crystract

crystract provides a suite of functions to parse Crystallographic Information Files (.cif), extracting essential data such as chemical formulas, unit cell parameters, atomic coordinates, and symmetry operations. It also includes tools to calculate interatomic distances, identify bonded pairs using various algorithms (Minimum Distance, Brunner’s, Hoppe’s, Voronoi, CrystalNN), determine nearest neighbor counts, and calculate bond angles. All data is extracted into nested data.tables, which can then be exported as an R Data Structure (RDS) or folders of .csv files. The package is designed to facilitate the preparation of crystallographic data for further analysis, including machine learning applications in materials science.

Note on Repository Structure

The crystract package is located within the packages/crystract/ subdirectory of the PrabhuLab/ml-crystals GitHub repository. You must use the subdir argument during installation, as shown below.

Key Features

Efficient CIF Parsing: Utilizes data.table for fast and robust extraction of metadata, unit cell parameters, atomic coordinates, and symmetry operations.
Symmetry and Supercell Generation: Applies symmetry operations to generate a full unit cell from the asymmetric unit and expands coordinates into a 3x3x3 supercell for neighbor searching.
Geometric Calculations: Computes interatomic distances using the metric tensor (correct for all crystal systems) and calculates bond angles.
Multiple Bonding Algorithms: Implements several algorithms to identify bonded atoms, including the minimum_distance (default), brunner, econ (Hoppe’s), voronoi, and crystal_nn methods.
Rigorous Error Propagation: Calculates and propagates experimental uncertainties from the CIF file into the final calculated bond lengths and angles.
Powerful Post-Processing Tools: Includes functions to filter results by chemical element, Wyckoff site, or to remove non-physical “ghost” distances caused by site disorder using a customizable atomic radii table.
Batch Processing & Export: The main analyze_cif_files() function is designed to process hundreds of files in a single run, and results can be easily exported to a structured directory of CSV files with export_analysis_to_csv().

Workflow Overview

The following diagram illustrates the primary data pipeline in crystract, from raw CIF input to final CSV export.

crystract Workflow Diagram

Best Practices & Decision Logic

To assist researchers in configuring crystract for their specific datasets, we provide the following decision trees for selecting atomic radii and choosing the most appropriate bonding algorithm.

1. Choosing a Bonding Algorithm

When invoking analyze_cif_files(..., bonding_algorithms = c(...)), we recommend choosing your target algorithm based on the chemical makeup of your structure and the electronegativity differences (\(\Delta EN\)) of the expected bonds.

Bonding Algorithm Selection Logic

2. Atomic Radii Selection

When utilizing functions like filter_ghost_distances() or algorithms that rely on distance cutoffs, crystract employs an internal logic to select the most appropriate atomic radius. You can also override this by injecting your own custom radii dictionary via set_radii_data().

Atomic Radii Selection Logic

Quickstart: A Complete Workflow & Real Outputs

The analyze_single_cif() (and its batch counterpart analyze_cif_files()) provides a complete, one-step workflow. Here we run it on an example crystal structure included inside the package itself, demonstrating the exact data outputs you can expect.

library(crystract)
library(data.table)

# 1. Load the built-in demo CIF file (Strontium Silicide)
cif_path <- system.file("extdata", "1590946.cif", package = "crystract")

# 2. Analyze the file
# This single function handles parsing, supercell expansion, geometric calculations, 
# bonding detection, and error propagation.
analysis_results <- analyze_single_cif(
  cif_path,
  bonding_algorithms = c("minimum_distance", "crystal_nn")
)

1. Extracted Metadata

The returned object is a single row data.table containing both high-level metadata and list-columns storing the detailed extracted measurements.

database_code	chemical_formula	space_group_name
depnum_ccdc_archive CCDC 1590946	Si1 Sr2	P n m a

High-Level Crystal Information

2. Unit Cell Metrics

The parameters defining the size and shape of the unit cell are securely parsed along with their experimental uncertainties (if available in the CIF).

_cell_length_a	_cell_length_b	_cell_length_c	_cell_angle_alpha	_cell_angle_beta	_cell_angle_gamma
8.11	5.15	9.54	90	90	90

Extracted Unit Cell Parameters (Å and Degrees)

3. Detected Bonded Pairs

crystract identifies bonded pairs using your chosen algorithms. Below is the output from the Minimum Distance algorithm. Notice the rigorous propagation of experimental error (DistanceError).

Atom1	Atom2	Distance	Weight
Si1	Sr1_1_0_0_0	3.163544	1.0000000
Si1	Sr1_2_0_0_0	3.245310	0.9748050
Si1	Sr1_4_0_-1_-1	3.184477	0.9934267
Si1	Sr1_4_0_0_-1	3.184477	0.9934267
Si1	Sr2_1_0_0_-1	3.261366	0.9700058
Si1	Sr2_3_-1_-1_0	3.465249	0.9129342

Predicted Bonded Pairs (Minimum Distance Method)

4. Calculated Bond Angles

Using the metric tensor, all connected triplets are evaluated to calculate the exact internal bond angles across the repeating periodic boundaries.

