鐵電電晶體之類比式操作與後段製程相容之設計
No Thumbnail Available
Date
2022
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
為達到人工智能（AI）之物聯網（IoT）及高速傳輸之5G/6G科技，高密度的記憶體內/近運算高度需求。近年來各方領域的近記憶體運算與記憶體內建邏輯紛紛被提出，利用各種新興非揮發性記憶體(Emerging non-volatile memory, e-NVM) ，以實現內部存取並執行邏輯操作減少耗時與耗能的問題。本論文便是討論鐵電電晶體之類比式操作與後段製程相容之設計。研究中採用直流 (DC) 掃描、脈衝測量、Endurance和Retention的方法來研究元件特性。因此，第二章會介紹實驗的測量設備和波形設置。在第三章中，驗證雙 HZO 鐵電場效應電晶體 (FeFET) 可多階操作 (MLC)以 提高NVM密度。與單HZO FeFET 相比，金屬層/鐵電層/金屬層/鐵電層/矽基板 (MFMFS) FeFET 能夠在 ±3 V 的超低寫入/抹除電壓 (VP/E) 下實現2-bit位操作，並具有穩定的數據保持能力>104秒和>107次循環的耐用性。此外，通過使用金屬層/鐵電層/介電層/鐵電層/矽基板結構將記憶窗戶(MW)擴大至2.6 V，讀取錯誤率(ER)比單HZO低600倍。兩種雙HZO FeFET都通過使用電壓調整的方案展示具有高度線性和對稱性的深度學習能力。
在第四章中，完成具有>106高開關電流比(Ion/Ioff)和4cm2/V⸳s 遷移率的無退火 In2O3 薄膜電晶體 (TFT)，採用20sccm的Ar和15 W的濺射系統沉積。最後將鐵電電容與In2O3-TFT串聯，成功觀察到磁滯特性，並完成FE電容與In2O3-TFT的面積比對磁滯差異進行實驗驗證。因此，In2O3-TFT 有望在未來與鐵電記憶體整合，用於後端製程 (BEOL)。
In order to achieve internet of thing (IoT) for artificial intelligence (AI) and high-speed communication with 5G/6G, the high-density in/near-memory-computing is urgently demanded. Recently, the in/near-memory-computing for kinds of Emerging non-volatile memory (eNVM) were proposed to reduce power consumption and time. In this thesis, the design of ferroelectric Field-Effect-Transistor (FeFET) for analog operation and Back end of line (BEOL) compatible process will be investigated.In Chapter 2, the measurement tools and technique will be introduced, such as direct current (DC) sweep, pulse measurement, endurance, and retention, as well as waveform setup for following experiment.In chapter 3, the double-HZO ferroelectric field-effect transistors (FeFETs) exhibits multilevel cell (MLC) to enhance NVM density. As compare to single-HZO FeFETs, metal/ferroelectric/metal/ferroelectric/Si (MFMFS) FeFETs has achieved 2-bit operation under ultra-low program/erase voltage (VP/E) of ±3 V with stable data retention of>104 s and robust endurance of 107 cycles. In addition, the enhancement of memory window (MW) with 2.6 V by using metal/ferroelectric/insulator/ferroelectric/Si leads to 600x lower error rate (ER) than single-HZO. Both double-HZO FeFETs demonstrated the capability of deep neural network (DNN) with nonlinearity and symmetry synaptic operation by using amplitude modulation scheme. In chapter 4, In2¬O¬3 thin film transistors (TFTs) with annealing-free exhibits high on/off-state current ratio of >106 with mobility of 4 cm2/V⸳s.¬, which fabricated by sputtering system with 20 sccm of Ar and 15 W. The proposed In2O3-TFT connected with ferroelectric capacitor in series to validate hysteretic characteristics. The Vt related to the area ratio of FE capacitor and In2O3-TFT. The In2O3-TFT is promising to integrated for ferroelectric memory toward back end of line (BEOL) in the future.
In order to achieve internet of thing (IoT) for artificial intelligence (AI) and high-speed communication with 5G/6G, the high-density in/near-memory-computing is urgently demanded. Recently, the in/near-memory-computing for kinds of Emerging non-volatile memory (eNVM) were proposed to reduce power consumption and time. In this thesis, the design of ferroelectric Field-Effect-Transistor (FeFET) for analog operation and Back end of line (BEOL) compatible process will be investigated.In Chapter 2, the measurement tools and technique will be introduced, such as direct current (DC) sweep, pulse measurement, endurance, and retention, as well as waveform setup for following experiment.In chapter 3, the double-HZO ferroelectric field-effect transistors (FeFETs) exhibits multilevel cell (MLC) to enhance NVM density. As compare to single-HZO FeFETs, metal/ferroelectric/metal/ferroelectric/Si (MFMFS) FeFETs has achieved 2-bit operation under ultra-low program/erase voltage (VP/E) of ±3 V with stable data retention of>104 s and robust endurance of 107 cycles. In addition, the enhancement of memory window (MW) with 2.6 V by using metal/ferroelectric/insulator/ferroelectric/Si leads to 600x lower error rate (ER) than single-HZO. Both double-HZO FeFETs demonstrated the capability of deep neural network (DNN) with nonlinearity and symmetry synaptic operation by using amplitude modulation scheme. In chapter 4, In2¬O¬3 thin film transistors (TFTs) with annealing-free exhibits high on/off-state current ratio of >106 with mobility of 4 cm2/V⸳s.¬, which fabricated by sputtering system with 20 sccm of Ar and 15 W. The proposed In2O3-TFT connected with ferroelectric capacitor in series to validate hysteretic characteristics. The Vt related to the area ratio of FE capacitor and In2O3-TFT. The In2O3-TFT is promising to integrated for ferroelectric memory toward back end of line (BEOL) in the future.
Description
Keywords
鐵電材料, 多階記憶體, HfZrO2, 氧化銦, ferroelectric material, multibit, multilevel cell, HfZrO2, indium oxide