SUPPLEMENTARY INFORMATIONS

Transcription

1 SUPPLEMENTARY INFORMATIONS Dynamic Evolution of Conducting Nanofilament in Resistive Switching Memories Jui-Yuan Chen, Cheng-Lun Hsin,,, Chun-Wei Huang, Chung-Hua Chiu, Yu-Ting Huang, Su-Jien Lin, Wen-Wei Wu *,, and Lih-Juann Chen National Chiao Tung University, Department of Materials Science and Engineering, Taiwan. National Tsing Hua University, Department of Materials Science and Engineering, Taiwan. National Central University, Department of Electrical Engineering, Taiwan. * To whom correspondence should be addressed, WWWu@mail.nctu.edu.tw Table of contents. 1. MATERIALS AND METHODS. 2. COMPOSISTION ANALYSIS AND ELECTRICAL PROPERTIES OF NANOFILAMENTS. 3. THE I-V SWITCHING CHARACTERISTIC. 1

2 1. MATERIALS AND METHODS. To prepare the specimens for in-situ TEM observation, SiO 2 (30 nm) / Si 3 N 4 (60 nm) bilayer thin film was deposited on a double-side-polished, (100)-oriented silicon wafer by a low pressure chemical vapor deposition chamber, as shown in Figure S1a. The oxide/nitride bilayer film was patterned at one side, serving as an etching mask for subsequent Si substrate removal, as shown in Figure S1b. Then the wafer was put into a 20% KOH solution at 85 C for 1.5 h to etch away the Si substrate through the predefined windows, leading to the formation of a suspended Si 3 N 4 membrane of an area of 200 µm 200 µm, as revealed in Figure S1c. The photolithography process was used to define the outer electrode on the front side of the membrane, and Au (40 nm) /Ti / (20 nm) electrode were deposited by electron beam evaporation, as shown in Figure S1d. The Pt/ ZnO /Pt layer was deposited by a radio-frequency sputter. The ZnO was polycrystalline with a thickness of 100 nm. This Pt/ ZnO /Pt layer was cut into 100 nm 50 nm 1 µm thin film by focus ion beam (FIB). The distance between two Pt electrodes was 100 nm; 50 nm was the thickness of the FIB specimen for investigation in the TEM; the width of the specimen was designed to easily observe the filament path through cutting the FIB specimen into 1 µm wide to reduce the observed area, as shown in Figure S1e. Finally, the Pt/ ZnO /Pt thin film was picked up and placed on the front side of the membrane. 200 nm wide Pt inner connect 2

3 electrode (connecting the outer Au/Ti electrodes and Pt of the specimen) was deposited by FIB. The advantage of this design is that the current flow in this sample is similar in the general MIM or MSM case. Figure S1f shows the scanning electron micrograph of the specimen on the observation window. The electron sequentially pass through outer electrode(au), top electrode(pt), ZnO, bottom electrode(pt), low resistance Si substrate, and outer electrode(au). Therefore, the actual resistance of our in-situ device should be less than measured value shown in Figure 1. The specimen was mounted on an in-situ TEM holder (Protochips Aduro300) which allows us to introduce an electric current through the specimen during TEM observation. The sample was loaded into the TEM (JEOL 2000V UHV-TEM and JEOL JEM-2100F) and pumped down to a base pressure of torr. The in-situ TEM observation was performed at room temperature to observe the electroforming behavior with an electric current limit of 10-4 A The resistance values are about Ω for LRS and Ω for HRS before fabricating the lateral device by FIB. The evolution of filament formation images was video recorded in real time. 3

4 Figure S1. The process sequences of preparing the specimens for in-situ TEM observation. (a) Deposition of a SiO2/Si3N4 bilayer thin film on a double-side-polished (100)-oriented silicon wafer. (b) Definition of the etching window on the SiO2/Si3N4 bilayer thin film. (c) Fabrication of the SiO2/Si3N4 membrane structure by etching the silicon substrate. (d) Deposition, patterning of a Ti/Au electrodes on the Si3N4 membrane. (e) The 3-D schematic illustrations and SEM image of the Pt/ ZnO/ Pt specimen. (f) The SEM image of the specimen in the observation window. 4

5 2. COMPOSISTION ANALYSIS AND ELECTRICAL PROPERTIES OF NANOFILAMENTS.. Figure S2. The TEM EDX mapping and line scan from cathode to anode after the external electric filed. No obvious migration of the top and bottom Pt electrode has been observed. Comparing with the element distribution of zinc and oxygen, the oxygen tends to gather toward anode as applying bias, while zinc does not. Zinc may be more mobile than oxygen in ZnO, but zinc would be dynamic equilibrium. The distribution of zinc is uniform in the EDX. Besides, the EDX shows clear contrast of the O distribution to support our explanation that the oxygen dominates the reaction under electric field while ZnO matrix acts as oxygen reservoir near the anode. 5

