In-materio computing in random networks of carbon nanotubes complexed with chemically dynamic molecules: a review

Abstract

The need for highly energy-efficient information processing has sparked a new age of material-based computational devices. Among these, random networks of carbon nanotubes (CNTs) complexed with other materials have been extensively investigated owing to their extraordinary characteristics. However, the heterogeneity of carbon nanotube (CNT) research has made it quite challenging to comprehend the necessary features of in-materio computing in a random network of CNTs. Herein, we systematically tackle the topic by reviewing the progress of CNT applications, from the discovery of individual CNT conduction to their recent uses in neuromorphic and unconventional (reservoir) computing. This review catalogues the extraordinary abilities of random CNT networks and their complexes used to conduct nonlinear in-materio computing tasks as well as classification tasks that may replace current energy-inefficient systems.

Physical Reservoir Computing Based on Nanoscale Materials and Devices

“…Such randomly connected nanotube systems are analogs to random resistive networks in percolation theory, which are conductive as long as the density of the connected resistors exceeds the percolation threshold. Physical reservoirs constructed from the mixture composite of CNT/ polymer [ 93–98 ] and CNT/polyoxometalate [ 99–103 ] network have been reported, as shown in Figure a,b–d, respectively. For the physical reservoir of CNT/ polymer network, [ 94 ] CNTs are highly conductive molecular‐scale pipes, forming unique circuit structures that act as coupling capacitors to maintain internal electrical coupling and provide nonlinear effects.…”

Reservoir computing using networks of memristors: effects of topology and heterogeneity

“…2 Much effort has been invested in developing neuromorphic integrated circuits based on complementary metal–oxide semiconductor technologies 3,4 but additionally several nanoscale systems and components show promise, especially spintronic oscillators, 5 and self-organised networks of memristive devices. 6,7 Self organised systems, which include carbon nanotube, 8,9 nanowire 10–14 and nanoparticle 15–18 networks, are appealing because they have the potential to naturally integrate large numbers of memristive devices into brain-like structures that are difficult (or practically impossible) to attain using top-down processes, with low fabrication costs.…”

The van der Pol physical reservoir computer

“…PRCs can be considered as hardware implementations of unconventional neural networks [41]. PRCs have been implemented as carbon nanotubes [42,43], soft robots [44,45], skyrmion [46], a Duffing array [47,48], a Hopf oscillator [49,50], origami structures [24], micro-electromechanical systems (MEMS) resonators [51], and even strawberry plants [52]. This computing scheme is often used for time series data processing, such as speech recognition [35,40], robotic gait [53], and chaotic time series predictions [33,35].…”