Australian researchers have achieved a world-first breakthrough by validating a quantum machine learning (QML) model for semiconductor fabrication using real-world experimental data – a step that could revolutionise the chip-making industry.
Led by CSIRO, the national science agency, the study shows that QML can outperform traditional AI in modelling Ohmic contact resistance, a complex but critical step in semiconductor manufacturing.
Published in Advanced Science, the research presents the first experimental validation of a quantum kernel method applied to semiconductor process data. The team developed a Quantum Kernel-Aligned Regressor (QKAR) that combines a Pauli-Z quantum feature map with a learnable kernel alignment layer.
Using datasets provided by Peking University, Songshan Lake Materials Laboratory, and City University of Hong Kong, CSIRO trained its model on just 159 samples of GaN HEMT devices. Despite the limited data, the quantum model outperformed seven classical AI baselines.
“Our results show that quantum models, when carefully designed, can capture patterns that classical models may miss, especially in high-dimensional, small-data regimes,” said lead author Dr Zeheng Wang.
“We validated the model by fabricating new GaN devices, which showed optimised performance, and, through quantum kernel spectrum analysis, confirmed QML’s ability to generalise beyond training data.”
Co-author Dr Tim van der Laan said the model was practical, even under real-world quantum noise conditions.
“By introducing a learnable kernel alignment layer into a shallow quantum circuit, we’ve demonstrated that useful performance gains are achievable even with limited qubit hardware,” he said.
The model retained accuracy even under simulated noise levels higher than typically found on today’s quantum systems, showing promise for near-term implementation.
Professor Muhammad Usman, team leader of CSIRO’s Quantum Systems group, said the study bridges the gap between quantum theory and industrial practice.
“This work goes beyond theory,” he said. “It offers a proof-of-concept for deploying quantum-enhanced modelling directly in semiconductor and device workflows. As quantum hardware matures, models like QKAR could enable real-time process optimisation.”
The study highlights how quantum machine learning can serve high-value industries such as nanotechnology and materials science, particularly in environments where large-scale data is unavailable.