Photo-Electroactive Non-Volatile Memories for Data Storage and Neuromorphic Computing

Su-Ting Han, Ye Zhou

• Publisher: Woodhead Publishing
• Year: 2020
• Print ISBN: 9780128197172
• eBook ISBN: 9780128226063
• Topic: Technology & Engineering
• Subtopic: Electrical Engineering & Telecommunications
• Index: Technology & Engineering

Photo-Electroactive Non-Volatile Memories for Data Storage and Neuromorphic Computing summarizes advances in the development of photo-electroactive memories and neuromorphic computing systems, suggests possible solutions to the challenges of device design, and evaluates the prospects for commercial applications. Sections covers developments in electro-photoactive memory, and photonic neuromorphic and in-memory computing, including discussions on design concepts, operation principles and basic storage mechanism of optoelectronic memory devices, potential materials from organic molecules, semiconductor quantum dots to two-dimensional materials with desirable electrical and optical properties, device challenges, and possible strategies.This comprehensive, accessible and up-to-date book will be of particular interest to graduate students and researchers in solid-state electronics. It is an invaluable systematic introduction to the memory characteristics, operation principles and storage mechanisms of the latest reported electro-photoactive memory devices.- Reviews the most promising materials to enable emerging computing memory and data storage devices, including one- and two-dimensional materials, metal oxides, semiconductors, organic materials, and more- Discusses fundamental mechanisms and design strategies for two- and three-terminal device structures- Addresses device challenges and strategies to enable translation of optical and optoelectronic technologies

1

Introduction to photo-electroactive nonvolatile memory

Jing-Yu Mao and Ye Zhou, Institute for Advanced Study, Shenzhen University, Shenzhen, P.R. China

Abstract

With the generation of explosive data and rapid development of information technology, urgent need for high-quality data storage devices need to be satisfied. Traditional von Neumann system has posed constraints in its inefficiency; however, frequent communication with limited bandwidth between central processing unit and memory, and the key factor lays on data storage device. Nonvolatile flash memory with stable performance has served as electrical-driven data storage device in practical applications for decades and emerging resistive switching memory is booming with various superiorities. Among existing modulation approaches, light acts as the most feasible and effective one which attracts great attentions in researches on optoelectronic devices which have been conducted recently. To achieve maximum exploitation in existing memory devices, photo-electroactive materials have been employed to construct photonic memory and novel device platforms since introduction of light leads to additional channel modulation with high bandwidth, high speed, and low power consumption. Concept of neuromorphic computing is brought into sight for more efficient and biorealistic implementation inspired by the highly parallel data processing inside human brain with numerous neurons and synapses in between. Targeted at future computing, integration of memory and processing functions into one offers a potential orientation in addressing present difficulties.

Keywords

Photo-electroactive materials; optoelectronic memory; resistive switching; flash memory; neuromorphic computing

In the Big Data era, enormous quantities of data are generated every second from various sources such as smart phones, personal computers, etc., which have stimulated the rapid development of advanced technologies in digital devices and networks. As the objects or physical information (signal) in real world are connected more tightly with networks through the Internet of things (IoT), versatile electronic devices with distinct functions will be entailed as building blocks [1,2]. None of these would have happened without progressive development of microelectronics especially data storage technology that advanced with time. This brings critical requirements on data storage devices of the present memory technology in terms of capacity, operation speed, device miniaturization, cost, and energy consumption [3,4]. To deal with the ever-growing amount of data, memory cells with decreased unit area and higher integration level were developed; thanks to the well-developed semiconductor technology [5]. During the past decades, however, continuous shrinking of device size is reaching its limits due to the failure of Moore’s law [6,7]. An effective way to address this issue is to create novel nonvolatile memory with various functionalities by modulation from additional physical channels. Among existing modulation approaches, light appears to be the most feasible and effective one which attracts great attentions and substantial researches on optoelectronic nonvolatile memory have been conducted [8–11]. Optoelectronic memory with its memory states modulated by both optical and electrical stimuli raises appealing prospect in data storage and other aspect beyond it [12–14].

In terms of enlarging the capacity of nonvolatile memory devices, unlike traditional binary storage, multilevel memory containing multiple memory states within one unit cell is critical to realize high-density data storage device, since more information can be recorded without enlarging the total device size. Large memory window and stable readout operations should be achieved to enable distinct separation of individual memory state and retention time of which stays as a key factor of this approach. It is critical to choose appropriate materials among versatile photo-electroactive materials to fulfill this requirement. In addition, in stark contrast to all electrical memory, light modulated memory has its superiority that the electrical readout is orthogonal to light modulation [15]. Other physical channels such as magnetic field or heat may also act as alternative ways to modulate the memory performance. Although, currently optoelectronic nonvolatile memory is unable to undermine the predominant status of electrically operated memory, vigorous potential has been exposed for the development of next-generation memory.

