非晶碳忆阻器件的构筑、性能调控及其在信息存储和突触仿生中的应用

The emerging memristive devices have great potential applications on the development of high-speed/high-density information storage technique and the fabrication of Artificial Neural Networks (ANNs) with inherent learning ability. Cabon based material is regarded as one potential material for memristive device due to its several advantages, such as controlled structure and electrical properties, stability, and biocompatibility. This project would start the study from the controllable growth of amorphous carbon film, the fabrication of memristive devices with sharp resistive switching and gradual resistive switching, and the study of their applications on storage information and synaptic emulation. The project would mainly solve the key problems of the impact of sp2 carbon cluster on the conducting filaments growth/rupture and the adjustment of resistance relax in time-domain. To obtain resistive memory device with excellent performance (e.g. high-speed, stability, high-endurance), this project would develop several new approaches to localize the resistive switching region, such as multi-layer structure design, arc-shaped electrodes fabrication and grains embedment. Based on the single memristive device and 1T1R structure, we plan to fabricate the artificial synaptic device and realize the synaptic emulation with several learning/memory functions. The project would develop in-situ TEM characterization technique to reveal the inherent mechanism of enhancement of sp2 clusters on the growth of conducting filaments, to confirm the drive force of conducting filaments relax, and to build the kinetics and thermodynamics model for conducting filaments growth and relax. It is expected to the project would advance the application of carbon materials on the emerging information storage area, which is an advanced issue of interdisciplinary field involving information, materials and physics.

新兴的忆阻器件在发展高速/高密度存储技术、构建具备认知能力的人工神经网络等方面极具前景。碳材料因其结构和电学特性易于调控、稳定性和生物兼容性良好等优势，成为潜在的忆阻材料体系。本项目在实现非晶碳薄膜可控生长的基础上，构筑电阻突变和渐变的碳基忆阻器件，并将其用于信息存储和突触仿生。重点解决SP2碳团簇对导电通道形成/断裂的影响，以及电阻弛豫的时域可调性等关键科学问题。发展多层结构、曲面电极、粒子包埋等方法，实现阻变位置的局域化，获得高速、稳定、长寿命的阻变存储器件。基于单一忆阻器件或1T1R构架，构筑人工神经突触器件，实现生物突触与认知、学习行为相关的多种功能模拟。发展忆阻器原位透射电镜表征技术，揭示SP2团簇促进导电通道生长的内在机制，明确导电通道弛豫的驱动力，建立导电通道形成与弛豫的动力学和热力学模型。本项目将推进碳材料在新型存储技术领域的应用，是信息、材料、物理等交叉领域的前沿课题。

碳材料因其结构和电学特性易于调控、稳定性和生物兼容性良好等优势，成为潜在的忆阻材料体系。本课题围绕非晶碳材料构筑忆阻器件，从忆阻机理探究、忆阻可靠性的提升以及忆阻器突触仿生功能/信息存储等方面开展了系列研究工作。取得主要结果如下:针对非晶碳材料局域结构调控及其对阻变行为影响问题，我们发展了限制电流调控和恒电流预置的方法，诱导非晶碳薄膜中sp2团簇的聚集，证实sp2导电团簇能够有效增强局域电场，提升忆阻器阻变性能；针对碳基忆阻器阻变波动性大和耐受性低的问题，发展了多孔复合结构、尖端电极、缓冲层等方法,促进了忆阻器导电细丝局域化，有效提升忆阻器件可靠性；在神经突触仿生器件方面，发展了一种数字型和模拟型混合的忆阻器件，实现了学习精度和速度可调的模式识别，发展了具有柔性、可转移特性的柔性可转移忆阻突触器件，为下一步柔性可穿戴神经形态计算系统的发展提供了研究方案。本项目对构筑高可靠性忆阻器件及其神经突触仿生/信息存储方面提供了科学依据和技术基础。相关结果发表在 Small、Adv. Electron. Mater.、Appl. Phys. Lett、IEEE EDL 等杂志报道，发表SCI论文45篇, 申请发明专利9项，其中授权非晶碳忆阻器相关专利3项。在国内外学术会议上作报告8次。申请人在项目执行期间获批国家国际科技合作专项项目1项，获批国家自然科学基金重点项目1项。