Finding adjacent minima from the saddle

Once a saddle is reached, the calculation can either be stopped, or continued to compute the adjacent minima. This is controlled by lpush_final flag.

In case the final push is performed, the algorithm will push the saddle structure once in the direction of the eigenvector, and then let it relax. When the first minimum is found, the structure will be brought back to the saddle, and pushed in the opposite direction of the eigenvector, and let to relax.

Size of the final push

The size of final push-over step is controlled by push_over.

Relaxation to minima

The minimzation from saddle is controlled by the FIRE parameters such as timestep, alpha, inc/dec factors, etc. The modifications of these parameters depends on your E/F engine, and the mode of using pARTn, as follows:

QE

The FIRE parameters can be modified from the QE input, in the IONS card: https://www.quantum-espresso.org/Doc/INPUT_PW.html#idm1103

LAMMPS

The FIRE parameters can be modified using the min_modify command, see: https://docs.lammps.org/min_modify.html

Siesta

The Siesta interface uses artn_step() functionality, thus the internal FIRE is used. Its parameters can be modified by adding a namelist &fire_params / into artn.in, unless a different infile was specified.

VASP

The VASP interface uses artn_step() functionality, thus the internal FIRE is used. Its parameters can be modified by adding a namelist &fire_params / into artn.in, unless a different infile was specified.