READ ME File For 'Dataset for 3D-Structured Mesoporous Silica Memristor for Neuromorphic Switching and Reservoir Computing'

Dataset DOI: 10.5258/SOTON/D2447

ReadMe Author: Ruomeng Huang, University of Southampton 

This dataset supports the publication: 3D-Structured Mesoporous Silica Memristor for Neuromorphic Switching and Reservoir Computing
AUTHORS: Ayoub H. Jaafar, Li Shao, Peng Dai, Tongjun Zhang, Yisong Han, Richard Beanland, Neil T. Kemp, Philip N. Bartlett, Andrew L. Hector and Ruomeng Huang
TITLE: Ayoub H. Jaafar,a,b Li Shao,c Peng Dai,a Tongjun Zhang,a Yisong Han,d Richard Beanland,d Neil T. Kemp,b Philip N. Bartlett,c Andrew L. Hector,c* and Ruomeng Huang a*
JOURNAL: Nanoscale
PAPER DOI IF KNOWN: DOI	https://doi.org/10.1039/D2NR05012A

This dataset contains:
The raw data of figure 1 to 8.

The figures are as follows:

Figure 1. Characterization of the 3D-structured mesoporous silica thin film and memristor. a) Fabrication process of the sol-gel mSiO2 thin films. b) The GISAXS patterns of the mSiO2 film with simulated Bragg peaks. The red circles and white squares represent the transmitted and reflected Bragg peaks, respectively. c) Cross-sectional TEM image of the as-deposited mesoporous silica thin film. d) Schematic of the mSiO2 based memristor. e) A cross section TEM image for mSiO2 memristor. f) Higher magnification TEM image for the mSiO2 thin film in the memristor. g-j) Cross-sectional TEM-EDX of the mSiO2 based memristor. The scale bar is 50 nm.

Figure 2. Resistive switching behavior of 3D-structured mSiO2 based memristor. a) I-V curve showing the forming process of the mSiO2 memristor. b) Consecutive volatile switching I-V characteristics of the mSiO2 memristor under a CC of 100 µA. c) Endurance characteristics, and d) Cumulative probability of HRS and LRS for the volatile switching of the mSiO2 memristor. e) Consecutive I-V characteristics showing the typical non-volatile resistive switching properties of the mSiO2 memristor under a CC of 5 mA. f) Endurance characteristics, g) Cumulative probability of HRS and LRS, and h) Cumulative probability of VSET and VRESET for the non-volatile switching of the mSiO2 memristor.

Figure 3. a) A cross section TEM image for the switched 3D-structuredd mSiO2 based memristor showing the conical nanoscale grown Ag filaments within the mSiO2 insulator thin film. b-e) Cross-sectional TEM-EDX of the mSiO2 memristor. The scale bar is 50 nm. f) The proposed switching mechanism for the mSiO2 based-memristor consisting of filament formation under the application of a SET process at Low CC (iii) followed by spontaneous breakage of the filament (iv). In contrast, a high CC results in the formation of a more robust filament (v) that can only be broken by a full RESET process (vi).

Figure 4. Multi-state switching behavior of the 3D-structuredd mSiO2 based memristor. a) I-V characteristics showing the multi-state switching behavior induced by varying the CC. b) Endurance of the multi-state resistance states. c) Consecutive volatile switching I-V characteristics with CC of 100 µA and showing the analog switching behavior of the device. d) Conductance versus pulses curve of the LRS of the device. The conductance was read at 0.2 V.

Figure 5. Analogy between biological-synapses and mSiO2 memristor and short-term dynamics under pulse stimuli. a) Schematic representation of a biological neural network and a memristor device showing the correspondence between biological and the electronic synapses. b) Gradual PSC change with a series of voltage pulses (+1 V, 5 ms duration) and the subsequent auto-decay showing the STP behavior. c) Current response of the device for pulses with different inter-spike intervals, emulating SRDP behavior. d) Current response of the mSiO2 memristor for pulses with different voltage amplitudes, emulating SVDP behavior. e) Current response of the mSiO2 memristor for pulses with different durations, emulating SDDP behavior.

Figure 6. Reservoir computing system based on mSiO2 memristor. a) Schematic of the RC system showing the reservoir x(t) with the input u(t) layer and output y(t) layer. Win is the input weight, W is the weight of the current connection in the reservoir, and Wout is the readout weight. b) Letter “A” as an example with size of 5 x 5 pixels for letter recognition. c) Schematic representation of the physical RC system including the inputs (pulse streams), the mSiO2 memristor reservoir and the readout network. d) The recorded reservoir states for input letter “A”. e) The response of memristor reservoir to six different pulse streams. 

Figure 7. Training and performance evaluation of the mSiO2 memristor-based reservoir computing system. a) The training and validation loss as a function of training epochs. b) The training and validation accuracy as a function of training epochs. c) Confusion matrix showing the prediction results from the RC system against the ground truth in the test dataset. Examples of letter patterns containing d) one, e) two, and f) three random noises. Confusion matrix showing the prediction results from the RC system against the ground truth for letter patterns in the noisy test dataset containing d) one, e) two, and f) three random noises.

Date of data collection: October 2020 to April 2022

Information about geographic location of data collection: United Kingdom

Licence: CC-BY
Related projects: This work is part of the ADEPT project funded by a Programme Grant from the EPSRC (EP/N035437/1). 


Date that the file was created: Nov 2022