Rust-accelerated powder X-ray diffraction simulation for high-throughput and machine-learning-driven materials...

Area of Science:

• Computational Materials Science
• Crystallography
• Software Development

Background:

• High-throughput powder X-ray diffraction (XRD) simulations are crucial for creating datasets for machine learning in materials analysis.
• The existing Python-based pymatgen powder XRD calculator can be computationally inefficient, limiting the scale of these simulations.
• There is a need for accelerated computational methods to improve the efficiency of XRD data generation.

Purpose of the Study:

• To develop a Rust-accelerated implementation of the pymatgen powder XRD calculator, named XRD-Rust.
• To maintain compatibility with existing Python-based workflows while significantly improving computational performance.
• To enable efficient high-throughput dataset generation for machine learning applications and enhance interactive diffraction analysis.

Main Methods:

• Reimplemented computationally intensive parts of the pymatgen powder XRD calculation in the Rust programming language.
• Integrated Rust code into existing Python workflows, retaining pymatgen for crystal structure handling.
• Utilized optional acceleration techniques including single instruction, multiple data (SIMD) vectorization and multi-threaded execution (Rayon library).

Main Results:

• XRD-Rust demonstrated substantial speedups on large crystallographic datasets (MC3D and COD).
• Median speedups ranged from 10.7× (serial SIMD) to 19.5× (eight-thread parallel execution) on the Crystallography Open Database.
• Runtime reductions were significant, e.g., from 1437 minutes to 1 minute for the COD dataset with parallel execution.

Conclusions:

• XRD-Rust offers a significant performance improvement over the original pymatgen implementation for powder XRD simulations.
• The acceleration enables more efficient generation of large datasets required for machine learning models in materials science.
• XRD-Rust enhances the feasibility of high-throughput computational materials analysis and interactive diffraction applications.