NANO-ORGANIC MEMRISTOR CIRCUITS WITH LEARNING CAPABILITIES – Moorea

MOOREA is a 3-year ambitious project targeting as final objective the experimental demonstration of the learning capabilities of a neuromorphic circuit composed of nano-scale organic memristors and their prospects for predictable performance, power consumption and scalability. It is based on the design, modeling, simulation and physical implementation of highly innovative nano-devices and on new architectures for hybrid circuits.
Miniaturization in microelectronics drives the digital economy but also faces very strategic challenges. In this context, long term research projects should contribute to this miniaturization, and most importantly, to (i) the multiplication and diversification of integrated functions and (ii) the management of power consumption. Nano-devices, thanks to their ultimate size and original functionality, will be key elements of this evolution, provided that technological variability inherent to the nano-scale can be handled. With their intrinsic tolerance to defects and their auto-compensation capabilities coming from a learning stage, neuromorphic architectures allow lifting this critical roadblock. Nano-memristors, which are programmable resistors, are ideally suited to be integrated in dense crossbars and used as nano-scale synapses in such architectures. Moreover, beyond synaptic function, the variety of behavior that allows the use of organic compounds gives access to entirely new utilizations of memristors for very high density implementation of the neurons including their learning capabilities. In addition, the non-volatile memory capabilities of the memristors allow designing circuits with minimized standby-power consumption and thus contribute efficiently to tackle the energy issue.
At the device level, the focus will be put on a new class of organic memristors, the active part of which will be redox and charge separation organic complexes grafted on the bottom electrodes of the crossbars. The top electrode will be fabricated by transfer printing. While ambitious, MOOREA is designed to minimize risks. Indeed, the functionality and robustness of a first type of organic memristive material have already been assessed by the consortium, as well as its compatibility with transfer printing processes. In addition, conventional inorganic memristors will also be fabricated to serve as a comparison in a benchmark effort. In the more exploratory part of the project, individual carbon nanotubes will be used as electrodes for fully organic nano-memristors. It will notably allow the evaluation of scaling laws for energy and speed performances.
We will develop the first physics-based and compact models of memristors adapted to circuit design and propose new architectures for adaptive circuits. The project will lead to the conception and realization of a full demonstrator of nano-memristor based circuit with learning capabilities.