[lammps-users] NPT with REAX

Dear users,

I’m trying NPT on vanadium oxides with REAX. But it always fails with different error.

After minimization, NPT stops. The error message is:

  1. ‘Too many bonds on atom. Increase MBONDDEF’
  • REAX precision is smaller than 1.0e-5. But I use MBONDDEF=60. MBONDDEF value larger than 60 makes compile error.
  1. making NAN values
  • REAX precision is larger than 1.0e-5.

Cutoff, box size, and time step made no effect on this.

Thank you.

I'll let Aidan respond to this. I think bonds > 60,
means something is fundamentally wrong with
your model.


2009/9/7 정광섭 (Kwangsub Jung) <[email protected]...>:

Sometimes ReaxFF does require a large number of bonds on one atom. For a
small system, you can try increasing the value of MBONDEF, reducing NATDEF,
and recompiling the library and LAMMPS.

However, it may also be that your vanadium oxide parameter file, which we
have never tested, requires a non-standard version of the ReaxFF potential.
Try talking to whomever you got the vanadium parameters from.

In situations like this, where you are using a LAMMPS potential for a system
that we have not tested LAMMPS on, it is a good idea to compare LAMMPS
forces and energies directly with the original code.