Thin-film transistors mimic biological synapses

Neuromorphic computing parallels the nervous system, distributing processing tasks to artificial synapses, the computational counterpart of neurons. While various types of devices are used as artificial synapses, those based on two-dimensional materials are especially promising due to their atomic thickness and advantageous electrical and optical properties, which allow these devices to be further miniaturized.

Wang et al. replicated the behavior of neurons using transistors fabricated with two-dimensional molybdenum disulfide (MoS₂). These electrically stable thin-film transistors mimicked the plasticity and transmission of information between biological neurons. The transistors were able to simulate various types of synaptic functions, including pulse-paired depression and short- and long-term potentiation and depression.

“This innovation contributes new insights into simulating the information transmission and plasticity between biological neurons, offering a fresh avenue for achieving synaptic behavior akin to that of biological synapses,” said author Yanmei Sun.

In addition to aiding in the development of neuromorphic computing, these findings could be used to develop biosensors and devices capable of interfacing with the nervous system to monitor biological signals. The authors plan to further explore and optimize the electrical characteristics of 2D material thin-film transistors for diverse applications.

“Building upon this foundation, we aim to simulate a wider range of artificial synaptic models to address more real-life challenges,” Sun said. “By extending our research on artificial synapses to different application domains and further studies, we aim to maximize the potential of this research, address practical issues, and advance the field of thin-film transistors.”

Source: “Bio-inspired synaptic behavior simulation in thin-film transistors based on molybdenum disulfide,” by Yufei Wang, Qi Yuan, Xinru Meng, and Yanmei Sun, Journal of Chemical Physics (2023). The article can be accessed at https://doi.org/10.1063/5.0174857 .