整合田口法與粒子群演算法應用於鐵酸鉍摻雜鈮MFIS電容器之最佳化
No Thumbnail Available
Date
2011
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
本研究主要是探討使用整合田口法的改良型粒子群演算法在鐵酸鉍摻雜鈮MFIS電容器最佳化上的應用。本論文可分為兩部分:(一)粒子群演算法整合田口方法(二)鐵酸鉍摻雜鈮MFIS電容器之最佳化。
粒子群演算法是近年來應用在諸多領域的最佳化技術。全域最佳型(gbest)和區域最佳型(lbest)是粒子群演算法的其中兩種變型,其分別擁有收斂性與探索性的優點。整合田口方法可結合兩者的優勢,使新衍生的變型兼具更好的最佳化效率和更好的精確度。此變型一開始先採全域最佳型快速收斂,接著再採用區域最佳型的探索能力,當最佳化效果不彰時,再使用田口法,自群體中萃取出具有潛力的元素,形成群體學習的對象,間接加強了群體的最佳化能力。實驗結果以t檢定驗證此改良型粒子群演算法的確結合了此兩種傳統方法的各自優點,在15個適應性函數的條件下展現擁有更好的表現。
我們將此整合田口法的粒子群演算法變型應用在鋁/鐵酸鉍摻雜鈮/二氧化鉿/p型矽MFIS結構之電容器的最佳化上,以期得到最佳的製程配方。鐵電材料因其特殊的鈣鈦礦結構,很適合當作記憶體單元的材料。其中鐵酸鉍因具有高居禮溫度、高尼爾溫度、低結晶溫度和很大的殘留極化值的優點,所以成為一種很具前景的記憶體材料。唯其漏電流太大的缺點仍待改善。藉由摻雜鈮可解決此問題,最終的目標是產生具有最大記憶視窗寬度和最小漏電流密度的電容結構。考量最大記憶視窗寬度與最小漏電流密度的情況,可得最佳化後的配方:鈮摻雜直流濺鍍瓦數15.5watt、氧化層厚度69.2nm、氬氧比17.3、快速熱退火850°C。
This study is about the application of improved particle swarm optimization (PSO) integrating Taguchi method over Nb-doped BiFeO3 MFIS capacitors. This thesis has two main subjects: (1) The variant of PSO integrating Taguchi method. (2) The optimization of Nb-doped BiFeO3 MFIS capacitors. PSO has been a popular optimization technique applied over many fields. The two of PSO variants, gbest and lbest, are reported to have the advantages of exploratory capability and exploitability, respectively. Integrating Taguchi method combines these two advantages for a newly-derived PSO variant with better efficiency and less error. The novel variant proceeds with gbest for fast convergence until the shrinking of the swarm stops. lbest succeeds the following optimization to the occurrence of deadlock. Then, the Taguchi method helps to extract best recipe from the swarm to continue the optimization. The experimental results are analyzed with t-test. The superiority of this variant has been verified under fifteen fitness functions. This proven PSO variant is utilized over Al/ Nb-doped BiFeO3/HfO2/p-Si capacitors for fabrication recipe. Ferroelectric materials are suitable for being memory cells with its unique “Perovskite” structure. BiFeO3 is one of the promising substitutes with high Curie temperature, high Neel temperature, low crystallization temperature, and large remnant polarization. But the major issue is its relatively large leakage current. Doping Nb can suppress the leakage. The larger memory window and the less leakage current is the contribution of this study. The optimized recipe is 15.5 W for DC power of Nb sputtering, 69.2 nm for insulator thickness, 17.3 for argon-to-oxyen ratio, and 850°C for RTA temperature.
This study is about the application of improved particle swarm optimization (PSO) integrating Taguchi method over Nb-doped BiFeO3 MFIS capacitors. This thesis has two main subjects: (1) The variant of PSO integrating Taguchi method. (2) The optimization of Nb-doped BiFeO3 MFIS capacitors. PSO has been a popular optimization technique applied over many fields. The two of PSO variants, gbest and lbest, are reported to have the advantages of exploratory capability and exploitability, respectively. Integrating Taguchi method combines these two advantages for a newly-derived PSO variant with better efficiency and less error. The novel variant proceeds with gbest for fast convergence until the shrinking of the swarm stops. lbest succeeds the following optimization to the occurrence of deadlock. Then, the Taguchi method helps to extract best recipe from the swarm to continue the optimization. The experimental results are analyzed with t-test. The superiority of this variant has been verified under fifteen fitness functions. This proven PSO variant is utilized over Al/ Nb-doped BiFeO3/HfO2/p-Si capacitors for fabrication recipe. Ferroelectric materials are suitable for being memory cells with its unique “Perovskite” structure. BiFeO3 is one of the promising substitutes with high Curie temperature, high Neel temperature, low crystallization temperature, and large remnant polarization. But the major issue is its relatively large leakage current. Doping Nb can suppress the leakage. The larger memory window and the less leakage current is the contribution of this study. The optimized recipe is 15.5 W for DC power of Nb sputtering, 69.2 nm for insulator thickness, 17.3 for argon-to-oxyen ratio, and 850°C for RTA temperature.
Description
Keywords
鐵酸鉍, MFIS電容器, 粒子群演算法, 田口方法, BiFeO3, MFIS capacitors, particle swarm optimization, Taguchi method