Spatial network conduction in carbon nanotube and Ag-Ag₂S-Ag atomic switch network device platforms
Networks of nanomaterials sit at a confluence of desirable features for the fabrication of advanced electronic devices, including facile fabrication, high conducting element density, and novel electrical characteristics. The spatial conduction through carbon nanotube (CNT) and Ag-Ag₂S-Ag atomic switch networks was investigated to determine how better to implement them in novel sensing and computation device platforms. Selective gating of localized regions of CNT networks with varying densities was investigated. To achieve this, lithographically defined FET structures were developed that allowed gating of localised regions of the CNT FET network area. The CNT FET device sensitivity to gating of different regions of the CNT network was measured for devices with network densities close to the percolation threshold. A 10² increase in sensitivity to local gating for CNT FET devices with low network densities was observed compared with high-density CNT networks. Networks densities were all well below a density where metallic shorts could be present, so the trends observed were attributed to m-s junction dominated gating of the network. A better understanding of the dominant conduction in CNT network FETs at low network densities is important for tuning their properties for use as novel biosensing platforms or a tunable connectivity conducting film. A CNT network simulation was developed to test the effects of local gating on networks of bundled CNTs with varying densities. Up to 70,000 bundles on a 60 µm x 60 µm simulated network area were used to generate an electrical network of field sensitive elements where the gate field could be spatially modified to investigate the effect of local gating. Monte Carlo methods were used to simulate large numbers of random networks with m-s junctions as the dominant gate-dependent element. Networks with 13.5% metallic bundles were shown to exhibit trends in local gating similar to the experimental systems. Current density maps showed key conduction paths in low-density devices, which supports a model of m-s junction dominance to explain the local and global gate responses measured in experimental CNT FET systems. Prototype Ag-Ag₂S-Ag atomic switch networks (ASN) device were fabricated using spray-coated silver nanowires which were sulfurised using gas-phase sulfur after deposition. Electrical formation of memristive junctions and hysteretic switching curves were shown under swept voltage bias demonstrating memristive behaviour. ASN devices have been demonstrated to show critical dynamics and memristive characteristics due to the complex connection of atomic switches formed at Ag-Ag₂S-Ag junctions between wires. A fabrication and measurement protocol for ASN based neuromorphic devices on multi-electrode array (MEA) platforms was developed. The electrical measurement system was designed and deployed to facilitate time-resolved measurement across multiple channels simultaneously on those MEA platforms. Under DC bias, MEA-based ASN devices showed switching events with a power-law distribution over two orders of magnitude of conductance changes and time intervals consistent with self-organized criticality within the network. The dynamic response of the critical system was measured across the network area. Changes in the relative voltage across the ASN network area were observed using 16 channel MEA platforms, showing spatiotemporal variation in voltage across the network. Novel application of principal component analysis to ASNswas used to demonstrate reduction of dimension while preserving relative voltage changes. This paves the way for scalable analysis of the complex dynamic signals from critical ASN systems.