Brain-inspired, or ‘neuromorphic’, computing has the potential to overcome the inherent limitations of conventional silicon technology. These problems arise from the power consumption and the challenges of sustaining ever-increasing miniaturisation as feature sizes shrink in the sub-10nm regime. Neuromorphic computing architectures promise a new generation of faster, low power computers plus added functions like pattern recognition, which we are familiar with in our own experience. The brain is a complex network of connecting neurons and synapses. Amongst multiple competing approaches, a film of nanoparticles deposited on a surface appears to offer a comparatively inexpensive route to self-organising networks that inherently display brain-like dynamics. The making and breaking of nanofilament links between clusters in a conducting (or “percolating”) film may offer the key to the electrical behaviour. The Lab’s approach couples cluster beam deposition and electrical measurements with advanced atomic imaging using multiple-tip scanning tunnelling microscopy and electron microscopy. The synapses (links between clusters) are highly sensitive to the arrival of molecules, which also makes metal cluster networks prime candidates as chemical sensors.

The miniMACS