Dear LAMMPS users,

I am trying to calculate the elastic constants of Al2O3 by using the LAMMPS example scripts and with the SMTBQ potential as provided in lammps. The minimization with CG (or sd) terminates with the message “line search alpha is zero”, for few of 6 deformation directions, where the fnorm is large. This leads to out of range elastic constants. Is there a way to relax the atoms positions after deformation such that the fnorm is at least in decimal if not around 10^{-5}?

Thank you in advance for your suggestions.

Regards,

Shyamal Roy

Hello Shyamal, the error message in a nutshell means that no suitable local minimum could be found. Can you try first minimizing the initial system (optionally using fix box/relax as well) and taking smaller increments when you deform the unit cell?

Giacom

Dear Giacomo,

Thank you for your suggestion. I minimized the initial structure using fix box/deform as well (with final force norm ~10^-7 and pressure ~10^-2 ), and deformed the unit cell with small increments say 10^-6 or 10^-8. However, for some deform directions the minimization after ‘change_box’ does not converge, and the stopping criterion is line search alpha zero. The force norm does not go to decimals. I am not sure what exactly I am missing in this case!

Regards,

Shyamal

Dear Giacomo,

Thank you for your suggestion. I minimized the initial structure using fix box/relax as well (with final force norm ~10^-7 and pressure ~10^-2 ), and deformed the unit cell with small increments say 10^-6 or 10^-8. However, for some deform directions the minimization after ‘change_box’ does not converge, and the stopping criterion is line search alpha zero. The force norm does not go to decimals. I am not sure what exactly I am missing in this case!

Regards,

Shyamal