Features#
Extract structured tables and ASE structures from engine outputs.
pint quantities end-to-end — no silent eV/kcal mix-ups.
Gradient-enhanced GPs via rgpycrumbs.surfaces, optional SurfaceFitConfig.
Call from rgpycrumbs eon plt-* with TOML --config for dense option sets.
Data flow#
flowchart LR
OUT[Engine outputs] --> PARSE[chemparseplot.parse]
PARSE --> COORDS[RMSD / energy arrays]
COORDS --> FIT{SurfaceFitConfig auto_thin?}
FIT -->|default off| FULL[Full cloud fit]
FIT -->|opt-in| THIN[Even subsample + endpoints]
FULL --> GP[rgpycrumbs.surfaces]
THIN --> GP
GP --> FIG[Publication figure]
Important
auto_thin defaults to false. Enable via SurfaceFitConfig or the
matching keys in rgpycrumbs plot TOML when dense force-eval movies make
grad_imq non-finite.
Parsing computational chemistry output files into structured data
Plotting with scientific color maps (camera-ready)
Unit preserving throughout via
pintComputation delegated to
rgpycrumbsfor surface fitting, interpolation, and structure analysisConcurrent RMSD landscape calculations via
ThreadPoolExecutor
Supported Engines#
ORCA (5.x)#
Geometry scan (
OPT) energy profilesNudged elastic band (
NEB) path visualization over the linearized reaction coordinate
eOn#
Saddle search parsing (Dimer, GPRD, LBFGS methods) with status tracking
NEB path energy profiles with RMSD landscape projections
Sella#
Saddle point optimization result parsing
ChemGP#
GP-based optimization visualization from HDF5 output (chemparseplot.plot.chemgp):
1D plotnine charts: convergence curves, RFF quality sweeps, NEB energy profiles, FPS subset scatter (PCA projection)
2D matplotlib surfaces: PES contour with NEB path overlay, GP surrogate progression panels, MAP-NLL landscape, variance overlay with hatching, trust region illustration, hyperparameter sensitivity grid
Trajectory Formats#
HDF5 trajectories from ChemGP output (pre-computed forces and reaction coordinates)
Generic ASE-readable formats (extxyz,
.traj) for NEB analysis with tangent force computation and RMSD landscape coordinateseOn ConFrame series (
parse.eon.frame_series): energies and trajectory DTOs from stamped frames for object-plot adapters.