In-Sensor Computing with Contactless MEMS: Binary Neurons for Harsh Environment

In-sensor computing remains limited in harsh environments due to the thermal and radiation sensitivity of conventional electronics. While MEMS microrelays enable high-temperature operation, their reliability is insufficient for the switching rates required by computing tasks. This work presents the first contactless MEMS-based binary neuron that performs weighted summation and thresholding within a single device. Based on electrostatic actuation and a tunable capacitive divider, neurons communicate electrically without any CMOS circuitry. Experimental validation demonstrates stable operation up to 723 K and under 40 MeV irradiation, with functionality up to 10 kHz and a compact <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.3 \text{mm}^{2}$</tex> footprint. This approach paves the way for scalable, robust neuromorphic processing for vibration pattern recognition, enabling seamless integration into MEMS sensor platforms for extreme-environment deployment.