Semiconductor spin qubits are a promising candidate for the building blocks of next-generation quantum computers due to their high potential for integration and compatibility with existing semiconductor technologies. Qubits - like the 0s and 1s of a traditional computer - serve as a basic unit of information for quantum computers. However, the practical realization of these computers requires a massive number of qubits, making the development of more efficient adjustment methods a critical challenge for the field.

A research group including Yui Muto from Tohoku University's Graduate School of Engineering, Assistant Professor Motoya Shinozaki and Associate Professor Tomohiro Otsuka from the Advanced Institute for Materials Research (WPI-AIMR), and their colleagues have successfully demonstrated a method that may help make this massive number of qubits much more manageable, moving us one step closer towards scaling up quantum computing.

A key challenge in scaling up is that reading information from these quantum dot systems requires researchers to manually identify the precise angles and positions of lines on measurements called charge stability diagrams. To address this issue, the researchers developed a method that automatically extracts charge transition lines from charge stability diagrams by utilizing U-Net, a sophisticated artificial intelligence (AI) model. "

As technology advances, future quantum computers will require an immense number of qubits, and adjusting each one by hand will simply be difficult," says Associate Professor Tomohiro Otsuka. "Our research leverages machine learning to automate the identification of charge transition lines and the definition of virtual gates. This allows us to determine single-electron regions with high efficiency, providing an essential tool for scaling up quantum dot systems."

By analyzing the data extracted by the AI using image processing and clustering techniques, the researchers demonstrated that it is possible to automate the configurations required for large-scale quantum dots. This breakthrough is expected to handle a vast number of qubits far beyond human capability, paving the way for the realization of practical, large-scale quantum computers.

Looking ahead, the research group aims to further refine this AI-driven approach. "We have demonstrated an automated tuning approach using this method," adds Otsuka. "Our next goal is to demonstrate the automatic adjustment of even larger arrays of spin qubits, contributing directly to the global effort to build powerful quantum computing systems."

The findings were published online in the scientific journal Scientific Reports on February 14, 2026.