Materials for Memristors

Abstract

Memristive devices are entering in our lives. They have started to be commercialized as non-volatile electronic memory, and despite their market size is still small, it is expected to increase up to 5.6 billion USD by 2026.[1] Moreover, recent investigations have shown that memristors hold great potential for applications in the fields of brain-inspired neuromorphic engineering, cryptography, and 5G/6G telecommunications.[2] This is surprising because such applications require figures-of-merit that are completely different and sometimes even opposite. For example, memristors used as non-volatile memory must exhibit very low variations of switching parameters to ensure a very high yield;[3] and, on the contrary, memristors for cryptography must exhibit very high and unpredictable variations of those same switching parameters.[4] The clue behind the versatility of memristors relies on the materials employed. Memristors are materials systems with two or three (sometimes even more) electrodes separated by an insulating or semiconducting material.[5] The use of different types of materials is a valuable tool allowing to adjust the electrical characteristics and figures-of-merit of the memristors. The reason is that different materials use different physical and chemical reactions to produce the resistance change, and those phenomena have associated different speed, energy, variability, and reliability. For example, one can find in the market memristive memories made of magnetic materials, phase-change materials, or metal-oxide materials, all of them showing strengths and weaknesses,[1] as shown in Figure 1. Intense research in the field of memristors is highly necessary for two reasons. The first one is to optimize the properties of those memristors that exhibited potential for some applications. Many articles have claimed that memristors could be useful for diverse applications but bringing them to the market requires intense work to optimize not only performance, variability, and yield.[2, 3] And the second reason is to discover new functionalities, not only in terms of electrical figures-of-merit but also mechanical, optical, and thermal properties that could enable their use in different objects and environments.[6] That is especially attractive for applications within the internet of everything. Some of these functionalities might be achieved via device engineering by experimenting with different types of materials, such as phase-change, metal-oxides, magnetic, ferroelectric, two-dimensional, perovskites, polymers, biomaterials, nanowires, nanotubes, and other novel materials.[7] In this special issue, Advanced Functional Materials brings you a Special Issue containing 20 articles in the field of materials for memristors written by some of the most brilliant scientists in this field. The articles touch upon multiple aspects related to related to fabrication, characterization, and modelling of memristive devices made of different materials, and many of them propose innovative applications. The first group of articles in this special issue focuses on tw-o-dimensional (2D) materials-based devices using MoS2, hexagonal boron nitride (hBN) and graphene oxide. Prof. Mingdong Dong from Aarhus University (article number 202213348) shows that the exfoliated MoS2 after oxygen passivation has interesting bipolar switching feature while the intrinsic exfoliated MoS2 cannot reset due to the lack of oxygen ions. To have intrinsic switching behaviours, the exfoliated 2D materials shall change to chemical vaper deposition (CVD) 2D materials due to the nature of grain boundary inside. Soon-Yong Kwon from Ulsan National Institute of Science and Technology (article number 202309455) shows that the wafer-scale highly polycrystalline semiconducting 2H-phase molybdenum ditelluride memristors can enable multilevel resistance states with small variation (<8.3%) and high yield (>83.7%). The CVD growth MoTe2 with small grains (≈30 nm) is responsible for nice switching behaviours. The Prof. Deji Akinwande from University of Texas at Austin (article number 202214250) also carefully found the reducing the top electrode (TE) deposition rate and increasing the thickness of MoS2 films can effectively improve yield up to 92% and DC cycling endurance 16 times. Moreover, an “effective switching layer” model compatible with both monolayer and few-layer MoS2 is built to give a clear physical image. Not only MoS2 after oxidation can have good switching behavior, Prof. Fei Hui from Zhengzhou University (article number 202302073) shows that the graphene oxide with Ag TE can have low leakage current (≈10−12 A), low operation voltage (≈0.3 V), and high endurance (>12000 cycles). They use a new method (called “chemical breakdown”) to obtain large-area uniform graphene oxide. The h-BN is also a powerful material, which can enable super resistive switching behaviors. Prof. Max C. Lemme from RWTH Aachen University (article number 202300428) shows that the Ni TE with hBN can enable threshold switches, low cycle-to-cycle variability (5%), and large On/Off ratio (107). The resistive switching mechanism is explained by the on and off of the Ni filaments. Prof. Mario Lanza from KAUST (article number 202213816) investigates the random telegraph noise in hBN memristors covering ∼7 orders of magnitude in current with consistent observation. More importantly, the ambipolar switching regime with very low resistance down to 50Ω can promote the high-frequency switches for 5G/6G communications soon. The second group is also based on 2D materials, but the devices are based on the creation of 2D heterojunctions. The 2D heterojunctions are formed based on the Van der Waals interaction, which quite is different from conventional heterojunctions with defects, covalent-bond, and interface mismatch. Prof. Tao Deng from Beijing Jiaotong University (article number 202302288) demonstrates the optical synapse based on 3D graphene/molybdenum disulfide (MoS2) heterostructure with photoresponsivity up to 105 A W−1 at 590 nm. The MoS2 can enable the