Fig. 2
From: A hardware Markov chain algorithm realized in a single device for machine learning
Resistive switching behavior based on SnSe RRAM. a AFM results showing the surface of SnOx/SnSe/SnOx with atomic flatness. Both top electrode (TE) and bottom electrode (BE) are shown. The scale bar is 1 μm. b AFM results showing generation of bubbles in filament region after 100 cycles operation. The scale bar is 1 μm. The inset shows the zoom-in image of AFM and the related thickness profile. The scale bar of inset AFM image is 200 nm. c The resistive switching behavior plotted in log-scale showing that the space-charge-limited mechanism dominates the electrical transport. The voltage region is fit with Ohmic law. d The zoom-in image of the dashed line in Fig. 3c showing the Child’s law relation at high voltage region. e The HRS, MRS, and LRS distribution. f The first, second set, and reset voltage distribution. g The first, second set voltage vs. SnSe total thickness. The error bars in y-direction are calculated based on the set voltage in 100 cycles for each device. The error bars in x-direction are calculated based on the thickness variations from AFM results for each SnSe flake. h The HRS, MRS and LRS in three devices with ~100 nm thickness. The error bars for HRS, MRS, and LRS are calculated from 100 cycles for each device. The two-step set/reset switching behavior of SnOx/SnSe/SnOx resistive memory under (i) 300 K and (j) 77 K, respectively
