In a world first, researchers at Australia’s national science agency, CSIRO (Commonwealth Scientific and Industrial Research Organization), have used quantum machine learning (QML) to improve the fabrication of semiconductors — a breakthrough that could transform how chips are designed and produced.
Semiconductor manufacturing is among the most intricate feats of modern engineering, requiring extreme precision and hundreds of steps, from etching to layering, just to create a single chip. At the heart of this process lies a critical challenge: accurately modeling the Ohmic contact resistance — the resistance that arises when a semiconductor meets metal, which directly impacts current flow efficiency.
Why Ohmic resistance is hard to model
Until now, engineers have relied on classical machine learning (CML) algorithms to model this resistance. But CML struggles in scenarios with small datasets and nonlinear relationships, which are common in semiconductor research. This has limited its effectiveness and left room for improvement.
The quantum leap
Led by Professor Muhammad Usman, head of quantum systems at CSIRO, the team explored an alternative: quantum machine learning (QML).
In a study published in Advanced Science, the team applied QML techniques to real-world experimental data from 159 samples of gallium nitride high-electron-mobility transistors (GaN HEMTs).
Their approach combined classical and quantum techniques in a novel way:
- They first narrowed down the many fabrication variables to only those with the most significant impact on performance.
- Next, they developed a new Quantum Kernel-Aligned Regressor (QKAR), which encoded classical data into quantum states to analyze patterns.
- Finally, a classical algorithm extracted the insights and trained on the optimized data to guide fabrication.
The QKAR technique outperformed seven different classical machine learning models tackling the same problem.
“These findings demonstrate the potential of QML for effectively handling high-dimensional, small-sample regression tasks in semiconductor domains and point to promising avenues for its deployment in future real-world applications as quantum hardware continues to mature,” the authors wrote in their paper.
Broader implications
This research could lower manufacturing costs and boost device performance in the semiconductor industry. Beyond that, it highlights how quantum technologies may soon tackle complex problems that are beyond the reach of classical computers, as quantum hardware continues to improve.
By showing that quantum machine learning can extract meaningful insights from limited, complex datasets, the CSIRO team has paved the way for a new era of smarter, faster, and more efficient semiconductor fabrication.
As quantum and classical computing converge, breakthroughs like this signal a future where even the most challenging engineering problems can be solved with unprecedented precision.