Ferroelectric Memcapacitive Dynamics from Nanoseconds to Milliseconds for Bio-Inspired Neuromorphic Computing and Control

Abstract

Achieving sub-volt, nanosecond-to-millisecond synaptic plasticity in a scalable ferroelectric platform remains a key challenge for neuromorphic hardware. Such capability is essential to replicate the multi-timescale efficiency of biological computation and enable real-time, low-power edge intelligence. Here, we present a memcapacitive platform based on sol-gel-derived Zr-doped HfO2 (HZO) thin films that integrates ferroelectric switching with spontaneous depolarization to realize fading memory, nonlinear signal amplification, and multi-timescale synaptic plasticity. The devices operate at sub-volt bias with coercive voltages of ±0.6 V, memory windows of 20-150 mV, and ultrafast transient responses down to 500 ns, supporting energy-efficient temporal information encoding. System-level integration enables a hardware-software co-design framework for biomedical signal processing, achieving 97.7 % accuracy in electrocardiogram classification across five cardiac rhythm categories. Beyond classification, end-to-end robotic manipulator control exploits threshold-dependent weak, moderate, and strong actions derived directly from device transients, emulating biological neuromuscular responsiveness. These results position HZO-based memcapacitive dynamics as a compact and robust platform for low-power neuromorphic computing, bridging ferroelectric physics with practical applications in real-time diagnostics, adaptive robotics, and edge intelligence.