Near‐Infrared‐Irradiation‐Mediated Synaptic Behavior from Tunable Charge‐Trapping Dynamics

Abstract

of stimulation will transfer the shortterm memory into long-term memory. To achieve artificial synaptic recognition for learning and memory in electronic devices, concomitant plasticity with different timescale is essential. As a result, the short-term plasticity (STP) modulates the transient dynamical efficacy during the synaptic transmission, while the longterm plasticity (LTP) shows the stabilizing effect by the given stimulation. [10][11][12][13] Therefore, the emulated synaptic plasticity of STP and LTP render themselves supportive to the sophisticated cognitive function and adaptive behavior pattern.The emphasis on neuroinspired computing so far has been predominantly in electrical stimulation-induced resistance state switching in phase change memories, [14,15] memristors, [16][17][18][19][20][21][22][23][24] as well as transistor-based memories. [25][26][27][28][29][30] In contrast with the existing electrical interconnect power loss as well as the limitation in trigger selectivity and spatially confinement inherently from the computing by electric signal, emerging optical stimulation based synaptic devices can tune the synaptic plasticity enormously by photons with low-power and high-efficiency. Therefore, the photonic synapse architecture is considered to be more favorable in handling the von Neumann bottleneck. [31][32][33][34] In addition, photonic synapse based on Parallel information storage coupled with storage density is a major focus for non-volatile memory devices to achieve neuromorphic computing that can work at low power. In this regard, a photoactive charge-trapping medium consisting of inorganic heteronanosheets for the fabrication of a synaptic transistor is demonstrated. This synaptic device senses and responds to near-infrared (NIR) light signals and mimics the memorization and dynamic forgetting process due to the reversible nature of photogenerated charge interaction. Device-level synaptic evolutions from short-term plasticity to long-term plasticity, paired pulse facilitation, and paired pulse depression are realized with light modulation on the weight update terminal. To understand the underlying mechanism of the synaptic behavior under NIR signals, systematic analysis is carried out using in situ atomic force microscopy based electrical techniques. With its photoactive architecture, this information processing analogue is validated for visual object recognition, which paves the way for implementing NIR-controlled neuromorphic computing.

An Optogenetics‐Inspired Flexible van der Waals Optoelectronic Synapse and its Application to a Convolutional Neural Network

“…Following the theoretical understanding of the PPC effect of the ReS 2 layer through computational analyses, we performed additional analyses based on temperature-dependent optoelectrical and KPFM measurements to experimentally understand the PPC effect of ReS 2 . As shown in Figure S13 He et al [44] Wang et al [45] John et al [46] Wang et al [30] Ahmed et al [47] This work…”

Artificial Synapse Based on Organic–Inorganic Hybrid Perovskite with Electric and Optical Modulation

“…Although the 4500 learning phases used in this work are far less than one epoch (i.e., 60 000 learning phase), the recognition rate of our device (Case 2: 83.7%; Case 3: 86.8%) is as high as those reported in previous studies. [ 45,52,53 ] Additionally, the obtained simulation results also illustrate the advantage of our light‐sensitive device, which improved the learning capability of MNIST handwritten numerical patterns and the recognition rate.…”

Near‐Infrared Artificial Synapses for Artificial Sensory Neuron System

“…A good response to NIR light is observed, and the photo-current grows with the enhancement of optical power density, which is consistent to the previous reports. [61,62] Obviously, after irradiation with NIR light (5 s and 3.6 mW cm −2 ), the channel current I DS grows sharply to around 120 nA. After switching off the input NIR light, the I DS decays gradually, exhibiting a typical LTP characteristic.…”

“…[5][6][7][8][9] Inspired by the structure and operation principle of the human brain, the establishment of a nervous system based on artificial synapses and neurons to process information in parallel, is expected to surpass the conventional computing power and effectively overcome the "von Neumann bottleneck". [10][11][12][13][14][15][16] For neural activities in the nerve system, when an action potential reaches the presynaptic neuron, the release of neurotransmitters induced by cations takes place in the synaptic cleft. As the neurotransmitters diffuse to the postsynaptic neuron, they couple with receptors along the postsynaptic membrane to generate an action potential, and then continuously transmit to the next neuron through axons or dendrites.…”