Researchers at Fudan University, China, are investigating memristive switching behaviours in In2Te5 asymmetrical hetero-structures for next generation computer memory applications inlcuding AI. Memristors are one of the most promising emerging technologies for next generation computer memory applications as they operate at high-speed, have non-volatile low-power operation and offer high-density integration. Memristors with structures that are based on metal-insulator-metal contacting layers, crossbar arrays and nanowires have all been widely investigated over the last decade. Electrodes with symmetrical and asymmetrical patterns have also been discussed in literature. The operation of reading, writing and erasing is achieved at high-frequencies and narrow positive voltages, thus resulting in low consumption and high-efficiencies. Memristors that are based on crossbar arrays are applied for logic operations and programmable circuits. Due to the exhibition of similar properties to synapses, these memristors are regarded as the best choice devices to achieve artificial general intelligence (AGI). Members of Professor Jing Li's research group in the sputtering lab at Fudan University. Professor Jing Li's research group at Fudan University. The resistive switching mechanism of memristors is the key factor for the performance of the whole device. This mechanism mainly depends on the interaction between the core layer and the different metal electrode layers. The performance of memristors can be effectively controlled by investigating the resistive switching (RS) mechanism of the related materials. The resistive switching mechanism is related to the material choices, whilst the different device performances should be affected by the preparation methods employed and by the electrode choices. Generally, the main methods to fabricate memristors are to use either a sputtering method, atomic layer deposition, pulsed laser deposition or a sol-gel method; and the work function of the electrodes is an effective tool for tuning the performance of memristors. In this issue of Electronics Letters, authors Jing Li and Yafei Yuan fabricate an In2Te5 memristor and resolve the underlying resistive switching mechanism, thus allowing the performance of the device to be modulated. Firstly, they successfully incorporated an In2Te5 phase-change film as a core layer with metal-semiconductor-metal structure in a memristor for the first time; and have shown the devices exhibit reliable bipolar resistive switching characteristics. Secondly, the switching mechanism in the memristor samples that are based on an In2Te5 core layer was found to be correlated to the trap-space charge limited current (SCLC). In addition, the research results in their Letter show that the sample devices display good endurance and retention. In2Te5, as a binary chalcogenide, has received little attention as a potential memristive material, and most related studies have focused on its phase change behaviour and topological properties. The RS mechanism in In2Te5-based memristive devices has not been well investigated. The results obtained will be a significant guide for memristive applications that are based on In2Te5. As a memristive device that is affected by the trapped SCLC, the memristive properties can be modulated by the injection current intensity. A high injection current could make the sample memristors show unpredictable switching behaviour due to thermal effects. A weak injection current could cause the sample memristors to exhibit slow switching behaviours. Importantly, the interfaces between the two electrodes and In2Te5 core layer are key factors to consider. The work presented in this Letter provides direct evidence that In2Te5 films are good core layers in memristive devices. The In2Te5-based memristor exhibits reliable bipolar resistive switching and can be applied in non-volatile, logic operation and programmable circuits. For longer-term developments, this work makes a case for the application of chalcogenide materials to play an active role in AGI research. Professor Jing Li's research group are also focused on carrying out fundamental and applied research into chalcogenide phased-change materials and other resistive switching materials. In the meantime, the group are carrying out AI-based deep learning research with the development of basic memristive components and programmable logic circuits. “We think that basic hardware realisations of neurons and synapses by the memristive application will be realised over the next decade” explains Jing Li. “This is causing a revolutionary change in the field of science and technology, industry and even in the way of social life.” The most difficult challenge for human beings is to understand the working principle of the brain intelligence. To achieve AGI, one practical approach is to build a so-called neurocomputer, which could be trained to produce autonomous intelligence. A neurocomputer imitates the biological neural network with neuromorphic devices, which emulate the bio-neurons, synapses and other essential neural components. A neurocomputer could perceive the environment via sensors and interact with other entities using a physical body. The Fudan team believe that the most probable way a neurocomputer could be realised is to use memristors and brain-like logic circuits.
Search from 250 M+ research papersSearch from 250 M+ research papers
Jan 13, 2020
Zhongrui Wang + 6
Open Access