| 摘要: | 在本次實驗中我們提出了一種利用氫電漿處理元件的方法，此方法提升了元件的穩定性、操作速度和可靠性。我們透過 XPS、TEM 等物性分析鐵電薄膜與各元件的材料差異，並針對氫電漿處理前後元件的 ION/IOFF 比、極化切換速度、耐久性、數據保留性、多層儲存單元操作等多項電性數據進行比較分析。此外，我們還透過將 1N-1P 的兩個鐵電電晶體組合成一個 FeCAM cell，並同樣對其進行電性及物性分析。透過多項分析，我們發現在 FeFET 方面氫電漿處理後介面缺陷密度由 8.04 × 1012????/????2 降低到了 5.5 × 1012????/????2，這就使得元件 ION/IOFF 比可以達到 8 個級距，比原本高出了兩個級距，漏電電流也明顯小於處理前 1~2 個級距。並且透過 XPS 發現 HZO 薄膜中的氧缺陷濃度降低了將近 50-60%，這將改善由氧空缺引起的電荷捕捉效應，可使元件 wake up free，並將耐久性提高到超過 1011次。此外，氫電漿處理降低了 pinning field，加速了極化切換，並減少了印痕效應。改善了多層儲存單元的操作，由原本 TLC 提升至 QLC，並且在變異性改善方面提升了 46%。透過將測試脈波、Gmax/Gmin、狀態數等參數輸入 MNIST 系統中，其手寫辨識度達到了 91.1%。而 FeCAM 方面氫電漿處理後的 VOFF 的標準差從64mV 降到 38mV，並且 MHD 甚至是原本的 6 倍，此外，Arrhenius 方程擬合結果顯示，該 FeCAM 在 99.5°C 下仍能保持十年的數據。;In this study, we proposed a method utilizing hydrogen plasma treatment to enhance the stability, operating speed, and reliability of devices. We performed material analysis on ferroelectric thin films and various components using XPS, TEM, and other techniques, comparing the electrical characteristics of the devices before and after hydrogen plasma treatment, including the ION/IOFF ratio, polarization switching speed, endurance, data retention, and multi-level cell (MLC) operation. Additionally, we combined two FeFETs (1N-1P) to form a FeCAM cell, conducting both electrical and material analysis.Through various analyses, we found that the interface trap density in FeFETs was reduced from 8.04 × 1012 μC/cm2 to 5.5 × 1012 μC/cm2 after hydrogen plasma treatment. This allowed the device′s ION/IOFF ratio to reach eight levels, which is two levels higher than before, with significantly reduced leakage current (1-2 levels lower). XPS also revealed that the oxygen vacancy concentration in the HZO film decreased by nearly 50-60%, improving the charge trapping effect caused by oxygen vacancies. This resulted in wake-up free devices and extended endurance to over 1011 cycles. Furthermore, hydrogen plasma treatment reduced the pinning field, accelerated polarization switching, and minimized imprint effects. It also enhanced MLC operation from TLC to QLC, with a 46% improvement in variability.By inputting parameters such as test pulses, Gmax/Gmin, and the number of states into the MNIST system, recognition accuracy reached 91.1%. Regarding FeCAM, after hydrogen plasma treatment, the standard deviation of VOFF was reduced from 64mV to 38mV, and the maximum Hamming distance (MHD) was six times greater than before. Additionally, the Arrhenius equation fitting results showed that the FeCAM could retain data for ten years at 99.5C. |
