N. Frick, T. LaBean
Keywords: neuromorphic, inks, memristor, nanocomposite, synapse
Summary:This work presents an experimental and theoretical investigation of neuromorphic nanoelectronic composite inks with memristive properties for the production of printed devices with resistive switching memory for sensing and computation applications. The main components of the nanocomposite inks are metallic and semiconductive nanoparticles that form complex conductive pathways that mimic the behavior of biological synapses. By combining two distinctly different materials, this nanocomposite provides a tunable platform for implementing artificial synaptic networks that could be used for training and inference with feedforward and recurrent neural networks. We will discuss the basic characteristics of these inks and demonstrate how to produce printed neuromorphic devices that can perform binary logic classification, such as AND, NOR, and XOR. We will also show some unique properties of these materials, such as an active negative differential resistance of the network of resistive switches and memory relaxation properties that can lead to the creation of in-materia liquid-state machines and reservoir computers for even more complex signal processing. Finally, we will provide recommendations on integrating printed neuromorphic devices with CMOS electronics to develop efficient computing architectures on modern platforms.