In a groundbreaking scientific advancement, researchers from the Commonwealth Scientific and Industrial Research Organisation (CSIRO), Australia’s national research agency, have announced the development of the world’s first semiconductors designed using quantum technologies.اضافة اعلان

This new approach, which relies on quantum machine learning (QML), opens unprecedented possibilities in semiconductor manufacturing, potentially revolutionizing the future of electronic devices.

Unlike classical machine learning (CML) techniques currently used in chip design, QML offers significant improvements in the efficiency and accuracy of semiconductor design.

Traditionally, semiconductor design has involved collaboration among experts in electrical engineering, materials science, and computing, all working together to create the integrated circuits that power today’s technology. This process has required extremely precise testing and stringent standards, according to Interesting Engineering.

With the evolution of AI and computing technologies, semiconductor design has already seen notable enhancements through classical machine learning. However, as systems become more complex and datasets shrink, CML has begun to show limitations—prompting CSIRO researchers to explore QML as a more advanced solution.

Tackling Semiconductor Design Challenges: Ohmic Resistance
The research team, led by Mohamed Osman, Head of Quantum Systems at CSIRO, focused on modeling ohmic resistance—a critical factor that determines how easily electric current flows through conductors when in contact with metals. Modeling this resistance accurately has long been a major challenge in improving chip design.

Using experimental data from 159 samples of GaN HEMT transistors—a more efficient alternative to traditional silicon-based semiconductors—the team initially evaluated 37 parameters and successfully narrowed them down to the five most influential ones.

How Quantum Machine Learning Works
Osman explained that given the limited computational power of current quantum devices, the team had to simplify their quantum model. To achieve this, they developed the Quantum Kernel-Aligned Regressor (QKAR), which converts classical data into quantum data using just five qubits.

QKAR extracts the most relevant features from the data, which are then processed by a classical algorithm to interpret results and guide manufacturing improvements. This approach allowed researchers to identify the key parameters affecting production quality and determine how to enhance performance.

Quantum Surpasses Classical AI
Tests showed that QKAR outperformed seven different classical machine learning algorithms in addressing this complex problem. Most notably, QKAR required only five qubits—making it feasible for immediate application in the industry without the need for large-scale quantum computing systems.

Zheng Wang, a CSIRO researcher, noted that quantum models can detect patterns missed by classical systems, especially in high-dimensional or small datasets. The model’s accuracy was validated by producing new GaN devices, which showed improved performance.

Promising Results
The study, published in the journal Advanced Science, demonstrates that quantum machine learning could represent a major leap forward in semiconductor design and manufacturing—and may pave the way for a revolution in the future of electronics.

Source: Al-Bayan