micromachines
Article
Amorphous ITZO-Based Selector Device for Memristor
Crossbar Array
Ki Han Kim 1 , Min-Jae Seo 2, * and Byung Chul Jang 1, *
Citation: Kim, K.H.; Seo, M.-J.; Jang,
B.C. Amorphous ITZO-Based Selector
Device for Memristor Crossbar Array.
Micromachines 2023, 14, 506.
https://doi.org/10.3390/
mi14030506
Academic Editor: Zhongrui Wang
Received: 22 January 2023
Revised: 12 February 2023
Accepted: 20 February 2023
Published: 22 February 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Micromachines 2023, 14, 506. https://doi.org/10.3390/mi14030506
1 School of Electronic and Electrical Engineering, Kyungpook National University, 80 Daehakro, Bukgu,
Daegu 41566, Republic of Korea
2 Department of Electronic Engineering, Gachon University, 1342 Seongnam-daero,
Seongnam 13120, Republic of Korea
* Correspondence: mjseo@gachon.ac.kr (M.-J.S.); bc.jang@knu.ac.kr (B.C.J.)
Abstract: In the era of digital transformation, a memristor and memristive circuit can provide
an advanced computer architecture that efficiently processes a vast quantity of data. With the
unique characteristic of memristor, a memristive crossbar array has been utilized for realization of
nonvolatile memory, logic-in-memory circuit, and neuromorphic system. However, the crossbar array
architecture suffers from leakage of current, known as the sneak current, which causes a cross-talk
interference problem between adjacent memristor devices, leading to an unavoidable operational
error and high power consumption. Here, we present an amorphous In-Sn-Zn-O (a-ITZO) oxide
semiconductor-based selector device to address the sneak current issue. The a-ITZO-selector device
is realized with the back-to-back Schottky diode with nonlinear current-voltage (I-V) characteristics.
Its nonlinearity is dependent on the oxygen plasma treatment process which can suppress the surface
electron accumulation layer arising on the a-ITZO surface. The a-ITZO-selector device shows reliable
characteristics against electrical stress and high temperature. In addition, the selector device allows
for a stable read margin over 1 Mbit of memristor crossbar array. The findings may offer a feasible
solution for the development of a high-density memristor crossbar array.
Keywords: memristor crossbar array; amorphous In-Sn-Zn-O (a-ITZO); selector device
1. Introduction
Digital transformation (DT) has accelerated the exponential expansion of total digital
data volume. In particular, the global pandemic has influenced individuals to not only
work remotely from home, study, and engage in e-commerce, but also to actively embrace
Internet of Things (IoT) technology. This trend has caused a paradigm shift from processor-
centric computing to memory-centric computing. To keep up with the new paradigm, a vast
volume of data must be efficiently processed on the edge device. Memristor and memristive
circuit are one of the candidates for the advanced computing architecture which differs
from the conventional von Neumann architecture suffering from data movement between
the central processing unit and memory. The memristor has been widely developed for
promising nonvolatile memory and artificial synapse device due to its simple two-terminal
structure, fast switching speed, and low power consumption [1–5]. In addition, memristive
circuit can implement a nonvolatile logic-in-memory circuit that enables normally off
computing [6–9].
The memristor crossbar array is the optimal architecture for enabling high packing
density, defect tolerance, vector-matrix product operation, and logic-in-memory opera-
tion [10–12]. During memristor operation, however, this architecture generates undesired
leakage currents, known as sneak currents, which flow through unselected neighboring
devices during memristor operation. These sneak currents lead to the operational error,
high power consumption, poor performance of memristive neuromorphic system, and
the limited logic operations in only a row of device [9,13], hence restricting the maximum
https://www.mdpi.com/journal/micromachines
Micromachines 2023, 14, 506 2 of 11

