Quantum-Annealing Enhanced Machine Learning for Interpretable Phase Classification of High-Entropy Alloys

TL;DR

This paper introduces a quantum-enhanced machine learning framework that improves phase classification accuracy and speed for high-entropy alloys by leveraging quantum annealing for interpretable and nonlinear modeling.

Contribution

It presents a novel integration of quantum annealing with machine learning for materials science, enabling faster and more accurate phase prediction in complex alloys.

Findings

Quantum classifiers match or outperform classical models in accuracy.

Quantum annealing reduces training time significantly.

The approach provides interpretable insights into alloy phase behavior.

Abstract

High entropy alloys (HEAs) offer unprecedented compositional flexibility for designing advanced materials, yet predicting their crystallographic phases remains a key bottleneck due to limited data and complex phase formation behavior. Here, we present a quantum-enhanced machine learning framework that leverages quantum annealing to enhance phase classification in HEAs. Our pipeline integrates Quantum Boosting (QBoost) for interpretable feature selection and classification, with Quantum Support Vector Machines (QSVM) that use quantum-enhanced kernels to capture nonlinear relationships between physical descriptors. By reformulating both models as Quadratic Unconstrained Binary Optimization (QUBO) problems, we exploit the efficient sampling capabilities of quantum annealers to achieve rapid training and robust generalization, demonstrating notable runtime reductions relative to classical baselines in our setup. We target six key phases: FCC, BCC, Sigma, Laves, Heusler, and AlXY B2, and benchmark model performance using both cross-validation and a rigorously curated test set of prior experimentally synthesized HEAs. The results confirm strong alignment between predicted and measured phases. Our findings demonstrate that quantum-enhanced classifiers match or exceed classical models in accuracy and offer insights grounded in interpretable physical descriptors. This work constitutes an important step toward practical quantum acceleration in materials discovery pipelines.