Networks of Functional Metal Oxides Towards Neuromorphic Materials

Artificial Intelligence (AI) has seen a massive increase in importance over the past years, and with it comes the high power consumption required for operation of deep learning networks. One way to reduce the power consumption of these systems is to move away from computing on central processing units (CPU), accelerated by graphics processing units (GPU) and look into brain-inspired (neuromorphic) hardware. Current neuromorphic chips outperform conventional computers when considering power consumption for tasks such as pattern recognition. But the large number of transistors required for these chips to emulate one neuron, limits their functionality. To lower the power consumption further, research is looking into new materials and novel device structures, with a special focus on memristors. A memristor (or memory resistor) is a device with a variable resistance that can be programmed and which depends on the history of the voltage that was applied to the device. Interestingly, these devices remember their resistance state even when the power is turned off, they are non-volatile.
In this work, we investigate the creation of self-assembled nanoscale networks of oxide materials with the intention of developing richer, more complex, and more tunable materials, including memristors. We achieve this by combining polymer imprinting and -templating to form highly ordered and interconnected networks of different functional metal oxides. We show that many materials are in reach for the combined templating and imprinting method and that these materials show promise for future applications in adaptable electronics.