The eon client provides three optimizers quick-min, conjugate gradients, and L-BFGS.
opt_method: The optimization method to use.
default: cg
options:
box: Optimizes the atom positions and box using quickmin
cg: Conjugate gradient
qm: Quickmin
lbfgs: Limited Memory Broyden-Fletcher-Goldfarb-Shanno QuasiNewton optimizer
fire: Fast inertial relaxation engine
convergence_metric: The metric to use to determine when an optimization is complete.
default: norm
options:
norm: The norm of the entire force vector
max_atom: The maximum force on any non-frozen atom
max_component: The maximum force on any non-frozen degree of freedom
converged_force: When the convergence_metric is smaller than this value (eV/A), the structure is considered minimized.
default: 0.01
max_move: Maximum distance that an atom may be moved in a single optimization step (Angstoms).
default: 0.2
time_step: The dynamical timestep for the quickmin algorithm (fs).
default: 1.0
max_iterations: The maximum number of optimization iterations that the will be performed.
default: 1000
The L-BFGS implementation in Eon resets its memory if a move larger than the max_move is attemped or if the angle between the force and the L-BFGS direction is larger than 90 degrees.
lbfgs_memory: The maximum number of previous gradients and positions that are kept in memory for the L-BFGS algorithm.
default: 20
lbfgs_inverse_curvature: The initial guess for inverse curvature at each step, which is used to construct the inverse Hessian matrix. This setting is only used when lbfgs_auto_scale is False.
default: 0.01
lbfgs_auto_scale: The initial guess for the inverse hessian at each step is constructed by estimating the curvature along the previous step direction. At the first step a forward finite difference calculation along the force is used to estimate the curvature.
default: True