memristor array size. Vertical integration of a memristor device and a two-terminal selector
device with nonlinear current–voltage (I-V) characteristics is required to tackle these issues.
The selector device requires a symmetric I-V nonlinearity for bipolar switching memris-
tor, which suppresses the current flow at a low-voltage region while permitting a much
greater current (about >100) in a high-voltage region [14]. Numerous researchers have
scrutinized various two-terminal selector devices, which include the threshold switching
device [14–17], 0T2M structure [18], mixed-ionic conductor device [19], multilayer tun-
neling device [14], and the back-to-back Schottky diode [20–22]. The threshold switching
based on the metal-to-insulator (MIT) transition of VO2 [15] and the fast relaxation of
Ag conductive filament [17] suffer from the poor thermal stability because the VO2 is a
well-known material occurring at the MIT transition at a low temperature of 340 K [23] and
the accelerated relaxation of the Ag conductive filament by high temperature. Furthermore,
although the 0T2M structure is effective to suppress the sneak current, the larger area of
0T2M is not suitable for high-density memristor crossbar array. Therefore, the back-to-back
Schottky diode is preferred for a stackable high-density memristor array due to its ease
of fabrication, low thermal budget, and simple structure design. Although there has been
research on back-to-back Schottky diodes, Ni/TiO2/Ni devices have demonstrated supe-
rior performance [24,25]. However, the TiO2 -based selector device requires well-controlled
stoichiometry, crystallinity of the TiO2 , and insufficient current density by the poor carrier
mobility of this intrinsic material.
Amorphous-oxide-semiconductor (AOS) thin films have received significant attention
as active layers in driving thin-film transistors (TFTs) in the display industry due to their
inherent amorphous properties enabling high device-to-device uniformity, low manufac-
turing cost, and high transparency by virtue of its wide bandgap (3 eV) [26,27]. Although
the amorphous In-Ga-Zn-O (a-IGZO) material is currently utilized in the display industry,
an amorphous In-Sn-Zn-O (a-ITZO) material is suitable for selector device applications
due to its higher electron mobility (20–30 cm2 /Vs) compared to a-IGZO [28]. Nevertheless,
a stable Schottky barrier formation with this material is challenging, because the surface
electron accumulation layer (SEAL) naturally arises from the oxide semiconductor (OS)
surface reduced by formation of oxygen vacancy in a vacuum chamber [29–32]. Deposition
of a conductive polymer [33], a chemical surface treatment [34,35], and oxygen plasma
treatment [31,32,36,37] have all been proposed to impede and remove the formation of
SEAL. Among these strategies, oxygen plasma treatment is suitable to develop the OS-
based selector device because it can provide a uniform and reliable Schottky barrier with a
high oxidation rate at low temperatures. Therefore, it is necessary to investigate the oxygen
plasma treated a-ITZO-based selector in order to develop a high-performance engineering
methodology.
In this study, we investigate the a-ITZO-based back-to-back Schottky diode type
selector device. Depending on the oxygen plasma treatment process conditions, the a-ITZO-
selector device demonstrated a clear elimination of SEAL with the improved nonlinear
I-V characteristics. With the optimized oxygen plasma process, the selector device with
Pd/a-ITZO/Pd structure showed a symmetric nonlinearity of I-V, reliable endurance, bias
stress, and temperature stabilities. In addition, the simulation results for calculating reading
margin with bipolar memristor revealed that the integrated memristor array with the a-
ITZO-selector is capable of achieving a maximum array size of more than 1 Mbit. Therefore,
these results provide not only a-ITZO-selector device is suitable for memristive crossbar
array, but also the engineering strategy to improve the performance of the a-ITZO-selector,
as well as other OS-based selector devices.

2. Materials and Methods

The a-ITZO-selector devices with the structure of Pd/a-ITZO/Pd were fabricated with
the crossbar architecture. In this device, the Pd with high work function (5.22 eV) [38]
is adopted for both top and bottom electrode to form a Schottky barrier with n-type
semiconductor a-ITZO which has 4.5 eV of electron affinity. First, 60 nm-thick Pd electrodes
Micromachines 2023, 14, 506 3 of 11

