Nanoscale Memristor Device and Circuits Design

Nanoscale Memristor Device and Circuits Design provides theoretical frameworks, including (i) the background of memristors, (ii) physics of memristor and their modeling, (iii) menristive device applications, and (iv) circuit design for security and authentication. The book focuses on a broad aspect of realization of these applications as low cost and reliable devices. This is an important reference that will help materials scientists and engineers understand the production and applications of nanoscale memrister devices. A memristor is a two-terminal memory nanoscale device that stores information in terms of high/low resistance. It can retain information even when the power source is removed, i.e., "non-volatile."

In contrast to MOS Transistors (MOST), which are the building blocks of all modern mobile and computing devices, memristors are relatively immune to radiation, as well as parasitic effects, such as capacitance, and can be much more reliable. This is extremely attractive for critical safety applications, such as nuclear and aerospace, where radiation can cause failure in MOST-based systems.

• Outlines the major principles of circuit design for nanoelectronic applications
• Explores major applications, including memristor-based memories, sensors, solar cells, or memristor-based hardware and software security applications
• Assesses the major challenges to manufacturing nanoscale memristor devices at an industrial scale

• Language: English
• Publisher: Elsevier Science
• Release date: Nov 8, 2023
• ISBN: 9780323998116

Book preview

Chapter 1 Memristor and spintronics as key technologies for upcoming computing resources

Piyush Duaa; Anurag Srivastavab; Parmal Singh Solankia; Mohammed Saif ALSaidia    a Department of Engineering, College of Engineering and Technology, University of Technology and Applied Sciences, Suhar, Oman

b Atal Bihari Vajpayee—Indian Institute of Information Technology and Management Gwalior, Gwalior, Madhya Pradesh, India

Abstract

The memristor is a fourth passive component, joining the resistor, capacitor, and inductor. It has been well established that memristance is present in all the devices, but due to the relatively large dimensions of the active area of the devices the magnitude of memristance is negligible. However, when we entered the era of nanodevices the magnitude of the memristor became significant and it can no longer be ignored in comparison to other counterpart quantities such as resistance. The journey started with Leon Chua’s mathematical model (Chua, 1971 [1]) and became important in 2008, when HP introduced its TiO2-based memristor with nonvolatile characteristics (Strukov et al., 2008; Chua, 2011 [2,3]). Currently, memristors are being explored for various applications including low-power memory devices, neural networks and neuromorphic systems, crossbar latches as transistor replacements, nanoelectronics devices for classical hardware security, and quantum computing (Valov and Yang, 2020; Karafyllidis and Ch, 2019; Wang et al., 2020 [4–6]).

In computational materials science, the problems of obtaining an optimized structure in a stipulated time with a high level of accuracy may be resolved by using high-performance computing with parallel processing. Other methods include use of memristor-based computing (Xia and Yang, 2019; Zidan et al., 2018; Ielmini and Wong, 2018 [7–9]), spintronics-based computing (Grollier et al., 2020; Fukami and Ohno, 2018; Grollier et al., 2016; Torrejon et al., 2017; Manipatruni et al., 2018 [10–14]), neuromorphic computing (Zhang et al., 2019; Marković et al., 2020; Song et al., 2020; Sangwan and Hersam, 2020; Sebastian et al., 2020; Ababei et al., 2021 [15–20]), and/or quantum computing (Karafyllidis and Ch, 2019 [5]). This is a smarter route that requires less time with high accuracy. Creating vaccines for Covid-19 was the latest major physical, chemical, and health-related problem (Wang et al., 2020 [6]) to be handled using computational resources. The status quo method is cyclic: explore new materials to simulate materials with high performance: for a given set of parameters, the only way to increase performance is to increase the quantity of computing resources, which in turn increases the consumed power; thus no more improvements are possible by increasing the quantity of computing resources only. In addition, Moore’s law is moving towards saturation, where increasing the number of transistors becomes a challenge. However, another way is to increase performance by altering the materials of the components to reduce the power consumption using the strength of the materials at the quantum level (Karafyllidis and Ch, 2019 [5]). Because computing resources with separate memory and processing units represent an obstacle, this chapter explores the strength of memristors/memristor materials, highlighted with real-time applications. It discusses the future of memristors/memristive devices in addition to the existing applications. The challenges and limitations with regard to properties of materials required to enhance the performance of memristive materials/devices are also addressed, along with the use of memristors for technologies such as artificial intelligence and their applications.

The memristor is a fourth passive component, joining the resistor, capacitor, and inductor. It has been well established that memristance is present in all the devices, but due to the relatively large dimensions of the active area of the devices the magnitude of memristance is negligible. However, when we entered the era of nanodevices the magnitude of the memristor became significant and it can no longer be ignored in comparison to other counterpart quantities such as resistance. The journey started with Leon Chua’s mathematical model [1] and became important in 2008, when HP introduced its TiO2-based memristor with nonvolatile characteristics [2,3]. Currently, memristors are being explored for various applications including low-power memory devices, neural networks and neuromorphic systems, crossbar latches as transistor replacements, nanoelectronics devices for classical hardware security, and quantum computing [4–6].

