Recently, a significant advancement has been made in the realm of piezoelectric acoustic wave memristive neural networks by the research team helmed by Wang Zhongrui from the Shenzhen-Hong Kong College of Microelectronics (SH-MICS) at the Southern University of Science and Technology (SUSTech). This achievement was made in collaboration with Yu Hongyu's team from Shenzhen Polytechnic University. Their groundbreaking research findings have been published in the esteemed academic journal Science Advances.

This study introduces a novel neural network architecture based on surface acoustic wave (SAW) memristors, leveraging the acoustoelectric effect. By incorporating non-volatile Ag/SiO₂ memristors, the team has successfully achieved stable and programmable phase modulation of SAW. They have constructed a fully functional SAW memristor neural network, which has demonstrated an impressive recognition accuracy of 91.7% in vector classification tasks. Notably, the system's size has been significantly reduced by five orders of magnitude compared to equivalent electromagnetic wave neural networks. This breakthrough offers a new physical implementation pathway for low-power, highly integrated intelligent radio frequency signal processing and edge computing applications.