| dc.description.abstract | With the rapid development of fields such as the Internet of Things (IoT), artificial intelligence (AI), and autonomous driving, the demand for memory technologies is steadily increasing. Among emerging memory technologies, non-volatile memory (NVM) is playing an increasingly important role. The main feature of NVM is its ability to retain stored data even when power is off, along with characteristics such as scalability, high speed, low power consumption, long lifespan, and durability. This makes NVM an indispensable part of future data storage and processing.
In this work, we focus on ferroelectric memories within NVM, specifically ferroelectric capacitors (FeCAP) and ferroelectric field-effect transistors (FeFET). Starting from process technology, we documented the fabrication process in detail and evaluated and compared these devices from the perspectives of material analysis and electrical measurement. Ultimately, we proposed a method to enhance device performance by utilizing a morphotropic phase boundary (MPB), effectively reducing defects and achieving a twofold increase in memory window. This approach improved device operating speed and reliability.
we explored improvements in the switching voltage, retention, and endurance of MPB FeFETs based on hafnium zirconium oxide (HZO) and ZrO?/HZO stacking. Using X-ray photoelectron spectroscopy (XPS), we demonstrated that MPB FeFETs effectively reduced oxygen vacancy concentration from 40.5% to 20.3%, lowered depolarization fields, and increased breakdown voltage from 4.2V to 6V, achieving low leakage current. Additionally, the heterojunction MPB FeFET exhibited stable post-read/write retention and enhanced endurance, with no failures observed even after 10? read/write cycles. Finally, using fixed pulses of ±2V to ±4V, with a width of 100 μs and a step size of 0.05V, we measured conductance in write-read operations. These results were fed into NeuroSim for machine learning, where an accuracy rate of 92% was achieved, surpassing other sample structures. This demonstrates that the experimental structure in this study contributes to improved overall performance, reliability, and durability of the device, highlighting its potential for broader applications in AI. | en_US |