What is suitable timestep and damping coefficients for metallic element uniaxial tension with LAMMPS involving 16000-32000 atoms?

LAMMPS LAMMPS General Discussion
1

I am working on creating nearly cube-sized BCC or FCC metallic element blocks (e.g. Al, W, Cu, Pt) containing several thousand atoms of same element (triply periodic boundary conditions) with x axis pointing towards various crystallographic directions (e.g. [1 4 2]) that is equilibriated for a while (say 0.01 metals unit for 5000 steps) under npt ensemble to a predetermined targeted temperature with 1.0 damping coefficient. All metallic elements utilize eam/alloy potential. Then it is deformed along x axis with strain rate ranging from 0.001 to 0.1 (again metals unit) with timesteps ranging from 0.005 to 0.02 metals unit, for 5000 to 100000 timesteps. Throughout the straining, along with attempt to fix temperature, an npt ensemble is maintained with zero imposed strains one two directions normal to loading, both with unity damping coefficient. The target is to observer stress-strain-temperature behaviour.

I am using automatic neighbor list recalculation with 2.0 units of ghost atom cutoff distance. I plan to show thermodynamic data in CLI after each 100 to 200 steps

Now my question is

  1. Are these values of damping coefficients and timesteps realistic enough for interior of a large-enough single crystal, provided that the temperature is 0.1 to 0.9 times (kelvin) melting point of the elements? do I need to change damping coefficients and timesteps? what is a suitable tradeoff as cited and benchmarked in reputed literature and benchmark files?
  2. Often I equilibriate once, store the data in a restart file, and run the deformations for various strain rates at same temperature. Is this scenario anyways erase or truncate some important system configuration that reduces realistic-ness of data?
2

These are questions for an expert in this specific area of science. You are not likely for find them here. In fact, this is not something that is specific to LAMMPS but would apply to any MD software that you would be using. So you need to discuss with your advisers and mentors and beyond need to devise mechanisms to determine what are suitable parameters. This is part of doing research and it will not do you well, if you have to explain your choices with “some dude on the internet told me they were OK”. What if we are wrong. Or - even worse - somebody here thinks that it is fun to give other people bad advice and make them look bad and himself good?

3

Axel is right that your questions would be better answered by experts (or books and publications). Still, I am curious about that bit:

Why do you vary the timestep? And is 0.02 really in pico-second? 20 fs is quite a huge timestep for atomic simulations, so you better have a serious argument for using it, otherwise I am sure that future reviewers will not receive your work very kindly.

Simon