[News] Chinese Research Team Achieves New Breakthrough in Semiconductor, Based on DRAM Principle

Currently, in fields such as autonomous driving, smart home systems, and industrial control, there is an increasing demand for edge-intelligent hardware to locally process real-time environmental data generated by sensors and smart devices, thereby minimizing decision latency. Neuromorphic hardware capable of accurately simulating various biological neuron behaviors is expected to drive the development of ultra-low-power edge intelligence. Existing research has explored hardware with synaptic plasticity (i.e., strengthening or weakening synaptic connections through adaptive changes), but to fully simulate learning and memory processes, multiple plasticity mechanisms—including intrinsic plasticity—must work together.

To address this issue, the joint research team led by Professor Bao Wenzhong from the School of Microelectronics at Fudan University, Professor Zhou Peng from the Institute of Integrated Circuits and Micro-Nano Electronic Innovation, and Professor Chai Yang from The Hong Kong Polytechnic University proposed a novel biomimetic neuron structure. Utilizing wafer-scale two-dimensional semiconductor (MoS₂) material and based on the principle of dynamic random-access memory (DRAM), they achieved—for the first time—the synergistic integration of intrinsic plasticity, spike-timing encoding, and visual adaptation in a single hardware unit.

According to the “Fudan University School of Microelectronics,” researchers developed a bio-inspired neural network (BioNN) for image recognition, in which the 2D DRAM neuron module is used as both an image preprocessing and computational layer. This neuron can simultaneously perform spike-timing encoding, regulate intrinsic neuronal plasticity, and emulate biological neural dynamics of visual adaptation. By breaking through the limitations of traditional neuromorphic hardware architectures, it integrates perception, memory, and computation into one, thereby achieving highly efficient brain-like visual event processing.

Looking ahead, the 2D neuron module could serve as a universal building block for expanding into large-scale neuromorphic computing systems, deeply integrating with advanced sensors, memory devices, and brain-inspired algorithms. This would enable the efficient construction of systems ranging from edge-intelligent terminals to large-scale distributed brain-like networks. Potential applications include autonomous driving, smart healthcare, robotic perception, and brain–machine interfaces, providing fundamental support for low-power, real-time intelligent systems and pushing neuromorphic computing technology toward forms closer to biological neural systems.

The research team stated that this breakthrough fully leverages the ultra-low-power advantages of 2D semiconductors, advancing AI computation toward a more biologically realistic, energy-efficient paradigm. Meanwhile, it opens up a brand-new pathway for applying 2D semiconductors in edge-intelligent hardware and neuromorphic vision systems. The team has also begun focusing on the engineering translation of their research outcomes, aiming to pave the way for the industrialization of 2D semiconductors, moving from “1 to 10.”

(Photo credit: FREEPIK)