npj Computational Materials

Quantum-annealed machine learning discovers ductile, high strength and corrosion-resistant high-entropy alloy

Diego Ibarra Hoyos, Peter F. Connors, Ho Jang, Nathan Grain, Israel Klich, Gia-Wei Chern, Peter Kaehuei Liaw, John R. Scully, S. J. Poon

Published Mar 6, 2026• 2 citations

Publisher

Abstract

Abstract Data scarcity remains a central challenge in materials discovery, where finding meaningful descriptors and tuning models for generalization is critical but inherently a discrete optimization problem prone to multiple local minima confounding the true optimal state. Classical methods often get trapped in these minima, while quantum annealing can escape them via quantum fluctuations, including tunneling, that overcome narrow energy barriers. We present a quantum-assisted machine-learning (QaML) framework that employs quantum annealing to address these combinatorial optimization challenges through feature selection, support-vector training formulated in QUBO form for classification and regression, and a new QUBO-based neural-network pruning formulation. Recursive batching enables quantum annealing to handle large feature spaces beyond current qubit limits, while quantum-pruned networks exhibit superior generalization over classical methods, suggesting that quantum annealing preferentially samples flatter, more stable regions of the loss landscape. Applied to high-entropy alloys (HEAs), a data-limited but compositionally complex testbed, the framework integrates models for fracture-strain classification and yield-strength regression under physics-based constraints. The framework identified and experimentally validated Al8Cr38Fe50Mn2Ti2 (at.%), a single-phase BCC alloy exhibiting a 0.2% yield strength of 568 MPa, greater than 40% compressive strain without fracture, and a critical current density in reducing acid nearly an order of magnitude lower than 304 stainless steel. These results establish QA as a practical route to overcome classical optimization limits and accelerate materials discovery.

DOI: 10.1038/s41524-026-02032-x

Related papers

From Quantum Annealing to Alloy Discovery: Towards Accelerated Design of High-Entropy Alloys

Ibarra-Hoyos, Diego, Connors, Peter, Jang, Ho +6 more

Data scarcity remains a central challenge in materials discovery, where finding meaningful descriptors and tuning models for generalization is critical but inherently a discrete optimization problem prone to multiple local minima confounding the true optimal state. Classical methods often get trapped in these minima, while quantum annealing can escape them via quantum fluctuations, including tunneling, that overcome narrow energy barriers. We present a quantum-assisted machine-learning (QaML) framework that employs quantum annealing to address these combinatorial optimization challenges through feature selection, support-vector training formulated in QUBO form for classification and regression, and a new QUBO-based neural-network pruning formulation. Recursive batching enables quantum annealing to handle large feature spaces beyond current qubit limits, while quantum-pruned networks exhibit superior generalization over classical methods, suggesting that quantum annealing preferentially samples flatter, more stable regions of the loss landscape. Applied to high-entropy alloys (HEAs), a data-limited but compositionally complex testbed, the framework integrates models for fracture-strain classification and yield-strength regression under physics-based constraints. The framework identified and experimentally validated Al8Cr38Fe50Mn2Ti2 (at.%), a single-phase BCC alloy exhibiting a 0.2 % yield strength of 568 MPa, greater than 40 % compressive strain without fracture, and a critical current density in reducing acid nearly an order of magnitude lower than 304 stainless steel. These results establish QA as a practical route to overcome classical optimization limits and accelerate materials discovery.

arXiv:2511.05750Nov 7, 20250 citations

Quantum Annealing-Driven Deep Learning Model Architecture Design

Jincheng Zhang

This paper proposes a novel approach to deep learning model architecture design by leveraging the principles of quantum annealing. The core idea is to map the complex search space of deep learning architectures to the energy landscape of a quantum annealing system. The quantum annealing algorithm is then employed to explore this landscape, iteratively minimizing the energy and ultimately identifying a globally optimal or near-optimal architecture. We introduce a framework integrating deep learning architecture search with quantum annealing, addressing the limitations of classical optimization methods prone to getting trapped in local minima. The method's effectiveness is demonstrated conceptually through a detailed architecture design process, highlighting the potential for accelerated discovery of superior models. The key contributions include a methodology for translating architectural parameters into a quantum annealing problem formulation and a strategy for interpreting the resulting quantum annealing solution as a viable deep learning architecture. Further research will focus on optimizing the annealing process and exploring different quantum annealing hardware platforms.

