Atomristor: Non-Volatile Resistance Switching in Atomic Sheets of

Transcription

1 Atomristor: Non-Volatile Resistance Switching in Atomic Sheets of Transition Metal Dichalcogenides Ruijing Ge 1, Xiaohan Wu 1, Myungsoo Kim 1, Jianping Shi 2, Sushant Sonde 3,4, Li Tao 5,1, Yanfeng Zhang 2, Jack C. Lee 1 and Deji Akinwande* 1 1 Microelectronics Research Center, The University of Texas at Austin, Austin, Texas 78758, United States. 2 Department of Materials Science and Engineering, College of Engineering, Peking University, Beijing , China. 3 Institute for Molecular Engineering, University of Chicago, 5640 South Ellis Avenue, Chicago, Illinois 60637, United States. 4 Center for Nanoscale Materials, Argonne National Laboratory, 9700 Cass Avenue, Lemont, Illinois 60439, United States. 5 School of Materials Science and Engineering, Southeast University, 2 Southeast University Road, Nanjing, , China. Equal Contribution Supplementary Fig. 1: MoS 2 materials characterizations Supplementary Fig. 2: Schematic of transfer processes Supplementary Fig. 3: Retention time of monolayer TMDs non-volatile switches Supplementary Fig. 4: Low-voltage Read operations Supplementary Fig. 5: Test crossbar device without TMD active layer Supplementary Fig. 6: Dependence of SET/RESET voltage on area Supplementary Fig. 7: Characterizations of few-layer MoS 2 Supplementary Fig. 8: Dynamic mechanical analysis (DMA) and switching I-V curves before and after bending Supplementary Fig. 9: Current sweep SET operation of monolayer MoS 2 crossbar device Supplementary Figure 10. RF switch simulation results Supplementary Table. Comparison with previous works. 1

2 a b c d e f Supplementary Figure 1. MoS 2 materials characterizations. (a) SEM image of fully covered MoS 2 film on Au foils. The white point indicated by the arrow is the bare Au substrate. (b) Low-magnification TEM image of the transferred MoS 2 film on TEM grid, suggesting the continuity of MoS 2 monolayer film. (c) High resolution TEM image on a folded edge showing the monolayer feature of the MoS 2 film. 2

3 (d) AFM image captured from the edge of MoS 2 film and the corresponding section-view along the white arrow showing the monolayer feature of the MoS 2 (~0.87 nm in apparent height). (e) Raman spectrum of as-grown MoS 2 on Au foil. (f) PL spectrum of transferred MoS 2 on SiO 2 since PL is not directly visible of as-grown monolayer on Au foil. The PL peak position located at 1.86 ev in good agreement with the theoretical value for monolayer MoS 2, implies high crystal quality of the transferred sample. 3

4 a MOCVD-grown MoS 2 10 μm b Supplementary Figure 2. Schematic of transfer processes. (a) SEM image of as-grown MOCVD monolayer MoS 2. (b) Simplified illustration of the steps for PDMS assisted pick-and-place transfer fabrication. 4

5 a b c Supplementary Figure 3. Retention time of monolayer non-volatile resistance switches. (a-c) The retention time of MoS 2, MoSe 2, and WS 2 memory devices at room temperature, respectively. The resistance of HRS and LRS is determined by the current read at 0.1 V. The area of the crossbar devices is μm 2 for MoS 2 device, μm 2 for MoSe 2 device, and 2 2 μm 2 for WS 2 devices. 5

6 a b Supplementary Figure 4. Low-voltage Read operations. (a,b) Read operation of monolayer MoS 2 crossbar device at HRS and LRS from -0.1 V to +0.1 V. The above results confirm linearity at low voltage, which is important for reading the memory state. 6

7 Supplementary Figure 5. Test crossbar device without 2D active layer. The structure is Au (60nm) /Cr (2nm) /Au (60nm). Cr serves as the adhesion layer in the lift-off process. No resistive switching behaviour is observed in this structure, indicating that the TMD active-layer plays the primary role. 7

8 Supplementary Figure 6. Dependence of SET/RESET voltage on area. SET & RESET voltages vs. area for MoS 2 crossbar devices. As the device area increases, SET voltage decreases from 3 V to 0.5 V, while RESET voltage remains relatively flat around -1 V. 8

