Unsensitive airebo cutoff for tensile streth of CNT

Dear all,

I want to simulate a stretching CNT and calculate its tensile strength. I need to calibrate the system for more complex geometries. I’d like to reproduce the results in https://www.sciencedirect.com/science/article/pii/S2095034915000525.

I have produced a 10x10 armchair, 4.8 nm CNT using scikitnano and want to compare rebo vs airebo vs airebo-m for several cutoff parameters with LAMMPS.

I perform energy minimization, followed by a NVE + Berendsen relax and a NPT relax.
Finally I apply a “fix all deform 1 z scale 2”. (I also tested the erate option instead of scale for comparison).

I calculate the strain rate as tot_ratio/tot_time and the stress as follows:

variable dt equal 0.001 #ps
timestep ${dt} #ps

compute 1 all stress/atom NULL
compute 2 all reduce sum c_1[3]
fix 1 all ave/time 1 100 100 c_2
variable sigmazz equal f_1

##--------------- STRAIN -----------------------
variable Zratio equal 2
variable srate equal ${Zratio}/${Nruns}/${dt} #/ps
variable strain equal step*v_srate*v_dt

fix 2 all ave/time 1 1 100 v_strain v_sigmazz file deform.dat
dump 1 all atom 100 dump_deform.lammpstrj
fix 3 all npt temp 300 300 0.1 x 0 0 0.5 y 0 0 0.5
fix 4 all deform 1 z scale ${Zratio}

I get at least three problems:

  1. airebo-m shows no improvement with respect to rebo: there is an un-physical hardening above 5% of strain
  2. there is almost no difference when changing the LJ cutoff from 2.0 to 5.0 Ang
  3. generally I’d expect a tensile strength one order of magnitude larger (about 100GPa).

I can imagine there is either a problem in my stress calculation or generally in my minimization/relaxation scheme. Could anyone suggest a solution or improvement?
Please find attached my input and geometry files for reproduction.

Thank you.

input.lammps (3.1 KB)
cnt_10.lmp (33.5 KB)

Why should it? As stated in the documentation, it is meant to improve the situation for high-density systems (e.g. poly-ethylene under high pressure) , so there is no reason why it should improve behavior for stretching of compounds.

As the documentation clearly states, the (optional) LJ cutoff is not given in angstroms but in multiples of sigma_CC = 3.4 Angstrom. since your interaction energy under straining conditions for your kind of system is dominated by the (strong) C-C bonds, it is no surprise that the LJ cutoff has very little impact.

I cannot comment on that. You should look up what others have found that studied the same kind of system under similar conditions. You are not likely the first. So you should check the published literature. The challenge is likely to find suitable publication among the many publications…