Proton-Assisted Redox-Based Three-Terminal Memristor for Synaptic Device Applications

Abstract

Emerging technologies, i.e., spintronics, 2D materials, and memristive devices, have been widely investigated as the building block of neuromorphic computing systems. Three-terminal memristor (3TM) is specifically designed to mitigate the challenges encountered by its two-terminal counterpart as it can concurrently execute signal transmission and memory operations. In this work, we present a complementary metal-oxide-semiconductor-compatible 3TM with highly linear weight update characteristics and a dynamic range of ∼15. The switching mechanism is governed by the migration of oxygen ions and protons in and out of the channel under an external gate electric field. The involvement of the protonic defects in the electrochemical reactions is proposed based on the bipolar pulse trains required to initiate the oxidation process and the device electrical characteristics under different humidity levels. For the synaptic operation, an excellent endurance performance with over 256k synaptic weight updates was demonstrated while maintaining a stable dynamic range. Additionally, the synaptic performance of the 3TM is simulated and implemented into a four-layer neural network (NN) model, achieving an accuracy of ∼92% in MNIST handwritten digit recognition. With such desirable conductance modulation characteristics, our proposed 3T-memristor is a promising synaptic device candidate to realize the hardware implementation of the artificial NN.

Control-Etched Ti3C2Tx MXene Nanosheets for a Low-Voltage-Operating Flexible Memristor for Efficient Neuromorphic Computation

“…The effectiveness of using a p-Ti 3 C 2 T x -based memristor as an artificial synapse was assessed in a four-layer neural network designed for pattern recognition simulation. The neural network comprised an input layer with 784 neurons, two hidden layers with 250 and 125 neurons, respectively, and an output layer with 10 neurons, all interconnected by synapses (Figure a) . The input data for training and testing consisted of handwritten digits (ranging from 0 to 9) from the Modified National Institute of Standards and Technology (MNIST) database, with each digit represented by a 28 × 28 pixels image.…”

“…The neural network comprised an input layer with 784 neurons, two hidden layers with 250 and 125 neurons, respectively, and an output layer with 10 neurons, all interconnected by synapses (Figure 7a). 59 The input data for training and testing consisted of handwritten digits (ranging from 0 to 9) from the Modified National Institute of Standards and Technology (MNIST) database, with each digit represented by a 28 × 28 pixels image. The memristor's behavior, in terms of linearity, dynamic range, and precision during the potentiation and depression processes (Figure 4d), played a crucial role in updating the synaptic weights during the training phase.…”

Unraveling the origins of the coexisting localized-interfacial mechanism in oxide-based memristors in CMOS-integrated synaptic device implementations

“…11,12 Concurrently, non-filamentary memristors have also been explored in neuromorphic research. [13][14][15] The analog nature of such memristive devices, due to their gradual resistive switching characteristics, aligns well with the intricacies of the human brain, allowing for the creation of artificial neural networks that closely mimic the behavior of biological synapses. The synaptic weights, also known as the connection strength between the neurons, are represented by the conductance of each device, and the analog properties enable different weights to be stored via a step-by-step weight adjustment during the neural network training process.…”

Proton Conducting Neuromorphic Materials and Devices

“…There is also a natural biological context to consider protons for neuromorphic computing owing to their important role in the animal brain. Protons can be inserted or extracted into the channel layers, or stored in solid or liquid/gel electrolytes to develop three-terminal memory devices, and also induce nonvolatile resistance change in two-terminal memristor devices. − Finally, proton-conducting materials present nonvolatile memory and rich biomimetic properties, offering an appealing path to neuromorphic hardware and efficient implementations of neuromorphic artificial intelligence since the intercalation and movement of protons as a charge carrier other than electrons provides a new knob for tuning the electrical properties of a material. ,− …”