riigid.convergence package

riigid.convergence.criterion module

class riigid.convergence.criterion.Compound_Criterion(cc1, cc2, operator)

Bases: Criterion

Compound convergence criterion.

Logical combination of two convergence criteria.

is_converged

Whether or not the convergence criterion is fulfilled.

Type:

bool

cc1, cc2

The two convergence criteria to combine.

Type:

riigid.convergence.Criterion

operator

How shall the two convergence criteria be combined?

Type:

str, options: “and”, “or”

check(optimization_history)

Check if the convergence criterion is fulfilled

If yes, self.is_converged is set to True.

Parameters:

optimization_history (list of riigid.Optimization_Step) – The history of the optimization, which shall be checked for convergence. (The optimization history is an attribute of the optimizer.)

class riigid.convergence.criterion.Criterion

Bases: object

Base class for RIIGID convergence criteria

is_converged

Whether or not the convergence criterion is fulfilled.

Type:

bool

check(optimization_history)

Check if the convergence criterion is fulfilled

If yes, self.is_converged is set to True.

Parameters:

optimization_history (list of riigid.Optimization_Step) – The history of the optimization, which shall be checked for convergence. (The optimization history is an attribute of the optimizer.)

riigid.convergence.displacement module

class riigid.convergence.displacement.Criterion_Displacement(cutoff=0.001)

Bases: Criterion

RIIGID convergence criterion: Displacement

If all atoms move less than cutoff, the criterion is fulfilled.

Note

This criterion could theoretically be fulfilled, even if the calculation is not actually converged. This is, because the displacement of the atoms not only depends on the force and torque acting on each fragment, but also on the stepsize. If the stepsize is very small, the atoms may move only very little, even though the forces and torques are still large. If you set the stepsize by hand, make sure not to choose a too small value, or use a different/additional convergence criterion. If you use an optimizer with automatic stepsize, the optimizer should prevent this from happening.

is_converged

Whether or not the convergence criterion is fulfilled.

Type:

bool

cutoff

If all atoms move less than cutoff, the criterion is fulfilled; [Å]

Type:

number

check(optimization_history)

Check if the convergence criterion is fulfilled.

If yes, self.is_converged is set to True.

Parameters:

optimization_history (list of riigid.Optimization_Step) – The history of the optimization, which shall be checked for convergence. (The optimization history is an attribute of the optimizer.)