Memristive devices by ALD: design aspects for high density 3D arrays for memory and neuromorphic applications

Currently ongoing 4th industrial revolution can be mainly described by two emerging technological fields: machine learning (ML) and internet of things (IoT). The success of the former relies on breakthroughs towards a full hardware implementation of deep neural networks. On the other hand, widespread adoption of IoT devices can be rapidly aided by advances in low power non-volatile memory technologies. In this talk, we will discuss the design aspects of simple nanoelectronic devices called memristors for implementation as both fully analog artificial synapses of a hardware neural network, or as a highly-scaled resistive memory element. We will focus on plasma-enhanced atomic layer deposition (PE-ALD) as a key technology providing homogeneous 3D deposition, nm-scale thickness control, as well as necessary stoichiometry tuning, as the ultimate pathway to high density 3D synaptic and memory arrays.