I am running a simple system of a 100 atoms during relaxation (CG). When I try to use the force criterion (say Etol=1e-15 and Ftol=1e-2), I get the following output:
Stopping criterion = failed linesearch
Energy initial, next-to-last, final =
-779.812204171 -779.812204172 -779.812204172
Force two-norm initial, final = 0.0234379 0.0234369
Force max component initial, final = 0.00559694 0.00559675
Final line search alpha, max atom move = 0 0
Iterations, force evaluations = 3 1088
In the user manual for the minimize command, I found this criterion: “the line search fails because the step distance backtracks to 0.0”
Is it what happened here? What is exactly the meaning (and what could be the origin) of the step distance “backtracking to 0.0”?
Thank you for you assistance.
Try bigger tolerance for energy (e-10) and hopefully you won’t get the following error.
Thanks a lot for the suggestion. I would need to understand what is really happening though.
If I use the energy tolerance, my calculation converges. But I believe the relaxed state is not the ground state, and the system would have to overcome some energy barrier to reach it. I used the “fix setforce” command in my simulation, and see a non-zero force in the relaxed system. So I would like to set forces as a criterion for convergence, but get this “failed linesearch” message.
This is what I am trying to understand: what is the meaning of this message and what could have happened during my simulation to cause it?
2009/11/9 niaz abdorrahim <[email protected]…24…>
Maybe you can try with this command: min_modify line quadratic
Try it without setting forces to a non-zero value. LAMMPS lets
you use fix setforce during a min, but it may mess things up.
When the backtracker fails, it is typically due to the energy not
being consistent with the forces, which could be due to your
use of fix setforce.