| 摘要: | 鐵電記憶體(FeRAM)因其優異的耐久性及操作速度能彌補馮·紐曼瓶頸的缺點，得益於這樣的優點使得在進行龐大數據計算能夠加速資料的處理，使成為人工智慧運算中有前瞻性的元件。目前鐵電記憶體廣泛使用的材料為氧化鉿鋯鐵電薄膜(HfO2-ZrO2, SL-HZO) ，因其展現出優越的極化特性、較小的變異性以及較高的切換速度，相較於固溶體 HfxZr1-xO2 (SS-HZO) 而受到關注。儘管已有研究表明晶格扭曲可能導致強極化，然而在超晶格結構中，週期膜厚與變異性、熱穩定性及保持特性衰退之間的關聯性仍缺乏全面的物理解釋。在本研究中，我們通過實驗的XPS 和電學測量觀察到這些行為，並結合從頭算分子動力學 (AIMD)、過度態 (NEB) 方法及COHP方法透過第一原理DFT計算，探討這些可靠性問題的根本原因。通過比較 HZ2.5, HZ5, HZ7.5 和HZ10 的缺陷形成（變異性）及熱穩定性，我們發現：(i) 較低的tp且有序的 Hf-O 與 Zr-O 原子排列可以縮短鍵長，防止扭曲傳播至材料內部，且聲子傳輸對於高溫保持有利。(ii) 較大的tp使SL-HZO能夠承受更大的應力偏壓；然而，Zr-O 鍵的失穩會累積缺陷，導致掃描操作中的高變異性。實驗表明，埃層堆疊的好處超過了奈米層堆疊的好處。
目前氧化物通道材料材料已經成為研究的焦點，因其中以銦鎵鋅氧化物 (IGZO)特別突出，IGZO由於通道與氧化物之間的介面層幾乎為零介電常數(Junction less)，因此並不會與之前使用矽通道的電性下降的問題，然而，隨著應用領域的不斷擴展和發展，傳統的IGZO面臨挑戰，因為在高解析度顯示器和 3D NAND 中應用時，需要更高的遷移率和保有原有的低關斷電流及優異的覆蓋性。本研究通過第一原理計算比IGZO更高遷移率及更容易微縮的二元化合物之N型半導體通道材料－銦氧化物(In2O3)，經過計算得出此材料單層膜厚的能隙高達2.27eV，功函數為4.2~5.3eV，並且實驗量測到的遷移率高達90 cm2V-1s-1。除此之外為了因應不同的應用，如設計雙極性電晶體、逆變器電路和透明薄膜電晶體，而先前NiO與Cu2O經過廣泛研究，但其遷移率都不超過100 cm2V-1s-1，因此研究具有高遷移率的寬能隙P型半導體也是迫切的，在此以二維氧化碲(TeO2)作為研究對象，其計算結果表明單層能隙高達3.79eV，功函數為3.9~4.3eV，並且遷移率高達200 cm2V-1s-1。本研究為高性能鐵電記憶體應用的實際設計指南提供了機會。
;Ferroelectric memory (FeRAM) can overcome the von Neumann bottleneck due to its excellent durability and operational speed, which enables faster data processing for large-scale calculations, making it a forward-looking component in artificial intelligence computation. Currently, the widely used material for ferroelectric memory is hafnium zirconium oxide thin films (HfO2-ZrO2, SL-HZO), which have attracted attention due to their superior polarization properties, lower variability, and higher switching speed compared to solid solution HfxZr1-xO2 (SS-HZO). Although studies have shown that lattice distortion may lead to strong polarization, a comprehensive physical explanation is still lacking for the relationship between periodic film thick-ness, variability, thermal stability, and retention decay in superlattice structures. In this study, we used XPS and electrical measurements to observe these behaviors and explored the root causes of these reliability issues through first-principles DFT calculations, using ab initio molecular dynamics (AIMD), nudged elastic band (NEB) method, and crystal orbital Hamilton population (COHP) analysis. By comparing defect formation (variability) and thermal stability in HZ2.5, HZ5, HZ7.5, and HZ10, we found: (i) Lower tp and an ordered arrangement of Hf-O and Zr-O atoms can shorten bond lengths, preventing distortion from propagating into the material, and phonon transport favors high-temperature retention. (ii) Larger tp allows SL-HZO to withstand higher stress biases; however, the instability of Zr-O bonds accumulates defects, resulting in higher variability during sweep operations. Experiments indicate that the benefits of Å-level stacking exceed those of nanolayer stacking.
Oxide channel materials have become a focal point of research, with indium gallium zinc oxide (IGZO) standing out in particular. IGZO is junction-less due to the near-zero dielectric constant at the interface layer between the channel and oxide, thereby avoiding the electrical degradation issues associated with silicon channels. However, as applications continue to expand and develop, traditional IGZO faces challenges in high-resolution displays and 3D NAND, where higher mobility, low off-current, and excellent coverage are essential. This study uses first-principles calculations to investigate an n-type semiconductor channel material, indium oxide (In2O3), a binary compound with higher mobility and easier scalability than IGZO. Calculations show that this material has an energy gap of 2.27 eV for a monolayer film thickness and a work function of 4.2–5.3 eV, with experimentally measured mobility reaching up to 90 cm²V⁻¹s⁻¹. Additionally, to accommodate different applications, such as the design of bipolar transistors, inverter circuits, and transparent thin-film transistors, we focused on high-mobility wide-bandgap p-type semiconductors. While NiO and Cu₂O have been widely studied, their mobilities do not exceed 100 cm²V⁻¹s⁻¹. In this study, we investigated two-dimensional tellurium oxide (TeO₂), which has a monolayer energy gap of 3.79 eV, a work function of 3.9–4.3 eV, and a mobility of up to 200 cm²V⁻¹s⁻¹. This study provides a practical design guideline for high-performance ferroelectric memory applications. |
