First Hardware Demonstration of Fully Integrated Hafnia-based Multimodal Neuromorphic Computing Platform

초록

We demonstrate for the first time a hafnia-based multimodal neuromorphic computing (HMNC) platform via co-integration of volatile and nonvolatile FeTFTs. Material-device co-optimization of the FeTFTs enables diverse functionalities for multimodal data processing. Full hardware demonstration achieves robust performance for multimodal MNIST data (88%) and emotion/sentiment recognition (66%), offering a scalable and energy-efficient neuromorphic solution.