Stopping criterion = trust region too small

LAMMPS LAMMPS Mailing List Mirror
1

Hello,

I am performing a minimization calculation. I have 1900 atoms in a periodic box. I am using the following criteria for minimization:

fix 1 all box/relax iso 0.0 vmax 0.001
min_style hftn # Hessian-free truncated Newton algorithm (similar to cg, not affected by min_modify)
min_modify line quadratic #not as robust, but will generally reduce the forces closer to 0
minimize 1.0e-20 1.0e-20 100000 10000000

The programs always stops with “Stopping criterion = trust region too small” even when the forces are very large. The above worked well for smaller number of atoms.

Thank you,

Vivek

2

That's a message from the hftn minimizer, which is a trust-region
method. Try a different one. Minimizers can be cantankerous,
esp when you are using fix box/relax. See the doc page on
it for various issues you can have. And try restarting the minimizer.

Steve