鐵電氧化鉿鋯材料於非揮發性電阻式元件之未來新興記憶體應用
No Thumbnail Available
Date
2021
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
鐵電二氧化鉿鋯(Hf1-xZrxO2)材料因具有雙穩態(Bi-stable)的特性，使其能在外加偏壓為零時仍具有兩個穩定的極化狀態，此特性使得它具備成為新興非揮發性記憶體(Non-volatile memory, NVM)的潛力，預期在未來人工智慧(Artificial intelligence, AI)和類神經運算(Neuromorphic computation)的應用中扮演至關重要的角色。截至現今已經有相當多關於鐵電材料於記憶體的研究，而本論文主要探討調變HZO中摻雜鉿(Hf)與鋯(Zr)的比例，並成功開發反鐵電(Anti-ferroelectric)材料應用於電阻式記憶體元件，且在記憶體特性上皆優於正鐵電(Ferroelectric)材料。本論文第二章研究結果為反鐵電介面二極體(Anti-ferroelectric junction diode)的記憶體具備單極性操作的能力，且記憶體的開關比例(On/Off ratio)達到100倍和耐受性(Endurance)可達到109次；而第三、四章則展示雙層反鐵電穿隧式記憶體(Bi-layer anti-ferroelectric tunneling junction)具有大於100倍的On/Off ratio和大於50倍的穿隧電阻比(Tunneling electro-resistance, TER)，耐受性與資料保存性(Retention)分別可達到108次與大於104秒，並且在調控不同寫入的脈衝電壓下，顯示具有多階儲存單元的能力（Multi-level cell）與深度學習(Deep Learning)的特性，使其具備成為高密度且低功耗的非揮發性記憶體應用於類神經運算的潛力(Neuromorphic computation)。
Ferroelectric (FE) material Hf1-xZrxO2 is one of the most promising candidates for emerging non-volatile memory (NVM) in artificial intelligence (AI) and neuromorphic computation due to its two stable polarization states at zero electrical field for data storage with natural bi-stable characteristics. A variety of prototypes utilizing ferroelectricity in different storage mechanisms have been proposed. In this work, we demonstrate the FE and antiferroelectric (AFE) material properties, which are modulated from doped Zr incorporated into the HfO2-system. The result shows AFE material enhances the capacity to modulate the current ratio/TER.We demonstrate two resistive switching memory devices, (anti-)ferroelectric junction diode((A)FE Diode) and bilayer-based (anti-)ferroelectric Tunneling Junction (Bilayer (A)FTJ) in chapter 2, 3 and 4 respectively. The proposed AFE diode shows memory On/Off ratio>100 and the switching endurance reaches > 109 cycles under unipolar operations. Bilayer (A)FTJ demonstrates >100x current ratio, >50x TER, leading to possible multi-level information storage (MLC) and neuromorphic computing. The endurance behavior shows 108 switching cycles with low energy consumption and data retention>104 seconds.
Ferroelectric (FE) material Hf1-xZrxO2 is one of the most promising candidates for emerging non-volatile memory (NVM) in artificial intelligence (AI) and neuromorphic computation due to its two stable polarization states at zero electrical field for data storage with natural bi-stable characteristics. A variety of prototypes utilizing ferroelectricity in different storage mechanisms have been proposed. In this work, we demonstrate the FE and antiferroelectric (AFE) material properties, which are modulated from doped Zr incorporated into the HfO2-system. The result shows AFE material enhances the capacity to modulate the current ratio/TER.We demonstrate two resistive switching memory devices, (anti-)ferroelectric junction diode((A)FE Diode) and bilayer-based (anti-)ferroelectric Tunneling Junction (Bilayer (A)FTJ) in chapter 2, 3 and 4 respectively. The proposed AFE diode shows memory On/Off ratio>100 and the switching endurance reaches > 109 cycles under unipolar operations. Bilayer (A)FTJ demonstrates >100x current ratio, >50x TER, leading to possible multi-level information storage (MLC) and neuromorphic computing. The endurance behavior shows 108 switching cycles with low energy consumption and data retention>104 seconds.
Description
Keywords
鐵電材料, 氧化鉿鋯, 鐵電穿隧式記憶體, Ferroelectric materials, HfZrO2, Ferroelectric tunnel junction