Interfacial AlO_x-insertion–driven synaptic enhancement and tunneling control in HfO₂-based ferroelectric MIFS memristors

Hafnia-based ferroelectric memristors are promising candidates for next-generation neuromorphic computing owing to their ability to emulate biological synaptic functions with excellent scalability, non-volatility, and CMOS compatibility. In this work, we investigate a metal–insulator–ferroelectric–semiconductor (MIFS) structure and compare it with a conventional metal–ferroelectric–semiconductor (MFS) device. Electrical characterization reveals that the MIFS device exhibits a wider memory window, a high tunneling electro-resistance (TER) ratio of ∼538 %, and superior array scalability (up to 253 × 253, as estimated by array level simulation) enabled by effective suppression of sneak-path currents. Moreover, the MIFS structure successfully reproduces essential synaptic behaviors such as paired-pulse facilitation (PPF), potentiation/depression, and excitatory postsynaptic current (EPSC), demonstrating its suitability for neuromorphic operation. When integrated as a reservoir layer in a reservoir computing (RC) system, the device achieved 97.48 % classification accuracy on the MNIST dataset, validating its potential for hardware-based neuromorphic computing. Additionally, the gradual and symmetric current modulation with stable retention enables controllable analog synaptic functionalities, highlighting the versatility of the enhanced MIFS architecture for next-generation neuromorphic systems.