pes-analyzer 0.7.0

Fast saddle-point analysis on dense N-D PES grids

pes_analyzer

Fast saddle-point and minimum analysis on dense N-D potential energy surface grids.

Python ≥ 3.10 · Rust 2024 edition · N-D grids for N ∈ [2, 7] · MIT

Why

Quantum-chemistry calculations produce potential energy surfaces (PES) as dense tables of energies on a multidimensional grid of geometric coordinates. Once that grid exists, the interesting analysis questions are topological: where are the minima, where are the saddle points, which basins are connected to which? pes_analyzer answers those questions on grids that may be too large for pure-Python approaches by pushing the inner loops into Rust.

What it does

• pes_analyzer.saddle.find_iwf_grid — imaginary water flow (watershed) saddle search between two grid points.
• pes_analyzer.extrema.find_minima_grid — local minima on the Chebyshev king-move stencil (default 3ᴺ−1; widen via confirm_range for a fast two-pass check, or via neighborhood_range for a direct wider check).
• pes_analyzer.extrema.find_maxima_grid — strict dual of find_minima_grid. Same stencil and same neighborhood_range / confirm_range semantics; output sorted descending by energy.
• pes_analyzer.extrema.find_extrema_grid — combined single-sweep search. Returns (minima, maxima) byte-identical to calling the two single-polarity functions separately, at the cost of one extra-list allocation but one fewer stencil walk per cell.
• pes_analyzer.topology.find_watershed_segmentation — full watershed flood: labels every cell by basin and records every basin-merge (saddle) event as a merge tree. The whole-surface generalization of find_iwf_grid.
• pes_analyzer.topology merge-tree helpers — pure-Python compute_persistence, prune_merge_tree, and the traversable MergeTree (whose nodes are BasinNodes) analyse that merge tree. MergeTree is physics-free: it exposes neutral traversal, membership, and geometry primitives that a consumer composes with its own predicates to label ground states, saddles, fission exits, etc.
• pes_analyzer.grid.build_dense — scatter helper that turns sparse (coords, value) rows into a dense numpy array indexed in axis order.

Installation

The package builds from source via maturin. From a checkout:

pip install maturin
maturin develop --release

For day-to-day development (editable installs, running tests, rebuilding after Rust changes) see DEVELOPMENT.md.

Quickstart

import numpy as np

from pes_analyzer.saddle  import find_iwf_grid
from pes_analyzer.extrema import find_minima_grid

# A toy 2x5 PES: two basins at (0, 0) and (0, 4) along the top row,
# separated by a hump that peaks at (0, 2). The bottom row is a high
# wall, so any path between the basins must cross the hump.
energies = np.array([
    [0.0, 1.0, 2.0, 1.0, 0.0],
    [3.0, 3.0, 3.0, 3.0, 3.0],
])

print(find_minima_grid(energies))
# [((0, 0), 0.0), ((0, 4), 0.0)]

print(find_iwf_grid(energies, start=(0, 0), end=(0, 4)))
# ((0, 2), 2.0)

API at a glance

Function	Purpose	Reference
grid.build_dense(coords, values)	sparse rows → dense N-D array	API.md
saddle.find_iwf_grid(energies, start, end)	watershed saddle search	API.md
extrema.find_minima_grid(energies, *, neighborhood_range=1, confirm_range=None)	local minima (Chebyshev stencil)	API.md
extrema.find_maxima_grid(energies, *, neighborhood_range=1, confirm_range=None)	local maxima (dual of find_minima_grid)	API.md
extrema.find_extrema_grid(energies, *, neighborhood_range=1, confirm_range=None)	combined single-sweep search	API.md
topology.find_watershed_segmentation(energies)	full basin labelling + merge tree	API.md
topology.compute_persistence(basins, merges)	per-basin topological persistence	API.md
topology.prune_merge_tree(basins, merges, threshold)	drop low-persistence basins	API.md
topology.MergeTree(labels, basins, merges)	traversable basin merge tree (physics-free primitives)	API.md

Documentation

• API.md — full API reference with examples.
• ARCHITECTURE.md — repo layout, Python/Rust seam, GIL handling.
• ALGORITHMS.md — how the watershed and minima algorithms work.
• DEVELOPMENT.md — building, testing, common workflows.

License

MIT — see Cargo.toml.