Nanowatt-level optoelectronic GaN-based heterostructure artificial synaptic device for associative learning and neuromorphic computing

Abstract In recent years, research focusing on synaptic device based on phototransistors has provided a new method for associative learning and neuromorphic computing. A TiO 2 /AlGaN/GaN heterostructure-based synaptic phototransistor is fabricated and measured, integrating a TiO 2 nanolayer gate and a two-dimensional electron gas (2DEG) channel to mimic the synaptic weight and the synaptic cleft, respectively. The maximum drain to source current is 10 nA, while the device is driven at a reverse bias not exceeding −2.5 V. A excitatory postsynaptic current (EPSC) of 200 nA can be triggered by a 365 nm UVA light spike with the duration of 1 s at light intensity of 1.35 μ W∙cm −2 . Multiple synaptic neuromorphic functions, including EPSC, short-term/long-term plasticity (STP/LTP) and paried-pulse facilitation (PPF), are effectively mimicked by our GaN-based heterostructure synaptic device. In the typical Pavlov’s dog experiment, we demonstrate that the device can achieve "retraining" process to extend memory time through enhancing the intensity of synaptic weight, which is similar to the working mechanism of human brain.