Best ensemble choice in stress-strain simulations

[marcelolpjunior]

Dear Ones,

I have recently been working on understanding scripts for obtaining the mechanical properties of structures in general.

I have noticed these two possibilities for traction of structures. One is fixing sets of atoms at the edges and using the set move, but since these atoms don’t evolve their positions beyond the one they are moving, I notice situations that don’t seem very physical after all… I mean, when stretching a 2D structure in the x-direction, always the breakage has happened at the edges since the structure is not allowed to adjust in the y-direction due to the clamped edges.

Another possibility that seemed physically better to me was to use the fix deform in periodic systems, but then I have had problems with the choice of ensemble, especially in 1D systems. Hence comes my first (and maybe main) question, how can I stretch a system with NVT ensemble in a physically correct way? I mean, the volume is constant, how is it possible to increase the box in one direction and keep the volume constant? Am I misinterpreting? In particular, when using the NPT system, I see that my structure is forced and sometimes “breaks” connections, and when using NVT, the stress-strain curve does not start with stress at zero, i.e., there is already stress associated with the structure before the start of the pull. That doesn’t seem correct to me either.

Also, even in two-dimensional systems, I have noticed that when stretching a direction, when plotting the stress-strain curve, there is stress associated with directions that are not being stretched, due to the simulation box not adjusting in that direction, i.e., with a simulation with fix deform in the x-direction, the stress-strain curve has stress in the x and y directions, even proportionally.

So the main discussions I would like to raise here are:

1. To what extent is it physically interesting to stretch finite systems, with fixed edges? Are there ways to make these simulations more realistic?

2. What is the accuracy/explanation of using an NVT with fix deform?

3. In practice, what is the difference between remap x, remap y, or not using remap in fix deform?

4. When using from NPT ensemble in stress-strain simulations, what is the best way to use constant pressure (iso 0.0 0.0 0.0, or x 0.0 0.0 y 0.0 0.0 …)?

Thank you for your attention.

[akohlmey]

@marcelolpjunior

Hence comes my first (and maybe main) question, how can I stretch a system with NVT ensemble in a physically correct way? I mean, the volume is constant, how is it possible to increase the box in one direction and keep the volume constant? Am I misinterpreting? In particular, when using the NPT system, I see that my structure is forced and sometimes “breaks” connections, and when using NVT, the stress-strain curve does not start with stress at zero, i.e., there is already stress associated with the structure before the start of the pull. That doesn’t seem correct to me either

Here is a major mistake in understanding. You do NOT choose an ensemble, you choose an integrator. If you use fix nvt, you will only get an NVT ensemble, if you have a fully periodic, equilibrated, bulk system and no further manipulations. Fix nvt does nothing to prevent volume changes, but - unlike fix npt - it does not change the volume either. This is documented.