Enhanced synaptic performance in hafnia-based ferroelectric memristors with MIFS structure for neuromorphic computing

Hafnia-based ferroelectric memristors show great potential for neuromorphic computing by emulating artificial synaptic behavior. These devices offer excellent scalability and CMOS compatibility, with performance further enhanced by techniques such as aluminum doping and dielectric layer insertion. In this study, we investigate the metal-insulator-ferroelectric-semiconductor (MIFS) structure and compare it with the conventional metal-ferroelectric-semiconductor (MFS) structure. Electrical measurements revealed that MIFS exhibits a wide memory window, a high tunneling electro-resistance (TER) ratio of ∼911 %, and superior array scalability (up to 145 × 145) due to reduced sneak-path currents. Furthermore, the MIFS structure demonstrates biologically inspired synaptic behaviors such as potentiation/depression, excitatory postsynaptic current (EPSC), and paired pulse facilitation (PPF). When applied to machine learning tasks using fashion modified national institute of standards and technology (Fashion MNIST) and Canadian institute for advanced research 10 (CIFAR10) datasets, the device achieved high classification accuracy, confirming its viability for neuromorphic applications.