases and at the edges, due to the atoms being attached, the diameter does not decrease. As a result, the rupture of the structure always occurs at the connection between the edge and the flake, which doesn’t seem very physically correct to me.

I thought that the solution was that the forces that act on the atoms take into account both the fix move (to have the rotation) and the integration in time with the NVT (so that the diameter adjusts).

I had a similar experience with LAMMPS, in tensile simulations of two-dimensional systems, where fixing the edge atoms always breaks the structure in the region between flake and edge. This problem was solved by using periodic boundary conditions, with deform command, which deforms the box instead of fixing atoms.

Is it possible to use move rotate on a group of atoms and simultaneously use an assembly in that group? Or some idea similar to the problem that I will present in detail later.

That is not correct. Fix setforce will remove forces, but does not exclude atoms from time integration. According to Newton’s Laws, an atom without a force will continue its motion unchanged with constant velocity.

This whole description doesn’t make much sense to me. It would probably help if you could provide some schematic or picture demonstrating what you mean.

That makes no sense at all. Both fix move and fix nvt do time integration, so they would conflict. You cannot have a prescribed motion and do time integration where atoms follow the forces. You would have to induce the rotation via forces or have some select atoms of the object you want to rotate be time integrated by fix move and the rest follow the forces and be time integrated by fix nvt.

It is not at all clear to me what using or not using periodic boundaries would change since this is not a bulk system.

I don’t understand what you are asking here. Perhaps you mistranslated some words?
The situation about time integration is quite clear. You cannot have the same atoms “move” by two different “moving” fixes. LAMMPS will print a warning in that case and the resulting trajectories are usually garbage or simulations will even crash.

1 Like

Hello, akohlmey!
Thank you for your interaction.

Enumerating your answers, one has:

(1) Ok. I misspoke about atoms with set force zero not being integrated in time. Thank you.

(2) This is the structure. I am twisting it, leaving the two edges fixed, with set force zero on one side and moving to rotate on the other end. The “middle” is simulated as usual, with an NVT ensemble. As the structure is twisted, the fixed atoms do not allow the diameter to decrease in that region, while it does in the “middle”. In this way, the structure breaks more because of this issue than because of the torsion itself. And this doesn’t seem very correct to me physically.

(3) and (4); Yes, I probably have some problems with the translation/English. Trying to rephrase my question as clearly as I can. I wonder if there is any way that I can make this rotation, according to the image you sent, and the atoms that make up the edges (fixed) can adjust somehow so that the diameter follows the rest of the structure (the “middle” part).

I apologize if all the questions are too simple, but I am really interested in solving this problem and need help.

I see two possible ways out of your dilemma:

  • create taller/longer bundles and keep the setup as is and only consider the part in the middle for your analysis. There will be a point where the finite size effect will become negligible.
  • instead of using fix move and immobilizing atoms, use the fix addtorque command — LAMMPS documentation at both ends (with opposite direction torque). There will be less of a finite size effect than in the other case, but I don’t think it is completely eliminated because your system is finite.

BTW: I am doubtful about the use of a thermostat. If there is no solvent, there is nothing to couple to and thus nothing that will exchange kinetic energy with the bundle. In general, a thermostat (in a production) simulation is supposed to do two things, 1) remove the (very small) amount of kinetic energy created by the discretization of the differential equations when solving the equations of motion numerically, and 2) induce kinetic energy fluctuations consistent with a bulk system. I don’t see either of those conditions apply here.