Direct Observations of Nanofilament Evolution in HfO2-Based RRAM Device
Diverse resistive switching (RS)-based devices such as resistive random access memory (RRAM) have been demonstrated to have promising potential applications in integrated circuits and internet of things industries due to the emerging non-volatile memory, bio-inspired neuromorphic devices, hardware security solution, etc. Understanding the microscopic mechanism of RS behavior is crucial for designing and optimizing the relevant electronic devices, propelling the theoretical research progress of the nanoscale condensed matter. So far, great efforts have been made to clarify the switching mechanisms. However, there is still a lack of details about the evolution of the conductive filament (CF) in oxygen vacancy type RS prototypes due to the difficulty in directly observing the oxygen vacancy defects. Up to now, several fundamental issues of the elusive switching mechanism in the transition metal oxides have been still under debate.
In situ electron holography has become a powerful tool to image the electrostatic potential distribution in a semiconductor device in TEM. The charges accumulated in the biased sample can change the phase of the transmitted electron wave, and such a phase disturbance can be retrieved from the electron interference patterns. So the formations and ruptures of the CFs in the valence change memory may be revealed in real space by the in situ electron holography. Moreover, in situ low energy-filtered images can describe the oxygen concentration changes in the hafnia films. Recently, Ph. D student LI Chao and Dr. YAO Yuan in Prof. YU Richeng’s group from Institute of Physics, Chinese Academy of Sciences, Dr. GAO Bin in Prof. KANG Jinfeng’s group from Peking University, and Prof. LI Junjie and Prof. GU Changzhi et al. from Institute of Physics, Chinese Academy of Sciences have directly observed nanofilament evolution in switching processes in HfO2-based RRAM device by in situ TEM studies.
The researchers found that the CFs are formed due to the fact that the oxygen vacancies are generated and ruptured at the top interface of the HfOx film. In the forming process, positive bias produces a lot of oxygen vacancies at the top interface but some vacancies also gather at the bottom interface, accompanied with injected electrons. As the positive bias increases, the oxygen vacancies multiply and diffuse in the HfOx film. Finally they form the continuous channels connecting two electrodes where the electrons can drift more easily and reach the anode. The oxygen vacancy channels may be broad but the bottlenecks of the CFs are near the top electrode (TE). In the RESET process, the CFs may shrink from the top interface and rupture near the TE side. These pictures reveal the details of the bipolar switching behaviors and clarify the physical mechanism behind the conventional I-V measurement.
This study entitled “Direct Observations of Nanofilament Evolution in Switching Processes in HfO2-Based Resistive Random Access Memory by In Situ TEM Studies” was published on Adanced Materials and was selected as cover picture.
The study was supported by the National Science Foundation, the Ministry of Science and Technology of China, and the Chinese Academy of Sciences.
|Fig. 1 Bias-induced phases featuring ∆φbias(x, y) under different positive biases in the forming process, where the intrinsic inner potential is removed by using the phase image of the pristine sample without any bias. The images are retrieved from the holograms taken at zero voltage after the corresponding bias setting. Only the HfOx film is highlighted here. Dark red is for higher potential while dark blue is for the lower potential. The dash curve outlines the boundary between the positive and negative phases (Image by Institute of Physics).|
|Fig. 2 Bias-induced phases featuring ∆φbias(x, y) under different negative biases in the RESET process, where the intrinsic inner potential is removed by using the phase image of the pristine sample without any bias. The images are retrieved from the holograms taken at zero voltage after the corresponding bias setting. Dark red is for higher potential while dark blue is for the lower potential. The dash curve outlines the boundary between the positive and negative phases (Image by Institute of Physics).|
|Fig. 3 Low-loss energy-filtered images of the RRAM under different positive biases, demonstrating that the concentration of the oxygen varies with increasing positive bias. (Red denotes the higher oxygen concentration.) The dash curve outlines OXYGEN-RICH domain which shrink under positive bias (Image by Institute of Physics).|
|Fig. 4 Schematic diagrams of the RS processes in HfO2-based RRAM. From the top to the bottom sequentially are the forming process, RESET process and SET process. First column: dynamic process, second column: simulated CF, third column: experimental potential maps (Image by Institute of Physics and Peking Unversity).|
Institute of Physics
Resistive switching; oxygen vacancy; in situ TEM; electron holography; energy filtered image;
Resistive switching of oxide-based metal-insulator-metal has been widely studied due to the fundamental research interest and the great potential applications. The fundamental understanding of the microscopic physical properties of resistive switching is necessary but still insufficient. Here, the resistive switching processes are studied by electron holography and in situ energy filtered imaging in HfO2-based resistive random access memory, which directly demonstrate the microscopic evolution behaviors of the nano-filaments in the formation and rupture processes. Electron holography reveals that the electrons migrate from the bottom (cathode) to the top (anode) interface of the HfOx film in forming process and move back from the top (cathode) interface at reset operation, which indicates definitely that the switch behavior occurs in the top interface of the hafnia layer. Meanwhile, in situ plasmon energy filtered images demonstrate that the oxygen vacancies are gradually generated in the insulator layer to form the channels connecting two electrodes under the positive bias. The results depict the dynamic evolutions of the nano-filaments in binary metal oxides in real time with a high spatial resolution, and clarify the relationship between the electrical property and the microscopic oxygen vacancy distribution, and provide direct evidences for understanding the resistance switching mechanism.