Robust Memcapacitive Synapse Array for Energy-Efficient Motion Detection.

Emerging neuromorphic systems offer a promising alternative for memory and sensing compared to traditional configurations, but face challenges with scalability and energy efficiency. Capacitive memories show great potential for addressing energy concerns due to their leakage-free nature. However, there is a lack of research on their scalability and robustness. In this work, we present a high-yield memcapacitor array that demonstrates reliable memory characteristics while also being capable of precisely sensing different types of vehicle motion with only a few picowatts of power consumption. Featuring a metal-oxide-semiconductor (MOS) structure with the most aggressively scaled dimensions compared to previously reported memcapacitors, we successfully established a 9 × 9 memcapacitor matrix with a yield of over 92.5%. The device exhibits tunable synaptic plasticity under varying pulsing schemes. We also demonstrate 64 distinct capacitance states and stable performance over 2 × 10⁴ electrical pulses. Additionally, we showcase its application in motion sensing for autonomous vehicles, leveraging the short-term potentiation properties of the device. This approach offers a scalable, energy-efficient solution for future motion sensing systems.

Duke Scholars

Author Tania Roy Electrical and Computer Engineering

Author Aaron D. Franklin Electrical and Computer Engineering