In computational materials science, the problems of obtaining an optimized structure in a stipulated time with a high level of accuracy may be resolved by using high-performance computing with parallel processing. Other methods include use of memristor-based computing [7–9], spintronics-based computing [10–14], neuromorphic computing [15–20], and/or quantum computing [5]. This is a smarter route that requires less time with high accuracy. Creating vaccines for Covid-19 was the latest major physical, chemical, and health-related problem [6] to be handled using computational resources. The status quo method is cyclic: explore new materials to simulate materials with high performance: for a given set of parameters, the only way to increase performance is to increase the quantity of computing resources, which in turn increases the consumed power; thus no more improvements are possible by increasing the quantity of computing resources only. In addition, Moore’s law is moving towards saturation, where increasing the number of transistors becomes a challenge. However, another way is to increase performance by altering the materials of the components to reduce the power consumption using the strength of the materials at the quantum level [5]. Because computing resources with separate memory and processing units represent an obstacle, this chapter explores the strength of memristors/memristor materials, highlighted with real-time applications. It discusses the future of memristors/memristive devices in addition to the existing applications. The challenges and limitations with regard to properties of materials required to enhance the performance of memristive materials/devices are also addressed, along with the use of memristors for technologies such as artificial intelligence and their applications.

1.1 End of Moore’s law

A transistor is a device with semiconducting properties required for amplification or switching of electrical signals and power. The vacuum tube transistor was the first transistor, fabricated at the beginning of the 20th century, which had a length typically between 1 and 6 in. The much smaller transistor was able to replace the bulky vacuum tube and mechanical relay. In 1956, John Bardeen, Walter Brattain, and William Shockley were awarded the Nobel Prize for their invention of the modern transistor, which led the entrance into the microelectronics era. At this moment it has reached to a limit of 14 nm and it is several thousand times smaller than components of microelectronics era [21,22].

The invention of the transistor has revolutionized the world of electronics, as the device became the basic component of all modern computers and power electronics [23]. This tiny device acts as an excellent electronic switch. The frequency of switching is so high that it can turn current on and off billions of times per second. The transistor is the basic component of modern digital computers and is the building block of integrated circuits, such as computer microprocessors or central processing units. Modern central processing units contain millions of individual microscopic transistors. The physical size of transistors has decreased over time, while their performance characteristics have improved drastically.

In 1965, the American businessman and engineer Gordon Earle Moore observed and predicted that the number of transistors would double every 18 months (for a given size of chip) as the size of the components was getting smaller, which is known as Moore’s law [24,25]. Moore’s law was based completely on observation of the performance of the devices.

Intel Corporation is one of the largest manufacturers of silicon microchips in the world and uses more than 290 million transistors in their processors. The current smallest size of transistors is 5 nm (corresponding to 25 atoms of silicon), which are being used in Apple’s new M1 and are manufactured by Taiwan Semiconductor Manufacturing [26]. To connect such components, wire of very few nanometers in size would be required, typically with a size of only 1–2 nm, i.e., just a few silicon atoms [27]. As the size of the components began to reach about 10 nm, the discussion began of reaching saturation; by the year 2016, when the size reached close to 7 nm, that was thought to be an unbreakable physical limit, but later on Intel launched a bid for 1.4 nm in 10 years [28]. Beyond this limit, the challenges faced were of leaking current and risk of overheating, as the size of the components was so small that a slight perturbation could lead to overheating [29], whereas this was not true for the central processing unit clock time, which has had mild increases in the last two decades [30]. The density of the components is now so high that the board looks like a continuum rather than discrete components, even though the components are all separate. At the same time, computing power has been increased enormously, while consumption of electricity is much less and heat dissipation has been reduced dramatically.

At the moment, the industry is talking about the possible saturation of Moore’s law, because the contraction of the size beyond a certain limit is not possible and at the same time the number of transistors is not increasing at the same pace as predicted. Now the question is how the development will continue after this saturation limit, keeping development and energy efficiency in focus. To overcome this challenge, research and development teams are looking to continue with multifunctional devices with a hardware route comprising quantum computing or neuromorphic computing [10–20] as one possible answer to this question. The idea is to obtain the optimum use of the available transistors on the chip, keeping the energy requirement in control while at the same time having processing and memory on the same component.

1.2 Life beyond Moore’s law: Multifunctional devices

To move beyond the era of Moore’s law, more options needed to be explored, with materials and devices having a multifunctional nature and small dimensions, such as 1-dimensional materials (nanowires) or 2-dimensional materials (nanosheets) with less power consumption and less heat dissipation. As far as computing is concerned, graphene is a multipurpose material that is lightweight and strong and flexible enough to be explored for mass scale production. Another possibility is to use memristors [7–9,31] and spintonics [10–20,32–36]. The memristor has been realized as a computer component along with the inductor, capacitor, and resistor, and could help to transform future integrated circuits by acting as one of four passive components along with the transistors, by controlling electrical flow. The memristor is treated as a circuit resistance switch [3] that can remember the amount of charge that had previously flowed through it. In electronic circuits and devices, electrons are the main charge carriers. Such action is performed with information processing; if both the charge and spin of the electron are controlled to conduct and provide information processing [12,14], such a phenomenon is known as spintronics, and it can be very useful in future computing technologies.

1.2.1 Features, strengths, and properties of multifunctional devices

Until the advent of the concept of quantum computing, supercomputing was considered to be the only option to improve the performance of computing resources. The numbers were added on to obtain thousands of nodes, to reduce the computational time and to increase the accuracy in the presence of many approximations. Algorithms were proposed to minimize the gap between computational results and observed experimental results. In this manner, both hardware and software means were set aside for the sake of development in the domain of computing [37].

Using hardware to improve performance required more energy to operate a large number of processors, even though the size of the hardware components was miniaturized following Moore’s law [38], and software was dependent on efficient algorithms. Another challenge was due to the increasing number of computing resources producing more heat, and cooling resources consuming a huge amount of