Recent Advances in Synaptic Nonvolatile Memory Devices and Compensating Architectural and Algorithmic Methods Toward Fully Integrated Neuromorphic Chips

Abstract

Nonvolatile memory (NVM)‐based neuromorphic computing has been attracting considerable attention from academia and the industry. Although it is not completely successful yet, remarkable achievements have been reported pertaining to synaptic devices that can leverage NVM capable of storing multiple states. The analog synaptic devices performing computation similar to biological nerve systems are crucial in energy‐efficient analog neuromorphic computing systems. To use NVM as an analog synaptic device, researchers focus on improving device characteristics. Among various characteristics, the most challenging one is linearity and symmetry of synaptic weight update that is required for on‐chip training. In this regard, this review paper discusses recent synaptic device improvements focusing on novel schemes tailored for each NVM device to improve the linearity and symmetry. In addition to device‐level studies, recent research achievements are reviewed expanded up to chip‐level studies because in realizing neuromorphic hardware systems beyond a single synaptic device, several considerations and requirements are needed to confirm for high‐level design, and accordingly, cooptimize among synaptic devices, synapse arrays, electrical circuits, neural networks, algorithms, and implementation. Also, this review paper introduces various circuit and algorithmic approaches to compensate for the non‐ideality of the analog synaptic device.

Artificial Neuron Devices

“…36,37,41,53 To overcome these drawbacks, researchers have developed novel artificial synapses based on a variety of materials and structures, typically implemented with 2-terminal memristors or 3-terminal transistors. 42,43,54,55 These devices are capable of achieving neuromorphic functions, such as short-term and long-term plasticity (STP and LTP), similar to the synapses in biological systems. 38,56,57 In recent years, there has been growing interest in using flexible electronics for the development of artificial neuron devices.…”

“…The plasticity of these synapses, meaning their ability to change their strength and connectivity over time, is crucial for learning and memory. − Neuromorphic electronic systems, proposed by Carver Mead in the late 1980s to early 1990s, aim to design electronic systems that mimic the structure, function, and plasticity of biological neural networks (Figure ). While neuromorphic circuits based on silicon complementary metal-oxide-semiconductor (CMOS) technology have been developed to replicate synaptic functionalities, classical computing systems traditionally relied on the von Neumann computing architecture and suffered from limitations due to the separation of memory from processing, leading to issues such as speed latency, high energy consumption, and limited communication bandwidth. ,,, To overcome these drawbacks, researchers have developed novel artificial synapses based on a variety of materials and structures, typically implemented with 2-terminal memristors or 3-terminal transistors. ,,, These devices are capable of achieving neuromorphic functions, such as short-term and long-term plasticity (STP and LTP), similar to the synapses in biological systems. ,, In recent years, there has been growing interest in using flexible electronics for the development of artificial neuron devices. ,, Flexible electronics refer to electronic devices and systems that can bend, stretch, and conform to their surroundings without breaking or losing their functionality. ,− Flexible electronics possess mechanical properties similar to human organs and tissues, showing great advantages for the development of artificial neuron devices. ,,− They can be easily integrated with biological tissues and structures, allowing for seamless interaction with the nervous system and the development of biointerfaces and biohybrid systems. − …”

Impact of Oxygen Reservoir Layer on 3T Oxygen Ion‐based Electrochemical Random Access Memory Performance

“…[4] To implement VMM in hardware, each synapse device consisting of the array should satisfy several requirements: linear and symmetric conductance modulation under identical pulse bias, wide on/off conductance ratio, data retention, and endurance. [5] These requirements can lead to precise and energy-efficient computing (training and inference) processes for neuromorphic systems.…”

Charge‐Mediated Copper‐Iodide‐Based Artificial Synaptic Device with Ultrahigh Neuromorphic Efficacy

“…Moreover, it is estimated that computing devices consume over 8% of global electricity, a rate that is doubling every decade. [10,11] This underscores the pressing need to develop new functional materials and architectural designs for implementing neuromorphic computation systems.…”