Date Published: April 14, 2018
Publisher: John Wiley and Sons Inc.
Author(s): Xing Wu, Kaihao Yu, Dongkyu Cha, Michel Bosman, Nagarajan Raghavan, Xixiang Zhang, Kun Li, Qi Liu, Litao Sun, Kinleong Pey.
Higher memory density and faster computational performance of resistive switching cells require reliable array‐accessible architecture. However, selecting a designated cell within a crossbar array without interference from sneak path currents through neighboring cells is a general problem. Here, a highly doped n++ Si as the bottom electrode with Ni‐electrode/HfOx/SiO2 asymmetric self‐rectifying resistive switching device is fabricated. The interfacial defects in the HfOx/SiO2 junction and n++ Si substrate result in the reproducible rectifying behavior. In situ transmission electron microscopy is used to quantitatively study the properties of the morphology, chemistry, and dynamic nucleation–dissolution evolution of the chains of defects at the atomic scale. The spatial and temporal correlation between the concentration of oxygen vacancies and Ni‐rich conductive filament modifies the resistive switching effect. This study has important implications at the array‐level performance of high density resistive switching memories.
Nanocrossbar arrays comprise a set of parallel bottom electrodes and perpendicular top electrodes with a thin layer of resistive switching material in between. Switching materials can be classified into different groups according to their physical mechanisms.1, 2, 3 Materials showing electrochemical metallization or valence change (VCM) effects have been extensively investigated.4, 5 Each crosspoint in a memory cell stores logic information as a high resistance state (HRS) or a low resistance state (LRS).6, 7, 8 A selection device, such as a transistor or a diode, is usually necessary in addition to the memory element for the scaling limit of the memory device. Therefore, the simplicity of the geometrical structure and the absence of transistors make the concept extremely interesting for low‐power nonvolatile memory circuitry and high integration. Conductive filament (CF) has been recognized as the key structural element that contributed to resistive switching performance.6, 7, 9, 10, 11 Recently, a number of approaches, such as electrochemical redox reaction,12, 13 metal nanodots doping,14 micrsostructural transitions,15, 16 and current compliance capping17 have been explored to address the controlled formation and rupture mechanism of CFs.
The device used in this study was a unipolar resistive switching device based on an asymmetric metal–insulator–semiconductor (MIS) structure: Ni as the top electrode, HfO2 as the insulator, and Si substrate as the bottom electrode. Ni has been used extensively in the mainstream complementary metal oxide semiconductor technology as a source/drain contact material.23 The bottom metal electrode was replaced with a highly doped n++‐type Si (ND ≈10−19 cm3) to study the role of asymmetric electrode in switching. It comprised of no additional diode and only one resistive switching element (0D1R), which allowed for the construction of highly dense passive crossbar arrays by solving the sneak path problem combined with the drastic reduction of power consumption and area.
In summary, a high‐performance diode‐free resistive switching cell was fabricated with a high on/off ratio. The real‐time in situ TEM analysis of asymmetric MIS structures allowed the study of the dynamics, evolution, and underlying physics governing the switching mechanism, which was previously speculated at best from electrical measurements without much direct evidence. Multiple conductive path formation and rupture events were temporally and spatially uncorrelated. The use of an asymmetric MIS structure clearly assisted in identifying the presence and source of the metal‐rich conductive filament, which originated from the anode electrode while preventing sneak path issues. This fundamental atomic study on the chemistry, morphology, and time‐dependent correlated/uncorrelated switching behavior of filaments provides a strong support to the feasibility of further scaling of future resistive switching technologies, paving the way for very high‐density data storage and future neuromorphic computing.
Electrical Characterization: The electrical performance of the device was determined ex situ using a probe station with the standard Keithley SCS‐4200 semiconductor characterization system at room temperature.
The authors declare no conflict of interest.