以過渡金屬氧化物為基礎的新世代非揮發性電阻式記憶體

Abstract

傳統的浮動閘極元件於40奈米以下非揮發性記憶體的應用上預期將遇到很大的障礙，而在眾多新世代記憶體技術中，以過渡金屬氧化物 (TMO) 為基礎的電阻式記憶體 (RRAM) ，利用簡單的兩極記憶單元、三維堆疊與多位準架構，是最具潛力實現兆位元級的超高密度非揮發性記憶體技術，然而在實際運用上仍有許多關鍵問題需要被解決，例如對阻質切換機制缺乏全面的瞭解、重置電流的持續降低、重複覆寫過程中的高變異性、及利用於1D1R架構中的高性能氧化物二極體的開發等等，本計畫中將針對上述當前RRAM技術的發展瓶頸做一深入與完整的探討。 我們將以雙層、複合層、雜質摻雜、梯度組成的不同TMO結構來有效控制氧原子缺陷的分佈，以導電性原子力顯微鏡與歐傑電子能譜儀，對局部絲狀傳導路徑的形成與破裂做最直接的觀察，並搭配電子自旋共振能譜儀對電性活化缺陷的分析，預期將可以對TMO 與金屬電極間的介面層在RRAM切換機制上所扮演的關鍵角色有更深一層的瞭解。同時我們也將調變異質的氧化物介面以及金屬與氧化物間的蕭基能障，開發出一能與單極切換特性RRAM搭配的高順向電流、高開關比的氧化物二極體。另一方面，我們將以電遷移及氧化層崩潰理論為基礎，瞭解在高電場、高電流、高溫的情況之下，在異質的金屬/TMO介面上，離子和電子間複雜的交互作用與移動，並以此建立一量化的RRAM物理模型，做為日後持續改建元件特性的藍圖。

The conventional floating-gate devices are expected to face tremendous scaling challenges for beyond 40nm nonvolatile memory applications. Among many alternative memory technologies, the transition-metal-oxide (TMO)-based resistive-switching random access memory (RRAM), owning to its simple two-terminal unit cell structure, easy three-dimensional stacking, and possible multi-level-cell implementation, has great potential to truly realize high-density terabit nonvolatile storage. However, there are several critical issues remained before RRAM emerges as the mainstream technology. For instance, the lack of physical understanding of resistive switching mechanism, the continuous reduction of reset current, the high variations on device parameters during repeated cycling, and the development of a high-performance oxide-based diode for the one-diode-one-resistor (1D1R) architecture. In this project, we will provide an in-depth study focusing on the above-mentioned key bottlenecks in the RRAM technology. Toward better comprehension on the pivotal role of the TMO/metal interface during the resistive switching, we will manipulate the distribution of oxygen vacancies through different bilayer, multilayer, impurity doping, and gradient TMO structures. Conductive AFM and Auger electron spectroscopy will be utilized to monitor the forming and rupture of localized conductive filaments. Moreover, electron spin resonance spectroscopy will be applied to probe the electrically active defects in TMO. Meanwhile, compatible with the unipolar switching and 1D1R architecture, a high-forward-current, high-on/off-ratio oxide diode will be developed through engineering the heterogeneous oxide interface and the Schottky barriers between TMO and metal electrodes. On the other hand, based on the solid theoretical foundation of eletromigration and oxide breakdown, we will also establish a quantitative RRAM physical model illustrating the complex interaction between ions and electrons under the high current, high electric field, and high temperature circumstances at the TMO/metal interface. This model will be proven invaluable in guiding the future RRAM technology development.