Memristive and memsensor devices for filament free sparse CNT networks

我们对水平和垂直记忆器件的新颖的实验和理论结果为我们提供了基本的理解，也揭示了新的应用可能性。实验和计算机模拟有力地表明，与文献结果相反，限制银离子的迁移率(例如，通过从纯银到AgAu合金纳米粒子的转变)对于允许记忆器件的无灯丝开关是必要的。现在，基于已确定尺寸和成分的金属团簇沉积的良好特性，我们计划进一步深入了解绝缘基质中的非丝状银离子输运，特别是通过现场研究，并实现用于稀疏网络应用的可靠的记忆器件。重点将放在非氧化物基质上，以排除氧缺陷的叠加影响。除了原位表征的优点外，水平几何结构还将被用于一种新的记忆传感器应用，即已经成功演示的忆阻器和传感器的组合。为了实现全电路，我们将设计和研究稀疏碳纳米管网络，用于以下几个目的：a)将记忆金属氧化物传感器与简单的神经网络联系起来；b)通过随后的Agau团簇沉积直接形成专用的神经网络；c)作为记忆器件与网络外围接触的互连层。最终目标将是一个独立的决策神经网络，对非人工输入信号做出反应。

Our novel experimental and theoretical results on horizontal and vertical memristive devices allow fundamental understanding and also reveal new application possibilities. Experiments and computer simulation strongly suggest that, in contrast to literature results, a limitation of Ag ion mobility (e.g. by transition from pure Ag to AgAu-alloy nanoparticles) is necessary to allow for filament free switching of memristive devices. Now, based on a well-characterized setup for metal cluster deposition of defined size and composition, we plan to get further insight into non-filamentary Ag ion transport within insulating matrixes, especially by in-situ studies, and to realize reliable memristive devices for sparse network applications. The focus will be put on non-oxidic matrices to rule out superimposed effects from oxygen defects. Beside the advantages for in-situ characterization, the horizontal geometry will be employed for a new memsensor application, a combination of memristors and sensors that could already be successfully demonstrated. To realize full circuits, we will design and investigate sparse CNT networks for several purposes: a) to contact memristive metal-oxide-sensors (memsensors) to simple neural networks, b) to form dedicated neural networks directly  by subsequent AgAu-cluster deposition, c) as an interconnection layer for memristive devices contacting the periphery of the network. The final goal will be a standalone decision making neural network responding to non-artificial input signals.