9 a b c d Supplementary Figure 7. Characterizations of few-layer MoS 2. (a) Raman spectra for different layers show the two most prominent peaks namely the E 1 2g and A 1g modes. The Raman shift of E 2g mode decreases whereas that of A 1g mode increases with increasing layer number. (b) Photoluminescence spectra showing decreasing intensity with increasing layer number. Intensity of 2-4 layer MoS 2 spectra is magnified by 10x for visibility. (c) TEM cross-section image of Cr/4 layer MoS 2/Cr litho-free device revealing the clear MoS 2 layer structure. The distance between the two arrows is 0.65 nm. (d) Elemental mapping of Cr/4-layer MoS 2/Cr litho-free device. The middle pink layer is the result of color combination of red Mo layer and blue S layer. 9

10 a b c Supplementary Figure 8. DMA apparatus and switching I-V curves before and after bending. (a) Photography of the bent bilayer MoS 2 crossbar devices on PI substrate via DMA apparatus. (b) Schematic of the bent MoS 2 device sandwiched by gold electrodes on flexible substrate. (c) Typical switching I-V curves of bilayer MoS 2 crossbar devices before and after 1000 cycles at 1% strain. This device is the same one with the LRS device in Fig. 5e. 10

11 Supplementary Figure 9. Current sweep SET operation of monolayer MoS 2 crossbar device. The current increases from 0 to 30 ma and then decreases to 0. The voltage increases with current until the voltage reaches 0.65 V, and then suddenly drops to around 0.2 V, which indicates the first resistance decreasing process (SET) at 0.65 V. From 0 to 30 ma, it has five separate SET steps. As the current decreases from 30 to 0 ma, the voltage follows a linear decreasing step back to 0 V. The inset shows the voltage sweep in the same device, with the SET voltage to be 0.63 V, corresponding to the first transition in current sweep at 0.65 V. The area of this crossbar device is μm 2. 11

12 Supplementary Figure 10. RF simulation results. 2DNS RF simulation in a 50 Ω system illustrating the outstanding prospects for switching GHz and THz signals. The 2D switch is a scaled version of the experimental results shown in the main text (Fig. 5d), with a 100x smaller area ( μm 2 MIM 1L- TMD). The simulated device parameters (R on ~6 Ω and C off ~0.26 ff) are predicated on experimentally achievable values. The insertion loss is less than 1 db in the ON state and the isolation greater than 20 db in the OFF state up to 500 GHz. Owing to the frequency-area scaling relation (f co~1/area), the figure of merit for this 0.01 μm 2 2D MIM switch is about 100 THz. 12

13 Supplementary Table. Comparison with previous works. The comparison of this work with other representative 2D non-volatile publications. Reference Active layer material Active layer thickness Device structure Forming voltage New application (RF Switch)* This work CVD-grown TMDs Monolayer Vertical Forming-free Yes Sangwan et al., NATURE NANOTECHNOLOGY, 10, 2015 CVD-grown MoS 2 Monolayer Planar ~ ±20V No Xu et al., APPLIED PHYSICS LETTERS, 104, 2014 Cheng et al., NANO LETTERS, 16, 2016 Bessonov et al., NATURE MATERIALS, 14, 2015 Hao et al., ADVANCED FUNCTIONAL MATERIALS, 26, 2016 Jeong et al., NANO LETTERS, 10, 2010 Puglisi et al., IEDM, , 2016 Hydrothermal MoS 2 nanosphere assembly ~ 200 nm Vertical N/P No Solution-processed 1T phase MoS nm Vertical N/P No Solution-processed MoO x/mos 2 and WO x/ws 2 Liquid-exfoliated black phosphorous Spin-casted graphene oxide MoO x/woo x <3 nm Top degraded BP, ~ 10 nm Vertical Forming-free No Vertical N/P No 15 nm Vertical Forming-free No CVD-grown hbn 5-7 layers Vertical 2-4 V No *New application refers to new usage of non-volatile resistance switching materials beyond contemporary applications in memory and neuromorphic circuits. N/P: data not provided 13