Ferroelectric FETs using hafnium zirconium oxide thin films and its applications

Abstract

The semiconductor market continues to grow, leading to personal computers and mobile phones, and is now entering a new phase with the emergence of the internet of things (IoT), autonomous vehicles, and artificial intelligence (AI) technologies. In particular, application processors required for various portable devices are attempting to incorporate non-volatile memory or neuromorphic systems to meet the needs of new markets. In the meantime, research on embedding next-generation memory such as RRAM, PRAM, and MRAM into the application processor has been carried out, but it is still difficult in terms of stable operation and mass production possibility. Therefore, the development of a new memory device compatible with the logic process can be regarded as a research topic suitable for the market demand in recent years. This research covers the development of ferroelectric field-effect transistors (FeFETs) for non-volatile memories and neuromorphic systems, utilizing ferroelectric hafnium zirconium oxide (HZO) thin films that can be easily embedded in the application processors. Highly scalable devices based on junctionless (JL) ferroelectric fin and gate-all-around (GAA) FETs are presented. The memory and synaptic behaviors of the JL metal-ferroelectric-insulator-silicon (MFIS) FeFETs were experimentally demonstrated after verifying ferroelectric characteristics of the HZO thin films using metal-ferroelectric-metal (MFM) capacitors. The fabricated device showed distinguishable polarization switching behaviors with controllable channel conductance. From neural network simulations based on the synaptic characteristics of the proposed JL ferroelectric FinFET, the pattern recognition accuracy for hand-written digits was validated to be approximately 80% for neuromorphic applications. The results of this research are expected to play an important role in easily integrating non-volatile memory and neuromorphic systems for brain-inspired computing into the next-generation application processors.