多孔非晶碳薄膜的忆阻效应及其超低功耗突触器件研究

Memristors have potential applications in next-generation information storage technology, brain-like artificial intelligence devices and other fields. The development of memristors with ultra-low power consumption comparable to that of the human brain is an important basis for building a neural network. Aiming at the contradiction between low power consumption and resistive-state retention, this project proposes to develop the ECM-type memristors and artificial synapses based on porous amorphous carbon (a-C) thin films. Some improved physical/chemical methods will be used to fabricate the porous a-C films with adjustable pore size and pore-wall modification. The influence of pore size and surface modification on the evolution of the metal conducting filament is carefully studied, and the thermal/kinetic model of the diffusion/migration of metal atoms in the pore is established. The size confinement, interface confinement and resistive-switching localization not only help to obtain the ultralow-power and long-retentive memristors, but also overcome the issue of large fluctuation and poor endurance caused by the high resistive-switching randomness in continuous dielectric medium. Based on the porous a-C memristor, an ultralow power artificial synapse is built to emulate the biological synaptic functions and brain-like cognition behaviors. The simple serial-parallel synaptic circuits will be demonstrated, and the simulation methods for the synergistic effect of multi synapses will also be studies, which will provide valuable experience for developing neural network. The project follows the research route of "material-device-function" and is the research frontier in the multidisciplinary field of materials science, information science and artificial intelligence.

忆阻器件在下一代信息存储技术、类脑人工智能器件等领域有重要应用。开发可与人脑相比拟的超低功耗忆阻器件是构筑神经网络的关键。本项目针对器件低功耗与阻态保持性存在矛盾这一难题，拟发展基于多孔非晶碳（a-C）薄膜的ECM型忆阻器件和人工突触。改进材料制备方法，获得孔道尺寸可调控、孔壁表面可修饰的多孔a-C薄膜；重点研究孔道尺寸、表面修饰对金属导电细丝演变的影响，建立孔道内金属原子扩散/迁移的热/动力学模型。孔道的尺寸限域、界面限域、阻变局域等效应，不仅有助于获得低功耗、高保持忆阻器件；而且有望克服传统连续均匀介质中阻变随机性高导致的波动性、耐受性问题。基于多孔a-C忆阻效应，构建超低功耗突触器件，实现对生物突触功能和类脑认知行为的模拟；探索简单串/并联突触电路，研究多突触协同的功能模拟方法，为发展神经网络积累经验。项目以“材料-器件-功能”为研究思路，是材料、信息、人工智能交叉领域的研究前沿。

低功耗、长保持特性的忆阻器件在高密度存储、存算一体化等领域具有重要应用潜力。本项目针对忆阻器件低功耗与阻态保持性存在矛盾这一难题，基于多孔限域效应发展高性能忆阻材料和人工突触器件。发展磁控溅射、脉冲激光沉积等制备工艺，结合氮掺杂以及热退火等方案，实现了多孔非晶碳（a-C）薄膜材料的可控制备。系统研究了a-C 孔道对金属导电细丝的限制效应及其对忆阻特性的调控作用，建立了a-C 孔道中金属导电细丝形成/断裂的数学分析模型。基于纳米孔道对导电通道的限域效应，实现了稳定可控的原子尺度导电通道并在单一器件获得了16个高可控量子电导态，基于量子电导实现了实蕴逻辑存算一体功能。基于多孔限域效应，结合固体电化学氧化理论，提出了忆阻器双缓冲层新策略，有效缓解了功耗与阻态保持性矛盾，并研制出具有超长寿命的a-C忆阻元器件。采用界面能带设计和电学性质调控，构建基于肖特基势垒调制的二阶忆阻器件，实现了对Bienenstock-Cooper-Munro学习法则的完整模拟。相关研究结果在Nature Communications、Advanced Materials、Nano Letters、Advanced Functional Materials、IEEE Electronic Device Letters等期刊发表论文33篇，申请发明专利5项，授权专利3项。与国际知名研究团队保持紧密合作，开展了多种样式的学术交流活动。培养毕业博士研究生3名，硕士研究生8名。项目执行期间，负责人作为“低维氧化物半导体同质/异质界面构建与应用基础研究”成果的第二完成人获2019年国家自然科学二等奖，并获得国家杰出青年科学基金资助