Chinese Researchers Develop Ultra-Low Power Advanced Transistor

Select Language:

A team of Chinese researchers has created the world’s smallest ferroelectric transistor that consumes ultralow power, shedding new light on advancements in the semiconductor industry, according to a recent publication in Science Advances.

In cutting-edge semiconductor manufacturing, logic chips now operate at voltages as low as 0.7 volts to maximize energy efficiency. However, traditional non-volatile memory devices like NAND flash typically require at least 5 volts for write operations.

This voltage discrepancy has historically necessitated complex circuits to step voltage up or down, allowing logic units and memory to interact. Such solutions add to power usage, occupy more space, and create data transfer delays between components.

In typical AI chips, a significant portion—between 60 to 90 percent—of total energy is spent on data transfer rather than computation, making this a major barrier to improving AI processing power and energy efficiency.

Led by senior researcher Qiu Chenguang and academician Peng Lianmao from Peking University, the team developed nano-gate ferroelectric transistors that operate at an ultralow voltage of just 0.6 volts. They successfully miniaturized the gate to a mere 1 nanometer in size.

Reviewers of the publication highlight that these nano-gate ferroelectric devices demonstrate excellent memory performance, achieving for the first time voltage compatibility between ferroelectric memory and logic transistors. This breakthrough has profound implications for memory technology development.

Qiu explained that their research addresses the longstanding voltage mismatch between memory and logic, enabling data transfer at the same low voltage with zero barriers and ultra-low power consumption for rapid interaction.

He further noted that the underlying principles of this technology are widely applicable to common ferroelectric materials and can be manufactured at scale using standard industrial methods, demonstrating strong compatibility with existing production processes.

This innovation is expected to find future applications in large-scale model inference, edge computing, wearable technology, and Internet of Things devices.