Aug 29, 20260 citations

QUBO-Driven Quantum Annealing: A Unified Cloud-Scale Framework for Resource Forecasting, Classification, and Deep Learning Training

Yulian Song, Guozheng Zhao, Nan Wang +3 more

This paper proposes an integrated QUBO-based quantum-annealing optimization framework for next-generation cloud computing. First, for resource-demand forecasting, we blend Autoregressive (AR) models with Quadratic Unconstrained Binary Optimization (QUBO) and solve the resulting problem via Simulated Annealing; monthly-scale tests reveal superior trend-capture accuracy compared with classical predictors. Second, we embed quantum optimization into Support Vector Machines by casting kernel selection and penalty-parameter tuning as QUBO instances; evaluations on the Iris benchmark—and on synthetically enlarged variants—show marked gains in both classification accuracy and training speed, especially under large-data regimes. Third, we extend the same paradigm to deep learning: a QUBO-formulated quantum-annealing routine optimizes weight initialization and layer-wise learning rates in convolutional networks for image classification, yielding higher top-1 accuracy and reduced wall-clock training time. Across all tasks the hybrid quantum-classical solver consistently outperforms its classical counterparts while maintaining linear scalability in cloud environments. These results show the practical potential of quantum - enhanced algorithms for resource prediction, classification, and deep learning, and pave the way for wider use of quantum technologies in cloud - based AI services.

Applied and Computational EngineeringAug 13, 20250 citations

A Hybrid Quantum Classical Framework for Enhanced Machine Learning Performance on High Dimensional Data

Nnaemeka Kingsley Ugwumba

Abstract This research presents a comprehensive hybrid quantum classical machine learning framework designed to overcome the computational limitations of classical algorithms when processing high dimensional data. The study integrates parameterized quantum circuits within classical neural network architectures and demonstrates their application across optimization tasks and supervised learning problems. The framework was evaluated on combinatorial optimization benchmarks including Max Cut and Job Shop Scheduling problems alongside classification tasks using quantum annealing based feature selection. Results demonstrate that hybrid quantum classical models achieve faster convergence rates and improved accuracy in high dimensional spaces compared to classical counterparts with the quantum algorithm for Max Cut optimization outperforming complete solution searches on larger problem instances. The findings provide machine learning practitioners and quantum computing researchers with validated hybrid architectures for near term quantum device deployment.

Mar 13, 20260 citations

Quantum Annealing for Combinatorial Optimization of Neural Network Training

Jincheng Zhang

This paper investigates the application of quantum annealing to the combinatorial optimization problem inherent in neural network training, specifically focusing on hyperparameter tuning. Traditional methods for hyperparameter optimization, such as grid search and random search, can be computationally expensive and often fail to efficiently explore the vast hyperparameter space. We propose a novel approach that maps the hyperparameter optimization task onto a quantum annealing landscape. The quantum annealer, leveraging quantum tunneling, is then utilized to navigate this landscape and identify optimal hyperparameter values. This approach aims to accelerate the training process and enhance the performance of neural networks. The core concept involves formulating the hyperparameter tuning problem as an Ising model, which can be directly implemented on quantum annealing hardware. We explore the theoretical underpinnings of this mapping and discuss the potential advantages and limitations of this method. The results, although not based on experimental validation due to the nature of the research, demonstrate a promising trajectory for integrating quantum computing with machine learning. Further research and development are required to fully realize the potential of this approach.

Sep 3, 20260 citations

Empirical Study of Quantum Machine Learning Algorithms for Accelerating Functional Materials Discovery

Doris Ngozi Nwachuku, Ericdavid E. Ugochukwu

Quantum machine learning (QML) represents a convergence of quantum computing and statistical learning theory with transformative potential for high-throughput computational materials discovery. This empirical study systematically benchmarks six QML architectures — quantum kernel support vector machines, variational quantum eigensolvers augmented by graph neural networks, quantum approximate optimization algorithm hybrids, quantum random forests, quantum Boltzmann machines, and parameterized quantum circuit-based neural networks — against their classical analogues across five material classes comprising perovskite oxides, battery cathode materials, two-dimensional semiconductor alloys, metal-organic frameworks, and high-entropy alloys. Using 14,720 DFT-computed property records drawn from the Materials Project and AFLOW databases as the common evaluation corpus, the QML methods achieved mean absolute error reductions of 38–61% and training time compressions of 29–61% relative to equivalent classical estimators at equivalent dataset sizes. The quantum random forest architecture exhibited the strongest overall performance (MAE = 0.026 eV; R² = 0.978), while hardware-native implementations on IBM Quantum Eagle and IonQ Forte achieved practical speed-ups of 28x–41x over CPU-based baselines. Barren plateau pathology and noise-induced gradient suppression remained the principal barriers to scaling beyond 50 qubits on current NISQ hardware. The empirical results of this study established principal conclusions that advance the theoretical and practical understanding of QML in materials discovery contexts. These findings provide grounding for deploying QML in production materials informatics pipelines.

Nigerian Journal of PhysicsJun 19, 20260 citations