Highly Stable Artificial Synapse Consisting of Low-Surface Defect van der Waals and Self-Assembled Materials

Abstract

The long-term plasticity of biological synapses was successfully emulated in an artificial synapse fabricated by combining low-surface defect van der Waals (vdW) and self-assembled (SA) materials. The synaptic operation could be achieved by facilitating hole trapping and releasing only via the amine (NH2) functional groups in 3-aminopropyltriethoxysilane, which consequently induced a gradual conductance change in the WSe2 channel. The vdW–SA synaptic device exhibited extremely stable long-term potentiation/depression (LTP/LTD) characteristics; its dynamic range and nonlinearity reproduced near 100 and 3.13/–6.53 (for LTP/LTD) with relative standard deviations (RSDs) below 2%. Furthermore, after conducting training and recognition tasks for the Modified National Institute of Standard and Technology (MNIST) digit patterns, we verified that the maximum recognition rate was 78.3%, and especially, its RSD was as low as 0.32% over several training/recognition cycles. This study provides a background for future research on advanced artificial synapses based on vdW and organic materials.

Flexible Artificial Synapses with a Biocompatible Maltose–Ascorbic Acid Electrolyte Gate for Neuromorphic Computing

“…The human brain is mainly composed of a neural network consisting of around 10 11 neurons and around 10 15 synapses, making it robust, plastic, and fault-tolerant, and it has extremely low energy consumption (<20 W). − Enlightened by the working mechanism of the brain, there have been many attempts to fabricate artificial neuromorphic electronics to imitate the behaviors of biological neural systems. − As early as 1989, Mead proposed the concept of neuromorphic engineering, which is a nonbiological computing system with functions similar to the brain . With continuous development and expansion, neuromorphic engineering has become an interdisciplinary subject that draws inspiration from biology, physics, mathematics, computer science, and electronic engineering to design artificial nervous systems, based on the visual system, tactile systems, and multisensory integration, the physical architecture and design principles of which are based on the principles of the biological nervous system.…”

Flexible artificial Si-In-Zn-O/ion gel synapse and its application to sensory-neuromorphic system for sign language translation

“…Because 100 positive V WC pulses of +5 V were applied to the weight control terminal (T WC ), G post increased linearly from 0.6 to 27.8 nS. It also decreased gradually to the initial conductance level after applying 100 negative V WC pulses of −5 V. This result is quite impressive because an artificial synapse should have a multistate that can be modulated linearly and reproducibly for the successful implementation of a hardware neural network (HW-NN) (36)(37)(38)(39)(40)(41). This conductance response originates from the gradual ion movement inside the ion gel according to V WC (Fig.…”

Electrolyte‐Gated Vertical Synapse Array based on Van Der Waals Heterostructure for Parallel Computing

“…In this light, the conductance difference representation between two equivalent devices can be one of the solutions to achieve both excitatory and inhibitory synaptic weights. [17,45,47] Next, the output values (f m ) were calculated using the sigmoid activation function (f (I m ) = 1 1+e −I m ) and compared with the corresponding label values (k m ). Finally, if necessary, the weight values were updated using the backpropagation algorithm (details are presented in the Experimental Section).…”

“…[12][13][14] In recent years, for the implementation of HW-ANNs, numerous studies on artificial synapses with various structures have been reported. [15][16][17][18] Early research on artificial synapses has focused mostly on two-terminal nonvolatile memory devices based on a crossbar array structure, such as resistive random access memory, conductive bridge random access memory, and phase-change memory. [4,[19][20][21][22][23] This is because the crossbar array can be simply fabricated, easily expanded, and highly integrated.…”

“…The artificial vdW-SA synapse showed extremely stable long-term potentiation/depression (LTP/D) characteristics with relative standard deviations (RSDs) below 2%. [17] Qian et al also proposed an organic double heterojunction to enable a nonvolatile step modulation of the conductance in a threeterminal synapse, where the double heterojunction was consisted of N,N-dioctyl-3,4,9,10-perylene tetracarboxylic diimide (PTCDI-C 8 ), copper phthalocyanine (CuPc), and para-sexiphenyl (p-6P). This organic synapse showed an excellent weight update characteristic with a nonlinearity (NL) below 0.01, in the LTP region and great controllability of the synaptic weight with nondestructive read-out.…”