Reax/c: problem with elastic coefficient reproduction

Dear All,

I have been looking into the reax potential for ZnO modelling. To start, I tried to reproduce the 0K wurtzite lattice constants and elastic constants of the reference paper (Surf Sci, 604, 741-752 (2010)). But so far have not been able to obtain similar values for the elastic constants.

Reference: C11=208, C33=287, C12=81, C13=113, C44=52 (GPa)

Calculated: C11=181, C33=230, C12=81, C13=113, C44=32 (GPa)

for strain increments of 1e-4 to 1e-2

I have obtained a reax parameter file from prof. van Duin and made sure that the values are the same as the LAMMPS provided ffield.reax.ZnOH. Also repeated the calculation for a morse type potential for ZnO and found good agreement with its published elastic coefficient value.

So I am wondering:

  1. How the ffield.reax.ZnOH parameters was tested against Surf Sci, 604, 741-752 (2010) i.e. for which phase/properties?

  2. Are there additional parameters in the control file that could be varied to give better reproduction?

Thanks for your help and any other suggestions.

Best Regards,


P.S. Used the in.elastic script file included with LAMMPS; Fix qeq tolerance: 1e-6, minimization etol and ftol: 1e-9.

Based on past threads, tried running without success: reax/c without the control file; min_style cg, sd, hftn,fire (so also without the initial box/relax); pairstyle reax; Varied the tbh_cutoff and tbh_cutoff_sq values from the control file, which gave better C33 values with C11, C44 unchanged.

I just ran examples/reax/ZNOH2/in.ZnOH2 and compared log.lammps with a copy I had from 2011 (when this example was first constructed and verified againt the serial code). Pressure and energy match to 8 digits on timestep 0 and the match is good after 3000 timesteps.

[athomps@…7878… ZnOH2]$ diff log.lammps …/lammps-17Feb12/examples/reax/Zn_O_H/log.reaxc.znoh.18Feb11.linux.1



< 3000 551.8036 -7939.3294 0 -7768.2679 -840.70849

< Loop time of 3.5449 on 1 procs for 3000 steps with 105 atoms