Researchers leverage quantum machine learning to fabricate semiconductors for the first time, potentially revolutionizing the production process of chips.

In a groundbreaking development, a team of scientists in Australia has created a quantum machine learning technique that could change the way microchips are made. Led by Professor Michael J. Taghavi from the University of Sydney, the research group published their findings in the journal Advanced Science in June 2021.

The new method, known as Quantum-Machine Nerve, is a blend of artificial intelligence (AI) and quantum computing principles. It takes classical data and encodes it in quantum states, allowing quantum computing systems to process complex mathematical relationships much faster than classical systems.

The heart of this innovation is the Quantum Kernel-Aligned Regressor (QKAR), a new machine learning architecture developed by the team. QKAR converts classical data into quantum states and enables the quantum system to identify complex relationships in the data. A classical system then takes over to interpret the results or apply them.

The researchers focused on modeling Ohmic contact resistance, a particularly difficult challenge in chipmaking. By identifying fabrication variables that had the biggest impact on Ohmic contact resistance and narrowing down the dataset to the most relevant inputs, they were able to outperform seven leading classical models, including deep learning and gradient boosting methods.

Semiconductor fabrication is a complex, multistep process that requires painstaking precision. It involves stacking and sculpting hundreds of microscopic layers onto a silicon wafer. Each step in the process must be performed with extreme accuracy. The new model could soon be applied to real-world chip production, particularly as quantum hardware continues to evolve.

The process begins with a photoresist coating, which applies a light-sensitive material that enables precise patterning. Deposition layers then thin films of material onto the wafer. These layers are crucial for determining the electrical properties of the chip.

The researchers demonstrated that quantum machine learning algorithms can significantly improve the process of modeling the electrical resistance inside a chip. This technique, which uses qubits that can exist in multiple states simultaneously, could potentially revolutionise the semiconductor industry.

QKAR was designed to be compatible with real-world hardware, meaning it could be deployed on quantum machines as they become more reliable. This development could mark a significant step forward in the application of quantum machine learning (QML) for effectively handling high-dimensional, small-sample regression tasks in semiconductor domains.

The study serves as a testament to the potential of QML and its ability to tackle complex problems in various fields. As quantum hardware continues to advance, we can expect to see more innovations like this one that leverage the power of QML to drive progress and solve previously intractable problems.