minimize

Syntax

minimize mini_style max_iteration tolerance
  • mini_style = cg or sd or fire or qm

  • max_iteration = positive integer

  • tolerance = positive real number

Examples

minimize cg 1000 1d-5
minimize fire 100 1d-6

Description

This command sets the style and two parameters for the energy minimization in quasistatic and hybrid CAC.

There are four mini_style: congjugate gradient (cg), steepest descent (sd), fast inertial relaxation engine (fire), and quick min (qm).

Both cg and sd use the negative gradient of internal energy as the initial direction; from the second step, however, the sd style uses the current negative gradient while the cg style uses the negative gradient conjugated to the current potential surface. Once the direction is set, the inner iterations begin in which a line search is conducted to determine the length by which the nodes/atoms need to move along the designated direction to find the minimized energy. For more information of the energy minimization with these two styles, read chapter 3 of Shuozhi Xu's Ph.D. dissertation.

The fire style is based on Bitzek et al., 2006 while the qm style is based on quenched dynamics which is used also in dynamic CAC. The difference is that only one quenched dynamics iteration is carried out at each simulation step in dynamic CAC while many quenched dynamics iterations are performed at each simulation step during the energy minimization until the internal energy converges at that step. For the fire and qm styles, the inner iteration is irrelevant.

The energy minimization is considered to converge when either the number of outer iterations reaches max_iteraction or the energy variation between successive outer iterations divided by the energy of the current iteration is less than tolerance.

This command is relevant only when simulation_style = statics or hybrid.

quasi_statics.f90, mini_init.f90, update_mini.f90, mini_energy.f90, hybrid.f90, conjugate_gradient.f90, steepest_descent.f90, quick_mini.f90, and fire.f90

Default

minimize cg 1000 1d-6

results matching ""

    No results matching ""