6 Figure S3. The ex-situ TEM image and the corresponding I-V measurement. (a) The high-resistance-state (HRS) after 2 times of the re-writing process. (b) The HRS after 3 times of the re-writing process. (c) The corresponding I-V electrical property. The I-V curve included the electroforming process and first RESET of the in-situ measurement (Fig. 2d). The I-V measurement indicates that this specimen has ReRAM properties. From HRS of the ex-situ TEM image (Figs S2a and b), the conductive filament region is approaching the Pt anode electrode. Also, the boundary of the matrix and conductive filament gradually blurred, implying the lack of the oxygen-ion supplement. This device may stay in LRS and may not work after several re-writing process. 6

7 Figure S4. The diffraction pattern of ZnO film. The left: diffraction pattern of the as-prepared specimen (without electrical operation); the right: diffraction pattern of the region containing the conductive filament and the matrix. There is a concern that the electrical operation would be able to crystallize the amorphous oxide material. The left side diffraction pattern shows the as-prepared film in order to illustrate its initial structure, indicating the ZnO film deposited by RF sputter was polycrystalline before the electrical operation. We have verified the possible element (Pt, Ti, Pt x Zn y, Pt x Ti y, Ti x Zn y etc.) with the extra spot and atomic structure and Zn is the only possible alternative. Besides, there is no diffraction spot ascribed to ZnO 1-x. 7

8 Figure S5. Temperature dependence of the LRS of the Pt/ZnO/Pt device. Resistance increases linearly with temperature from 295 to 355 K, indicating that the conducting nanofilament in the device is a metallic phase. 8

9 Figure S6. The TEM EDS analysis of chemical compositions. The labels of the EDS spectra and table are corresponding to the red dot marked in the TEM image at lower-right corner. Despite the semi-quantitative nature of EDS, we can still find the tendencies in the TEM EDS data: Firstly, the concentration of oxygen decreased after the writing process, which confirms that the oxygen ions escaped from the ZnO. Secondly, the concentration in the erasing region is almost the same as the ZnO matrix. It is an indirect evidence that the erasing region is transformed back into ZnO 9

10 composition. Finally, there are still oxygen signal detected in the conductive filament, which is attributed to embedment of the Zn-dominated ZnO 1-x filament in the ZnO matrix and the imprecision in EDS measurement. 10

11 3. Ga damage via FIB cutting technique Ga ion damage is an issue in FIB preparation. The FIB sputter would cause the material amorphization. As a result, we take caution in using Ga ion beam of small voltage, small current and small angle to handle the ZnO thin film after the FIB cutting to remove the damaged ZnO. Furthermore, the electron diffraction pattern of as-prepared ZnO indicates that the ZnO film remains polycrystalline (shown in Fig. S3) and the Si substrate is still single crystalline (not shown). The TEM EDX data (shown in Fig. S4) did not reveal the presence of the Ga. It is therefore inferred that the Ga does not have significant influence on our experiment. 4. THE I-V SWITCHING CHARACTERISTIC. The concept of a series of insulator and metal filaments between the parallel conductive plates was used to describe the switching model. In our in-situ device, the resistance values are Ω for LRS and Ω for HRS. The following equation describes the I-V switching characteristics of the series connection of conductive filament and rectifier 4,16,18 : Va I = R( V a = ) R Va V + ( V ) R ZnO a on Zn = A eff AT 2 eφ B eva exp exp kt nkt m 1 + ω V R a Zn,where A eff, A, n, R Zn, ω, m, and ψ B are the effective area of the rectifier region, the Richardson constant ( A=4πemk 2 /h 3 (A/cm 2 )), the ideality factor (n 1), the resistance 11

12 from the metallic filaments composed by zinc, the value to describe the ReRAM in the ON state or the OFF state (normalized between 0 (OFF) and 1(ON)), a free parameter in the model to consist with the experimental observations, and the barrier height between metal and semiconductor, respectively. The first term presents the I-V approximation for the rectifier and the current is attributed to thermionic emission, where the exponential term was ascribed to the Schottky barrier junction. The second term in the equation is the I-V approximation for the ON state of the ReRAM device, which is mainly the electron current through the connective metal filaments. This approximate equation is applicable to the insulator with embedded metallic filaments. 12