were deposited SiO2 /Si substrate using thermal evaporation method through a shadow
mask. Then, a 30 nm thick a-ITZO film was deposited on the bottom Pd electrodes using
radio-frequency (RF) sputtering under high vacuum conductions (0.088 Pa). The atomic
ratio of the deposited a-ITZO film is investigated using X-ray photoelectron spectroscopy
depth profile (Figure S1). The atomic ratio of a-ITZO film without oxygen plasma treatment
has In:Sn:Zn:O = 22:4:18:56. After that, the oxygen plasma using an inductively coupled
plasma asher system was treated to remove the SEAL in a-ITZO film. Finally, 60 nm thick Pd
top electrodes were deposited perpendicular to the Pd bottom electrodes using the thermal
evaporation method. Both line widths of Pd were identical at 60 µm. The fabricated a-ITZO-
selector device was electrically characterized by providing a voltage to the Pd TE while
the Pd BE was grounded using a Keithley 4200 semiconductor parameter analyzer in an
air environment. To evaluate the performance of a-ITZO-selector for suppression of sneak
current in the memristor crossbar array, we performed the numerical simulation through
MATLAB using the experimental values for the LRS/HRS resistances and SET/RESET
voltages of TiO2 -based bipolar switching memristor and the resistance of a-ITZO-selector.
The TiO2 -memristor with the structure of AI/a-TiO2 /AI was fabricated using amorphous
titanium oxide films which are deposited by plasma enhanced atomic layer (PEALD)
deposition method (ASM Genitech MP-1000) [39]. Titanium tetraisopropoxide (TTIP) and
oxygen plasma were used as precursors for Ti and oxygen, respectively. For a-TiO2 film
deposition utilizing PEALD, 100 cycles of ALD at 80 ◦ C were performed. The cycle was
comprised of TTIP (2 s), Ar (4 s), O2 gas (2 s), and Ar (2 s). The radio frequency (RF) pulse
was applied for 1.5 s with 150 W of RF power during the injection of O2 gas. The Al bottom
electrode was deposited by thermal evaporation through a metal shadow mask and then
an a-TiO2 film was formed on the Al bottom electrode. The top Al electrode was deposited
using thermal evaporation method, which forms a crossbar array structure.

3. Results and Discussion

The sneak current path in the memristor crossbar array is illustrated in a schematic
format in Figure 1a, which depicts the operating process. If there is no selector device, the
sneak currents will flow through the devices in the surrounding region that have not been
selected, as shown by the red arrow line. This will result in an error in the operation of the
memristor crossbar array. To suppress the sneak current, we developed the back-to-back
Schottky diode with a Pd/a-ITZO/Pd structure on SiO2 /Si substrate. The Schottky diode
should operate the thermionic emission over the Schottky barrier while both thermionic
emission and thermionic field emission across the Schottky barrier are dominated in the
high-voltage region, as shown in Figure 1b. Note that the conduction mechanism in the
low-voltage region only works with thermionic emission and is not dominated by tunneling
current across the Schottky barrier to suppress the sneak current in the memristor crossbar
array by increasing nonlinearity of the Schottky diode-based selector device. This results in
a high level of nonlinearity of I-V characteristics. Generally, tunneling current across the
Schottky barrier happens when the width of the Schottky barrier is reduced as a result of
a high doping concentration. It is essential to inhibit the formation of oxygen vacancy in
OS material, also known as dopant, in order to achieve the high nonlinearity of I-V that is
achieved by suppressing the tunneling component. This will allow for the achievement of
high nonlinearity of I-V. Therefore, we employed Pd, which is a noble metal with a high
work function (5.22 eV) [38] when compared to the electron affinity of a-ITZO (4.5 eV).
This is due to the fact that the noble Pd electrode has a very high Gibbs free energy of
oxide formation.
The fabricated a-ITZO-selector device is shown in Figure 1c as a cross-sectional high-
resolution transmission electron microscopy (HRTEM) image. It was confirmed that an
a-ITZO layer with a thickness of 30 nm was formed uniformly between the top and bottom
Pd electrodes. The as-deposited a-ITZO-selector device showed the almost ohmic behavior,
which indicates the formation of SEAL (Figure S2). In contrast, the a-ITZO-selector devices,
which had an oxygen plasma treatment applied to the surface of the a-ITZO before the top
Micromachines 2023, 14, 506 4 of 11

Pd electrode was deposited, showed nonlinear I-V characteristics, as shown in Figure 1d.
The I-V characteristics of the oxygen
 plasma-treated a-ITZO-selector can be fitted by the
thermionic emission equation (ln TI2 ∝ V 0.5 ). Figure 1e,f shows the currents in both the
low-voltage and high-voltage regions, where the currents are exponentially proportional
Micromachines 2023, 14, x FOR PEER REVIEW 4 of 11
to voltage on a linear scale, are well fitted to the thermionic emission, indicating that
thermionic emission is the dominant conduction mechanism.

