Skip to content

Module: pathbias

Description Usage
Frozen complete-path bias functions
Authors: Pengchao Zhang used in 0 tutorialsused in 0 eggs

Details

Frozen complete-path functions

This optional module provides stateless functions for fixed complete-path biases. It does not perform learning, define a PIMD integrator, or supply a reference distribution. PATH_LOGMEANEXP provides stable replica-root aggregation of probability ratios, and PROBABILITY_MIX combines frozen path potentials with a global partition-function ratio. CONDITIONAL_PATH uses a distinct coordinate-dependent normalizer. Consult each action for its derivative, domain, synchronization and stationary-sampling contract.

The module is distributed under the GNU Lesser General Public License, version 3 or later, like PLUMED. Enable it with --enable-modules=+pathbias.

Terminology and composition

The order of averaging determines the path potential:

  • Coordinate-centroid bias: evaluate the CV on the mean bead coordinates, then evaluate the bias.
  • Bead-averaged CV bias: average the bead CVs with ENSEMBLE, then evaluate the bias.
  • Bead-averaged bias energy: evaluate one common field on each bead, then average its energies. This is the LAMMPS bead_density convention.
  • Mean probability-ratio bias: average exp(-v_b/kBT), then take -kBT times the logarithm. PATH_LOGMEANEXP supplies the stable logarithm and its derivatives; it does not average the bias energies or an estimated density.

PROBABILITY_MIX combines two frozen path potentials using a global normalizer. CONDITIONAL_PATH uses a coordinate-dependent normalizer with its own derivative. Their COUPLING parameters do not contract coordinates. LAMMPS path_contraction instead changes the geometry supplied to the CV; its intermediate values are generally not a linear mixture of endpoint CVs. These descriptions leave existing action and keyword names unchanged.

Actions

The following actions are part of this module

FUNCTION

Name Description Tags
CONDITIONAL_PATH Evaluate a fixed conditional path log mixture, with its full chain rule. FUNCTION
PATH_LOGMEANEXP Compute a stable log arithmetic mean across synchronized replicas. FUNCTION
PROBABILITY_MIX Mix two frozen path potentials with a fixed global normalization. FUNCTION