斜角蒸鍍系統製備奈米種子應用於可控制多重導電絲之電阻式記憶體

Controlling Multiple Filaments by Embedding Nano-Seed Layer in Resistive Random Access Memory via Glancing Angle Deposition

在這個大數據時代，傳統電腦計算方式將導致延遲和大量的能量消耗；神經形態計算的出現可以加快數據移動並減輕能源消耗。電阻開關切換的記憶體具有開關速度快、功耗低和尺寸縮小等優點，在下一代非揮發式記憶體和神經形態計算系統中具有高潛力的應用。然而，在燈絲狀電阻式記憶體中實現神經形態計算是很困難的。研究表明，多燈絲型電阻式記憶體有利於用於類比訊號及神經元計算。在此，我們開發了一種奈米種子功能層嵌入電阻式記憶體，以限制導電燈絲路徑並實現多重燈絲型電阻式記憶體。通過斜角沉積來製備奈米種子氧化鋁結構，並通過原子層沉積製作氧化鉿層，用作主要轉換層。相較於具有約50 nm強導電細絲的純氧化鉿層記憶體，植入奈米種子氧化鋁層的氧化鉿層記憶體限制了幾個10-12奈米級的導電細絲。此外，這種奈米種子氧化鋁層的氧化鉿層記憶體展示了多位元特性和雙向類比訊號行為。這種新穎的架構可為神經形態計算系統中實現人工突觸提供一個方向。

Accessing data from the memory units to processing units is a main bottleneck in conventional von Neumann computing, which will give rise to latency and large energy consumption in this era of data explosion. A non-von Neumann computational approach, neuromorphic computing, has been come out recent year to speed out the data movement and alleviate the energy consumption. Resistive switching (RS)-based memristors possess several advantages such as fast switching speed, low power consumption and shrinking size that serve high potential applications in the next generation nonvolatile memory and neuromorphic computing system. However, it is difficult to realize the neuromorphic computing in filamentary resistive random access memory (RRAM). Studies have shown that multiple-filaments type RRAM is beneficial for analog switching. Herein, we developed a nano-seed functional layer embedded memory to confine the conducting filament path and achieve multiple weak filaments type RRAM, which is suggested for realizing analog behavior. Nano-seed Al2O3 architecture was fabricated by glancing angle deposition to confine the oxygen vacancy regions. HfO2 layer was fabricated by atomic layer deposition and served as the main insulator layer. The nano-seed Al2O3 layer embedded HfO2-based memory restricts several conducting filaments of 10–12 nm, which are much thinner than pure HfO2-based memory with a strong conducting filament of around 50 nm. Besides, this nano-seed Al2O3 layer embedded HfO2-based memory demonstrates multilevel cell characteristic and bi-directional analog switching behavior. This novel architecture is believed to provide a direction for realizing artificial synaptic devices in implementation of neuromorphic computing systems.