Paper title: A Memristive Switching Uncertainty Model

Authors: Spyros Stathopoulos¹, Alex Serb¹, Ali Khiat¹, Maciej Ogorzałek² and
    Themis Prodromakis¹

Corr. author: Dr Spyros Stathopoulos (s.stathopoulos@soton.ac.uk)
ORCID: 0000-0002-0833-6209

Institution 1: University of Southampton
Department: Zepler Institute for Photonics and Nanoelectronics
Group: Electronic Devices and Materials Group
Address: Highfield Campus, Southampton, SO17 1BJ, UK

Institution 2: Jagiellonian University
Department: Department of Information Technologies
Address: ul. Łojasiewicza 11, 30-348, Kraków, Poland

Period of data collection: Oct 2018 to May 2019

Manuscript submitted: 21 January 2019
Revision #1 submitted: 19 March 2019
Revision #2 submitted: 14 May 2019
Manuscript accepted: 17 May 2019

Journal: IEEE Transactions on Electron Devices
Data DOI: 10.5258/SOTON/D0929

##############################################################################

Instructions on reading the provided data
=========================================

The data for each figure is contained within its respective folder. All figures
are accompanied by their corresponding gnuplot script used to generated.
Gnuplot is open source software [1]. The scripts can be used either by
double-clicking on them (provided gnuplot is installed) or loaded directly
through the command line (quotes are mandatory).

> load 'FILENAME.gpl'

[1] http://www.gnuplot.info

Figures 1 and 2
---------------
Conceptual information; no actual data used

Figure 3
--------
Data used for the illustration are generated using the provided Python 3.x
script. Upon running the script folder `fig3-data` will be generated and
populated with the data. The graphs can then be generated with the gnuplot
script `fig3.gpl`.

Figure 4
--------
Subfigure (a) is an excerpt from the generated data (3e). The gnuplot script for
subfigure (b) is `fig4b.gpl`. Error metric plot for different values of `N` are
available in the files ks_XX.dat for N = 01, 02, 05 and 10 as presented in the
paper. Gnuplot script `fig4c.gpl` will generate the plot. These plots are used
to find the `k` that minimises the error function.

Figure 5
--------
Switching data for the device as shown in fig 5a are provided in the
W11B11-{bias,read}-{neg,pos}.txt files. These contain the resistance of the
device under constant bias for voltages 1.5, 1.6, 1.7, 1.8 and 1.9~V (one
column for each) in W11B11-*-pos.txt files and -1.5, -1.6, -1.7, -1.9 and
-1.9~V (one column for each) in W11B11-*-neg.txt files. Figure 5a presents
an excerpt of this data (column 5 → ±1.9~V). This plot is produced with the
gnuplot script `fig5a.gpl`.

Data for the increment plots 5b (programming) and 5c (readout) are available
for the voltages of ±1.9~V (the same as in 5a). Files W11B11-V4-increment*.txt
contain six columns. As measurements are repeated three times each set of two
columns represents a repetition of said measurement. Fig~5{b,c} portray the
second repetition (columns 3 and 4). Odd columns represent resistance (R); even
columns represent differences in resistance (ΔR). Script `fig5bc.gpl` will
generate the increment plots for programming and readout. In the case of figure
5b quadratic fits are used to guide the eye only and are included in the
gnuplot script as well.

For the figure 5d separate planes have been fitted to programming and readout
results. The parameters for the planes fitted to the programming results are the
same as in the column representing DUT 2 in Table III. To extract the readout
plane all the readout measurements after programming (W11B11-*_FITDATA.txt) are
lumped together and are fitted with the linear fit the R-σ coordinates of which
are available in the file `linfit.txt`. The linear fit is extended along the
V-axis. The plot is produced by script `fig5d.gpl`.

Finally `fig5e.gpl` will plot the N(R,V) surface as explained in the article
(equation 4) using the parameters already used for plot 5d.