Gradient Zn-doped HfCeOx ferroelectric synapse with fast domain switching

Ferroelectric synaptic devices are pivotal in artificial intelligence applications, particularly for image recognition tasks. Their outstanding capabilities in neuromorphic information processing, biologically plausible signal emulation, and faithful neural mimicry render them indispensable. However, for HfO 2 -based ferroelectric synaptic devices, concurrently attaining high remnant polarization and rapid domain switching which are essential for high-speed processing and precise pattern recognition remains a formidable challenge. In this study, gradient Zn-doped HfCeO x (Zn: HCO) ferroelectric film was fabricated. It displays enhanced ferroelectric properties with a remnant polarization (2 P r ) reaching 77.2 µC/cm 2 . High-resolution transmission electron microscopy analyzes and phase-field simulations reveal that the gradient Zn doping induces a flexoelectric effect in HfCeO x thin films, which promotes the formation of the ferroelectric orthorhombic phase. Additionally, this gradient design accelerates domain switching speed by enhancing the domain wall mobility under applied electric fields. To demonstrate the neuromorphic computing potential, we implemented an artificial neural network using Pt/Zn: HCO/p + -Si synaptic devices, achieving an 86.5 % recognition accuracy in standard pattern classification tasks. This work establishes a viable strategy for simultaneously enhancing both ferroelectric polarization and switching speed in HfO 2 -based memory and neuromorphic devices.