ezff.errors - Error calculation modules

This module provide functions for computing errors from previously completed MD runs

ezff.errors.error_atomic_charges(MD=None, GT=None)[source]

Calculate error due to difference between MD-computed atomic charges and ground-truth atomic charges

Parameters:
  • MD (xtal.AtTraj object) – Relaxed structure after MD run
  • GT (xtal.AtTraj object) – Initial Ground-Truth structure used as input for MD calculations
ezff.errors.error_energy(MD, GT, weights='uniform', verbose=False)[source]

Calculate error between MD-computed potential energy surface and the ground-truth potential energy surface with user-defined weighting schemes

Parameters:
  • md_disp (1D np.array) – MD-computed potential energy surface
  • gt_disp (1D np.array) – Ground-truth potential energy surface
  • weights (str or list) – User-defined weighting scheme for calculating errors. Possible values are uniform - where errors from all points on the PES are weighted equally, minima - where errors from lower-energy points are assigned greater weights, dissociation - where errors from highest-energy points are assigned greater weights, and list - 1-D list of length equal to number of points on the PES scans
  • verbose (bool) – Deprecated option for verbosity of error calculation routine
ezff.errors.error_lattice_constant(MD=None, GT=None)[source]

Calculate error due to optimization of lattice constants in the initial structure.

Parameters:
  • MD (xtal.AtTraj object) – Relaxed structure after MD run
  • GT (xtal.AtTraj object) – Initial Ground-Truth structure used as input for MD calculations
ezff.errors.error_phonon_dispersion(MD=None, GT=None, weights='uniform', verbose=False)[source]

Calculate error between MD-computed phonon dispersion and the ground-truth phonon dispersion with user-defined weighting schemes

Parameters:
  • MD (2D np.array) – MD-computed phonon dispersion curve
  • GT (2D np.array) – Ground-truth phonon dispersion curve
  • weights (str or list) – User-defined weighting scheme for calculating errors provided as a list of numbers, one per band. Possible values are uniform - where errors from all bands are equally weighted, acoustic - where errors from lower-frequency bands are assigned greater weights, and list - 1-D list of length equal to number of bands
  • verbose (bool) – Deprecated option for verbosity of error calculation routine
ezff.errors.error_structure_distortion(MD=None, GT=None)[source]

Calculate error due to relaxation of atoms in the initial structure. The error is the sum of root mean square displacement of atoms.

Parameters:
  • MD (xtal.AtTraj object) – Relaxed structure after MD run
  • GT (xtal.AtTraj object) – Initial Ground-Truth structure used as input for MD calculations