CentralAtom	Neighbor1	Neighbor2	Angle
Si1	Sr1_1_0_0_0	Sr1_2_0_0_0	109.37260
Si1	Sr1_1_0_0_0	Sr1_4_0_-1_-1	125.55190
Si1	Sr1_1_0_0_0	Sr1_4_0_0_-1	125.55190
Si1	Sr1_1_0_0_0	Sr2_1_0_0_-1	129.28796
Si1	Sr1_1_0_0_0	Sr2_3_-1_-1_0	69.08689
Si1	Sr1_1_0_0_0	Sr2_3_-1_0_0	69.08689

Calculated Interatomic Angles

Data Dictionary

Here is a comprehensive overview of the columns generated in the Master Analysis Object and its nested tables.

1. Master Analysis Object

Column Name	Data Type	Description
`file_name`	Character	The name of the processed CIF file.
`database_code`	Character	The unique identifier from the source database.
`chemical_formula`	Character	The chemical sum formula extracted from the CIF.
`structure_type`	Character	The name of the structure type.
`space_group_name`	Character	Hermann-Mauguin space group symbol.
`space_group_number`	Character	International Tables space group number.
`unit_cell_metrics`	List (DT)	Nested table containing lattice parameters.
`atomic_coordinates`	List (DT)	Nested table of primary asymmetric atoms.
`symmetry_operations`	List (DT)	Nested table of symmetry operators.
`transformed_coords`	List (DT)	Nested table of the full unit cell atoms.
`expanded_coords`	List (DT)	Nested table of the supercell (3x3x3) atoms.
`distances`	List (DT)	Nested table of all calculated interatomic distances.
`bonded_pairs_*`	List (DT)	Nested table of bonds detected via requested methods (e.g. `_minimum_distance`).
`neighbor_counts_*`	List (DT)	Nested table of coordination numbers for requested methods.
`bond_angles_*`	List (DT)	Nested table of calculated bond angles for requested methods.

2. Atomic Coordinates (Nested)

Column Name	Data Type	Description
`Label`	Character	Unique atom label (e.g., “Fe1”).
`WyckoffSymbol`	Character	The Wyckoff letter (e.g., “c”).
`WyckoffMultiplicity`	Numeric	The site multiplicity (e.g., 4).
`Occupancy`	Numeric	Site occupancy factor (0.0 to 1.0).
`x_a`, `y_b`, `z_c`	Numeric	Fractional coordinates along axis \(a, b, c\).
`*_error`	Numeric	Standard uncertainties for coordinates.

3. Bonded Pairs (Nested)

Column Name	Data Type	Description
`Atom1`	Character	Label of the central atom (from the asymmetric unit).
`Atom2`	Character	Label of the neighbor atom (from the expanded supercell).
`Distance`	Numeric	Calculated Euclidean distance in Angstroms (Å).
`DistanceError`	Numeric	Propagated standard uncertainty of the distance.
`DeltaX`, `DeltaY`, `DeltaZ`	Numeric	Difference in fractional coordinates (\(x_1 - x_2\)).
`Weight`	Numeric	Calculated bond weight/strength depending on the algorithm.

Licensing

crystract is offered under a dual-license model to accommodate a variety of use cases:

For Open-Source Projects: The package is licensed under the GNU General Public License v3.0 (GPL-3.0). If you are developing other open-source software, you are free to use, modify, and distribute crystract under the terms of the GPL-3.0.
For Commercial Use: If you wish to use crystract in a commercial product, for commercial services, or for any other commercial purpose, you must obtain a separate commercial license. Please contact the package maintainer to arrange the terms.

Installation

Installing crystract involves a few steps, as it is currently hosted on GitHub. We use the remotes package to facilitate installation directly from the repository.

Prerequisites

Install the latest version of R.
Install the free RStudio Desktop IDE.

Installation Steps

Open R or RStudio and run the following commands:

# First, ensure you have the remotes package
install.packages("remotes")

# Install crystract from the GitHub repository
remotes::install_github("PrabhuLab/ml-crystals", subdir = "packages/crystract", build_vignettes = TRUE)

Verifying the Installation

To make sure the package was installed correctly, load it into your R session.

library(crystract)

If this command runs without any errors, the installation was successful.

Learning More

For a detailed, step-by-step guide explaining each function, the crystallographic principles, and the formulas used for calculations, please see the package vignette.

You can access it with the following command after you have successfully installed the package:

# This command opens the detailed package guide
vignette("crystract")

Community Guidelines

We welcome and appreciate all forms of community engagement. To ensure a smooth and productive collaboration, we have established guidelines for contributing, reporting issues, and seeking support.

All participants in this project are expected to abide by our Code of Conduct. Please read it to understand the standards of behavior we expect.

For detailed instructions on how to contribute to the software, report bugs, or suggest new features, please review our Contributing Guidelines.

Author

Author and Maintainer: Don Ngo (dngo@carnegiescience.edu)