中文
Home Research Publications Team Courses News Contact

Our research group is dedicated to innovating and developing artificial neuromorphic devices and arrays with ultra-high computing efficiency, ultra-low computing power consumption, high accuracy and large-scale integration. The main research direction is:

1)Through the microscopic characterization of the ion dynamics process to realize the exploration and recognition of the working mechanism of the device;

2)By simulating the working processes of synapses and neurons in the organism, new types of high-precision, low-power artificial neuromorphic device based on ion migration is designed and prepared;

3)Based on resistive memory and new-type neuromorphic devices, the large-scale integration of devices is realized by using the array structure to realize artificial neural networks and related logic applications.

Microscopic characterization of ion kinetic processes

       An in-depth understanding of device microdynamics is crucial to the development of artificial neuromorphic devices and their arrays represented by memristors. This research group systematically and deeply conducted the processes of oxygen ion migration in transition metal oxide memristors, silver ion migration based on cation migration devices, and lithium ion migration in synaptic transistors based on two-dimensional layered semiconductor materials. Through the high-resolution transmission electron microscope (HR-TEM), scanning electron microscope (SEM), conductive scanning probe microscope (CAFM) and other characterization methods, the ion migration process in the memristor is intuitively and deeply displayed, and a variety of artificial The working mechanism of the demeanor device has played a guiding role in device design and array integration.

Artificial neuromorphic device based on ion migration

      Traditional neuromorphic devices usually only attempt to simulate biological characteristics in electrical behavior. This research group designed the synaptic transistors and Hetero-synapse devices that can accurately simulate biological synapses from micro-dynamic processes to macro-electrical characteristics through detailed research on the working process of biological synapses. In addition to the conventional transition metal oxide material system, this research group has also studied synaptic transistors based on two-dimensional layered semiconductor materials, organic electrolytes, and artificial neural components based on metal-insulator conversion. An ultra-low power consumption of about 30 fJ / spike is achieved experimentally.

Artificial neural network based on neuromorphic device array

      Based on the structure of two-terminal memristors and three-terminal heterologous synaptic devices, this research group is committed to achieving array integration with high integration density and ultra-low power consumption. Based on this array structure, efficient logic operations, machine learning problems, etc. can be achieved. This is of great significance for breaking through the "von Neumann" bottleneck, accelerating the realization of the "Internet of Things" and other concepts, and promoting the rapid development of neuromorphic computing.