pbctools

A lightweight Python package for periodic boundary condition calculations, neighbor analysis, and molecular recognition in trajectory data.

Features

• PBC Distance Calculation: Compute distance vectors between atom sets across multiple frames
• Nearest Neighbor Detection: Find nearest neighbors with PBC support for trajectory data
• Molecule Recognition: Identify molecular species using bond detection algorithms
• High Performance: C++ backend with OpenMP acceleration
• Easy Integration: Simple NumPy-based Python API

Installation

Install from source:

git clone https://github.com/jhaens/pbctools.git
cd pbctools
pip install .

Installation via pip (PyPI):

pip install pbctools

Quick Start

import numpy as np
from pbctools import pbc_dist, next_neighbor, molecule_recognition
from ase.io import read, write

# OPTION 1
# Load single trajectory frames data
coords1 = read('coord1.xyz')
coords1.set_cell(np.loadtxt('pbc1.txt'))
coords2 = read('coord2.xyz')
coords2.set_cell(np.loadtxt('pbc2.txt'))

# Calculate distance vectors between all atom pairs (ASE objects auto-extract PBC)
distances = pbc_dist(coords1, coords2)
print(f"Distance shape: {distances.shape}")  # (1, n_atoms1, n_atoms2, 3) for single-frame inputs

# Find nearest neighbors
indices, nn_dists = next_neighbor(coords1, coords2)
print(indices.shape, nn_dists.shape)  # (1, n_atoms1), (1, n_atoms1) for single-frame inputs

# Analyze molecular composition (single frame)
molecules = molecule_recognition(coords1)
print(f"Found molecules: {molecules}")  # e.g., {'H2O': 100}

# OPTION 2
# Load trajectory data
coords1 = read('traj1.xyz', index=':')
coords2 = read('traj2.xyz', index=':')
pbc = np.loadtxt('pbc.txt')
for frame in coords1:
	frame.set_cell(np.array(pbc))

# ...

API Reference

pbc_dist(coord1, coord2=None, pbc=None)

Calculate periodic boundary condition distance vectors.

Supports flexible inputs:

• ASE Atoms (single frame) or list of Atoms (trajectory) — PBC auto-extracted
• NumPy arrays

Parameters:

• coord1: ASE Atoms | list[Atoms] | np.ndarray
  • If ndarray: shape (n_frames, n_atoms1, 3) or (n_atoms1, 3) for single-frame
• coord2: ASE Atoms | list[Atoms] | np.ndarray | None (default: coord1)
  • If ndarray: shape (n_frames, n_atoms2, 3) or (n_atoms2, 3)
• pbc: np.ndarray | None
  • Required for ndarray inputs; shape (3, 3). Ignored when using ASE objects (read from Atoms).

Returns:

• np.ndarray with shape (n_frames, n_atoms1, n_atoms2, 3)
  • For single-frame inputs, n_frames = 1

Notes:

• To compute within-set distances, pass the same object/array for coord1 and coord2.

next_neighbor(coord1, coord2=None, pbc=None)

Find nearest neighbors between two atom sets (PBC-aware).

Parameters:

• coord1: ASE Atoms | list[Atoms] | np.ndarray (…, 3)
• coord2: ASE Atoms | list[Atoms] | np.ndarray (…, 3) | None (default: coord1)
• pbc: np.ndarray (3, 3) | None

Returns:

• indices: np.ndarray of int with shape (n_frames, n_atoms1)
• distances: np.ndarray of float32 with shape (n_frames, n_atoms1)

molecule_recognition(coords, atoms=None, pbc=None)

Identify molecular species in a single frame via bond detection.

Parameters:

• coords: ASE Atoms | np.ndarray, shape (n_atoms, 3)
• atoms: list[str] | np.ndarray[str] | None — required for ndarray input; ignored for ASE Atoms
• pbc: np.ndarray (3, 3) | None — required for ndarray input; ignored for ASE Atoms

Returns:

• dict[str, int] — molecular formulas and counts

Performance

pbctools is optimized for large trajectory analysis:

• Multi-threaded C++ backend
• Support for both orthogonal and triclinic unit cells
• Optimized distance calculations with PBC

License

MIT License - see LICENSE file for details.