Single ferroelectric memristor shows synaptic plasticity and habituation

Our brains transmit and process information efficiently thanks to synaptic plasticity — the ability of synapses to grow and strengthen. Mimicking this behavior in the artificial synapses of neural network computing would make high-efficiency artificial neural computing possible.

Recent efforts at artificial synaptic plasticity have focused around memristors, which use external voltage signals to reconfigure resistance states. Li et al. take memristors to the next level by designing a device that mimics habituation behavior, a basic type of plasticity that allows an organism to adapt to the environment, in a single device.

“We found a new memristor that can implement habituation by a unique conductivity mechanism,” author Zhibo Yan said. “This work can complete the architecture of neuromorphic computing in hardware and decrease operational complexity and energy consumption.”

The device relies on conductive domain walls to change the conductivity. Unlike previous memristors, this conductivity mechanism allows it to mimic habituation by making non-monotonic changes under monotonic voltage. An analysis of the electrical transport behavior showed the evolution of conductive domain walls gives rise to the habituation.

The authors plan to reduce the size of the device so it can be easily integrated into large neural networks. They also aim to increase its synaptic plasticity.

“We hope that our device can be applied to future neuromorphic computing, promoting advances in hardware implementation of neuromorphic functions and further simplifying artificial neuromorphic networks,” Yan said.

Source: “Implementation of habituation on single ferroelectric memristor,” by Xinyu Li, Guangyuan Li, Zhihang Zhang, Wenjing Zhai, Wenhao Zheng, Liufang Chen, Lin Lin, Xiaohui Zhou, Zhibo Yan, and J.-M. Liu, Applied Physics Letters (2023). The article can be accessed at https://doi.org/10.1063/5.0141710 .