An organic electrochemical synaptic transistor array for neuromorphic computation of sound localization

Abstract

Neuromorphic devices have a potential to accelerate high-performance parallel and low-power memory computing, artificial intelligence, and adaptive learning. In this work, a facile and high-resolution patterning process is introduced to fabricate an organic electrochemical synaptic transistors (OESTs) array using a laser etching process and screen-printing ion gel. The OESTs show an excellent electrical-pulse-modulated conductance updating for synaptic functions and also remarkable mechanical flexibility and low energy consumption. Based on the linear, repeatable, and stable long-term plasticity, the long-term potentiation statistics of 2205 count points have been simulated to explore the regularity of their conductivity states. Furthermore, the sound-localization function was simulated by constructing a cross-grid array of OESTs. The normalized mean square error of sound localization results was reduced by ∼37.5% from the untrained period. This work provides a platform for designing a high-performance, flexible, and highly efficient neuromorphic computation for artificial neuromorphic systems.