Traditional von Neumann system has posed constraints in its inefficiency, however, frequent and necessary communication with limited bandwidth between central processing unit and memory [16]. One way of addressing this issue is optoelectronic interconnection. Different from conventional electrical data transmission, highly efficient optical communication operates at the speed of light between data processing unit and memory unit, leading to a better way of data processing [17]. Besides, more densely interconnected and integrated device components are strongly needed to fulfill the urgent requirements for constructing innovative computing paradigm [18,19]. One could envision the prosperous future by combining these optimizations since the present computing system made of all-electronic components, in which massive energy consumptions and dissipations are inevitable from electronic devices operations, has failed to provide a satisfactory answer to better future computing [20]. Such limitations spark emergent demand on novel memory device concept and new types of computing architecture [21–23]. Parallel operations of data processing and memorizing enabled by in-memory computing show the potential to renovate the present computing architecture [24]. With the help of optical communication and processing, superiorities such as high bandwidth, high-speed data transmission, and energy-efficient operations will be accompanied with the appearance of novel memory devices.

They are inspired by the highly parallel data processing inside human brain with numerous neurons and junctions (synapses) in between [25]. The transmission of information is dependent on the synaptic connection strength which can be modulated by chemical approaches upon the arrival of input signals [26]. These series of actions will affect synaptic plasticity, which is the foundation of learning and memory in human brain [27,28]. Human neural network outperforms modern computers with respect to computing efficiency, although the computing speed of processor is significantly faster than that of its biological counterpart. Besides, a surprisingly low energy consumption per event (about 10⁻¹⁶ J) found in brain activities with low cost makes it superior to any present machine [29]. Therefore, energetically friendly brain-like computing systems with learning and memory capabilities are suitable for more effective and efficient computation [30]. Electronic and optoelectronic synaptic devices of resistive switching type and flash memory–based transistor type were employed to mimic the short- and long-term synaptic plasticity [31–36]. Concept of neuromorphic computing was brought into sight for more efficient and biorealistic implementation, which pave the way for the construction of artificial neural network [21,37–40].

Electrical-driven memory can be sorted into two categories including volatile memory and nonvolatile memory in view of data retention time. For volatile memory, transient response to program operation occurs before it recovers to its original state. Namely, the recorded data can only be preserved for a short period and will dissipate soon after power off, and this phenomenon is frequently used in dynamic random access memory (DRAM). For nonvolatile memory, the programed state maintains for a relatively long period after the removal of external voltage supply. Normally, the program and erase operations can be repeatedly performed for many times. Here, nonvolatile memory will be discussed in two most common categories, including three-terminal transistor-based memory and two-terminal resistive switching memory. Considerable efforts have been carried out in enabling reliable performance of memory cell by employing different functional materials. In these devices, photonic modulation was introduced in multifunctional devices concerning information communications, signal sensing, and logics, offering a broad new vista to achieve high computation standards.

Transistor-based flash memory is a well-studied data storage device that lies the foundation in basic field effect transistor (FET). Normally, a FET despite of its configuration, contains a gate electrode, a gate dielectric layer, a semiconductor layer, and a pair of source and drain electrodes. The current in semiconductor channel mostly depends on gate voltage modulation at constant source–drain voltage application. Flash memory has been widely employed for commercial applications like solid-state drives and flash disks in our daily lives due to the compatibility with complementary metal-oxide-semiconductor (CMOS) circuit. Before the flash memory came out, electrically erasable programmable read-only memory (EEPROM) as a revised form of electrically programmable read-only memory (EPROM) suffered from high fabrication cost. The operation mechanism flash memory with single transistor realization is based on charge trapping and detrapping within floating gate layer or charge trapping element inserted between semiconductor and gate dielectric layer or flipping of molecular dipoles within ferroelectric materials leading to the shift of threshold voltage [41–44]. This technology has proved its reliability in storage capacity, cyclic operation, an...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• List of contributors
• Preface
• 1. Introduction to photo-electroactive nonvolatile memory
• 2. Characteristics and mechanisms in resistive random-access memory
• 3. Memory characteristics and mechanisms in transistor-based memories
• 4. Two-terminal optoelectronic memory device
• 5. Three-terminal optoelectronic memory device
• 6. Synaptic devices based on field-effect transistors
• 7. Ionic synergetically coupled electrolyte-gated transistors for neuromorphic engineering applications
• 8. One-dimensional materials for photoelectroactive memories and synaptic devices
• 9. Novel photoelectroactive memories and neuromorphic devices based on nanomaterials
• 10. Organic and hybrid photoelectroactive polymer for memories and neuromorphic computing
• 11. Metal oxide materials for photoelectroactive memories and neuromorphic computing systems
• 12. Perovskites for phototunable memories and neuromorphic computing
• 13. Chalcogenide materials for optoelectronic memory and neuromorphic computing
• 14. Device challenges, possible strategies, and conclusions
• Index