A research team led by Professor Il Jeon of the Department of Nano Engineering and the Sungkyunkwan Advanced Institute of Nanotechnology (SAINT) at Sungkyunkwan University, including Dr. Sihyeok Kim and Dr. Jang Woo Lee, has developed a next-generation artificial intelligence semiconductor device capable of independently implementing long-term and short-term memory within a single device using surface acoustic waves (SAWs).

The research team proposed a new concept of a reconfigurable artificial synapse that overcomes the limitations of conventional memristors by selectively controlling long-term and short-term memory through electrical signals and surface acoustic waves, respectively. The developed technology is expected to contribute to the realization of ultra-low-power neuromorphic computing and next-generation AI semiconductor devices.

As the performance of artificial intelligence continues to improve, the amount of power required for data processing is also increasing rapidly. Neuromorphic computing, which performs memory and computation simultaneously in a manner similar to the human brain, has therefore attracted considerable attention as a next-generation computing technology.

Artificial synapses capable of implementing both long-term and short-term memory are essential for neuromorphic computing. However, conventional memristors generally rely only on electrical stimulation to control memory behavior. Repeated electrical stimulation can cause device degradation and reduced reliability, while also making it difficult to independently control long-term and short-term memory.

To address these limitations, the research team employed surface acoustic waves, which are mechanical waves that propagate along the surface of a solid, as a new control signal. By integrating a monolayer molybdenum disulfide (MoS2) memristor and a SAW device onto a single platform, the researchers designed the system so that electrical signals were responsible for forming long-term memory, while SAWs were used to control short-term memory through a non-contact mechanism.

This configuration enabled the team to realize a reconfigurable artificial synapse in which previously stored long-term memory remained intact, while short-term memory could be selectively generated and erased.
The researchers reproduced biological short-term synaptic plasticity by adjusting the intensity, pulse width, and interval of the SAWs. They also confirmed that short-term memory could be repeatedly controlled without damaging electrically stored long-term memory.

In addition, the device maintained stable operation without performance degradation even after more than 10,000 seconds of repeated operation. When the device was applied to reservoir computing, it achieved a recognition accuracy of 96.1% in a character classification task, demonstrating its potential as practical neuromorphic AI hardware.

The researchers stated, “Conventional memristors rely on repeated electrical stimulation to implement both long-term and short-term memory, which limits device reliability and reconfigurability. This study is significant because it presents a new neuromorphic device platform in which electrically stored long-term memory can be preserved while only short-term memory is selectively controlled in a non-contact manner using surface acoustic waves.”

They added, “In the future, we plan to combine large-area integration technologies with surface acoustic wave control over a wide range of frequencies to develop more energy-efficient next-generation AI semiconductors and neuromorphic computing systems.”