Supplementary Information

Transcription

1 Normalized Intensity Current (A) Supplementary Information x30 m 2 V set V reset Electroforming Pt/ / /Pt Voltage (V) Pt/ / /SiO 2 /Pt Supplementary Figure S1: The d.c. I-V characterization of Pt (Top electrode)/ / /Pt structure with/without a SiO 2 (1.5nm) layer at the Pt/ interface. A thin SiO 2 layer at the Pt/Ta 2 O 5 x interface reduces switching currents. Black line indicates I-V curve characteristics of the initial electroforming process. Resistive switching was performed at 30 x 30 m 2 cell size. Binding Energy (ev) Supplementary Figure S2: A comparison of XPS spectra taken at the and layers. The XPS spectra show that the TaO 2 x (red) layer has a higher concentration of Ta metal than the Ta 2 O 5 x (lue) layer. 1

2 Current (na) Current (A) Intensity ( a.u.) Fermi E Energy F Ef Binding energy (ev) 0-2 Supplementary Figure S3: A comparison of valence and XPS spectra taken at the and layers. The valence and XPS spectrum of the a-tao 2 x (red) layer shows a strong peak near the Fermi level, E F, indicating that there exist many metallic phases in the ase layer in contrast to the a-ta 2 O 5 x (lue) layer. a Set 1 Set 2 Set 3 1E-5 LRS E-6 1E-7 1E E Voltage (V) Numer of cycles Supplementary Figure S4: Multi-cycle test of the TaO x -ased switching device in a TEM. (a) Local I-V curves for the Set operations, which were otained during three consecutive switching cycles. () Corresponding multi-cycle data otained from the same switching device of a. Switching was carried out y -2.2V, 200 ms pulse for LRS and +3.2 V, 10 ms pulse for. 2

3 a LRS Supplementary Figure S5: In-situ HAADF-STEM images of the Pt/SiO 2 / / layers. Low magnification HAADF-STEM images oserved at the same location under (a) LRS and (). Magnified HAADF-STEM images shown in Fig. 2d,f otained from these lue oxed areas. Scale ar: 10 nm. a Pt SiO 2 LRS SiO 2 Pt Supplementary Figure S6: Magnified HAADF-STEM images of the Pt/SiO 2 / layers. Magnified HAADF-STEM images under (a) and () LRS, specially taken in the oxed area B (Fig. 2d). Tantalum oxide clusters, still remained in the SiO 2 layer after the Reset operation, were indicated y yellow arrows. Scale ar: 2 nm. 3

4 a LRS Supplementary Figure S7: In-situ HAADF-STEM images of the Pt/SiO 2 / / layers. In-situ HAADF-STEM images showing the (a) formation and () annihilation of conducting filaments taken at the same location under two different resistance states (LRS and ). Nanoscale filaments in the LRS were indicated y yellow arrows. Scale ar: 5 nm. a LRS Supplementary Figure S8: In-situ HAADF-STEM images of the Pt/SiO 2 / / layers. In-situ HAADF-STEM images showing the (a) formation and () annihilation of conducting filaments taken at the same location under two different resistance states (LRS and ). Nanoscale filaments in the LRS were indicated y yellow arrows. Scale ar: 5 nm. 4

5 Intensity of HAADF STEM image Filamentary TaO x Insulating TaO x SiO Distance (nm) (SiO 2 TaOx cluster) Supplementary Figure S9: Intensity profiles of the HAADF-STEM images taken from the same tantalum oxide cluster of Fig. 3a after the Set (red) and Reset (lue) operations. In a qualitative analysis of the HAADF-STEM images, we found that the intensity ratio of the filamentary tantalum oxide cluster to the SiO 2 area (I filament /I SiO2 = 27500/12000 = ~2.3) was higher than that of the insulating tantalum oxide cluster to the SiO 2 area (I insulating cluster /I SiO2 = 19000/12000 = ~ 1.6). 5

6 a LRS 1 SiO 2 c LRS 2 Supplementary Figure S10: In-situ HAADF-STEM images oserved during multiple resistive switching cycles. In-situ HAADF-STEM images showing the repetitive formation and annihilation of conducting filaments taken at the same location during multiple resistive switching cycles (LRS1 (a) -> () -> LRS2 (c)). Nanoscale filaments in the LRS were indicated y yellow arrows. Scale ar: 5 nm. 6

7 a LRS 1 SiO 2 c LRS 2 Supplementary Figure S11: In-situ HAADF-STEM images oserved during multiple resistive switching cycles. In-situ HAADF-STEM images showing the repetitive formation and annihilation of conducting filaments taken at the same location during multiple resistive switching cycles (LRS1 (a) -> () -> LRS2 (c)). Nanoscale filaments in the LRS were indicated y yellow arrows. Scale ar: 5 nm. 7

8 Intensity (a.u.) Intensity (a.u.) a F G O-K Energy loss (ev) Energy loss (ev) Supplementary Figure S12: Compositional analysis of the right region indicative of conducting paths in the TaO 2 x ase layer. (a) Low-energy electronloss spectroscopy and () O-K edge EELS spectra taken at the right region in Fig. 4a (Set state). The ulk plasmon peak taken at the right region is similar to that of the metallic TaO 1-x phase. Supplementary Figure S13: Cross-sectional HAADF-STEM images of the a- layer taken in the initial state. (a) Original HAADF-STEM image. () Pseudocolor map otained y converting the raw STEM image of a. Scale ar: 3 nm. 8

9 O/Ta (at %) a c Supplementary Figure S14: Aerration-corrected HAADF-STEM images. Atomic scale HAADF-STEM images showing the structural changes of the a-tao 2 x layer after the (a) Set, () Reset, and (c) reakdown operations. The conductive clusters in the a-tao 2 x layer formed after the Set operation showed atomic short-range order structure and evolved into a long-range ordering y the excess voltage sweep (V> V Set ) over a long period of time (Break down operation). Scale ar: 1 nm. a nm, 20 spectra 0.5 middle ottom Distance (nm) Supplementary Figure S15: Oxygen concentration in the lower part of the a- layer. (a) Cross-sectional HAADF-STEM image of the Ta 2 O 5 x /TaO 2 x ilayer with an arrow indicating where EDS line profiles were otained. Scale ar: 10 nm. () Corresponding EDS line profiles for the O/Ta atomic ratio otained from the lower part of the layer after Set (red) and Reset (lue) operations. 9

10 Normalized Intensity 0.10 t = ln (I t /I 0 ) : inelastic mean free path I 0 : sum of zero-loss peak counts I t : sum of total spectrum counts Energy Loss (ev) Supplementary Figure S16: Sample thickness of the regions P1 and P2. Lowloss EELS data measured at the regions P1 (lue) and P2 (red) in Fig. 3e, showing that the sample thickness was varied within 1% (t p1 = 0.91, t p2 = 0.90 ). Thickness was measured from the log-ratio formula t = λ In (I t /I 0 ). Pt (TE) Supplementary Figure S17: Cross-sectional HAADF-STEM image of the Pt/Ta 2 O 5 x interface after the Set operation taken at the region far from the Pt-Ir proe. This image shows a decrease in the numer density of nanoscale filaments compared to Fig. 6c. Scale ar: 10 nm. 10