optical sensing while the roll up 3D graphene can enhance the optical field and make the device polarization sensitive. Prof. Wenwu Li from fudan University (article number 202309642) creates graphene/α-In2Se3/h-BN/Cr-Au memristor with ultralow off-state current (4.2 × 10−13 A) using 8 nm hBN interlayer. The device shows on/off ratio up to 109 and 32 distinct resistance states. Prof. Yang Chai from The Hong Kong Polytechnic University (article number 202304242) develops an ultrathin two-terminal n-p-n selector using 2D transition metal dichalcogenides. By fine tune the MoS2/WSe2/MoS2 n-p-n selector, it shows high nonlinearity (≈230) and high current density (2 × 103 A cm−2). Prof. Saptarshi Das from University Park (article number 202308129) provide a comprehensive review about the 2D-memtransistor devices from fundamental operational mechanisms, device operations and applications. Moreover, large-scale integration of 2D electronics shows great protentional as next-generation electronics. The third group is based on the materials with dipoles, such as ferro- and piezo-effect. Prof. Peng Zhou from Fudan University (article number 202305822) give a new device concept of All-In-One optoelectronic neuristor based on the 2D ferroelectric α-In2Se3/SnSe p–n heterojunction. The co-exist p–n junction and photo-inducing ferroelectric polarization switching can enable high paired-pulse facilitation index (457%), short synaptic plasticity, long synaptic plasticity, and even retina-like optical adaptations. Not only for application, Prof. Tianxiang Nan from Tsinghua University (article number 202308944) also investigate the fundamental switching behaviors based on Multiferroic Insulator. His team find the interesting Magnon-mediated Spin Torqu, which could enable wide applications for next generation storage. Prof. Baker Mohammad from Khalifa University (article number 202305869) innovates an ITO/ZnO/Yb2O3/Au structure combined with a high-sensitivity piezoelectric nanogenerator ITO/ZnO/Al. Interestingly, the nanogenerator generating the electrical signals can drive analogue unidirectional neuromorphic device, which can mimic the process of a neuron without additional circuits. Prof. Hao Guo from North University of China (article number 202304648) use the four noncorrelated structures in the diamond crystal structure for the multilevel encoding. Such encoding method based on the energy level does not depend on physical structure parameters, which can enable high security. The fourth group is based on nanowire electronics for vision and security applications. Prof. Lili Wang from Institute of Semiconductors, Chinese Academy of Sciences (article number 202304119) demonstrates a parallel photoelectron storage and visual preprocessing in nanowire. The combination of image perception, visual memory, and in-sensor preprocessing functions can be explained by the cascaded defect engineering. Prof. He Tian from Tsinghua University (article number 202304758) reports an Ag nanowire-based physical unclonable function (PUF) with high encoding capacity. The ternary bit encoding method enables the internal topological connection structure of the PUF, resulting in a larger encoding capacity. For example, a 7 × 7 array PUF of the estimated decryption time is around 1.5 × 1058 years. The fifth group is related to phase change memory and organic memory. Prof. Baker Mohammad from Khalifa University (article number 202214615) shows that a 100 nm-thick GeTe device has highly reproducible phase change characteristics, while 70 and 200 nm-thick devices do not have reliable memory characteristics. Such behaviors can be explained by the phase loops and defect loops in 100 nm of GeTe. Prof. Yuchao Yang from Peking University (article number 202300458) demonstrates a 4 Mb phase change memory chips with in-memory multiply-accumulate and in-memory rank functions. The PCM-based neural architecture search shows the energy and time efficiency by 4779 times and 123 times, respectively. Prof. Sahika Inal from KAUST (article number 202304103) demonstrates two types of device architectures and applications using the high performance of the organic electrochemical transistor. Importantly, it shows a low threshold voltage (0.02 V), and fast switching speed (τON, τOFF = 336 µs,108 µs). Prof. Yoeri van de Burgt from Eindhoven University of Technology (article number 202307729) reviews the recent advancements in the field of organic transistors for neuromorphic systems and smart sensing applications. The authors declare no conflict of interest. Mario Lanza got the PhD in Electronic Engineering in 2010 at Universitat Autonoma de Barcelona. After postdocs at Peking University and Stanford University, in October 2013 he joined Soochow University as Associate Professor, and after 3.5 years he promoted to Full Professor. In October 2020 he moved to the King Abdullah University of Science and Technology (KAUST), where he is currently an Associate Professor in Materials Science and Engineering. Prof. Lanza has published over 140 research articles (including two Science and five Nature Electronics) and has registered four patents (one of them granted with 1 million USD). He is a Distinguished Lecturer of the Electron Devices Society (IEEE-EDS), and has received multiple funding projects (EU, NSFC, MOST) and international awards (Young 1000 Talent, Marie Curie, Elsevier YIA, Wiley Rising Star). He Tian received the Ph.D. degree from the Institute of Microelectronics, Tsinghua University in 2015. After postdoc at Yale University and University of Southern California, he joined Tsinghua as Assistant Professor. He is currently an Associate Professor and Deputy Director, Institute of Integrated Electronics, School of Integrated Circuits, Tsinghua University. He has co-authored over 200 papers (including one Nature, one Nature Machine Intelligence and four Nature Communications) and more than 8000 citations. His current research interests include various 2D material-based devices. He is selected as the World's Top 2% Scientists for year 2022 and life career.

Topics & Concepts