Figure 1.
Figure 1. (a)
(a)Schematic
Schematicofofthethesneak
sneakcurrent path
current in in
path memristor
memristor crossbar array
crossbar during
array operation.
during (b)
operation.
Band diagram and conduction mechanism of back-to-back Schottky diode
(b) Band diagram and conduction mechanism of back-to-back Schottky diode under low- and under low- and high-
voltage regions. (c) Cross-sectional HRTEM image of the a-ITZO-selector device. (d) I–V character-
high-voltage regions. (c) Cross-sectional HRTEM image of the a-ITZO-selector device. (d) I–V
istics of a-ITZO-selector device before and after O2 plasma treatment prior to deposition of top Pd
characteristics of a-ITZO-selector device before and after O2 plasma treatment prior to deposition
electrode. The results of the I-V characteristics are fitted by thermionic emission in (e) the positive-
of top Pd
voltage electrode.
region The
and (f) theresults of the I-V characteristics
negative-voltage are fitted
region. The shaded redby thermionic
regions are theemission in (e)
thermionic the
emis-
positive-voltage region and (f) the negative-voltage region. The shaded red regions
sion region where the current increases exponentially with the voltage on a linear scale. are the thermionic
emission region where the current increases exponentially with the voltage on a linear scale.
This is due to the fact that the noble Pd electrode has a very high Gibbs free energy
These
of oxide characteristics indicate the suppression of current flow in low-voltage regions
formation.
of both positive and
The fabricated negative voltage
a-ITZO-selector while
device permitting
is shown a larger
in Figure 1c as current flow in either
a cross-sectional high-
direction
resolutionintransmission
high-voltage regions.
electron These nonlinear
microscopy (HRTEM)I-Vimage.
characteristics demonstrate
It was confirmed that
that an a-
the oxygen plasma treatment is an effective method for Schottky barrier formation by
ITZO layer with a thickness of 30 nm was formed uniformly between the top and bottom
eliminating SEAL.
Pd electrodes. The as-deposited a-ITZO-selector device showed the almost ohmic behav-
ior, which indicates the formation of SEAL (Figure S2). In contrast, the a-ITZO-selector
devices, which had an oxygen plasma treatment applied to the surface of the a-ITZO be-
fore the top Pd electrode was deposited, showed nonlinear I-V characteristics, as shown
in Figure 1d. The I-V characteristics of the oxygen plasma-treated a-ITZO-selector can be
𝐼
fitted by the thermionic emission equation (𝑙𝑛 ( 2) ∝ 𝑉 0.5 ). Figure 1e,f shows the currents
𝑇
in both the low-voltage and high-voltage regions, where the currents are exponentially
Micromachines 2023, 14, 506 5 of 11

The oxygen plasma treatment process is the essential technology to form the Schottky
barrier of a-ITZO material. We investigated the effect that the oxygen plasma treatment
had on the a-ITZO-selector device by analyzing the electrical properties of an a-ITZO-
selector device to determine how they changed as a function of the time that the device was
exposed to oxygen plasma. Considering that the RF power of oxygen plasma can affect the
penetration depth of the oxygen radical, a plasma process with a power more than 100 W
is capable of oxidizing the surface region of a-ITZO. This oxidation of a-ITZO with the
harsh oxygen plasma treatment results in an increase in the resistivity of its selector device,
which can result in a degradation of the nonlinearity of its I-V characteristics. We adopted
the 100 W of oxygen plasma treatment process to get rid of SEAL solely on the surface of
a-ITZO. When the oxygen plasma treatment condition is maintained for more than 15 min,
the current in the high-voltage region decreases rapidly. On the other hand, the currents
in the positive low-voltage region and negative-voltage region decrease abruptly at the
15 min and the 20 min of oxygen plasma treatment, respectively (Figure 2a). Figure 2b
exhibits the nonlinearities of I-V characteristics which are extracted from the current ratio
at V Read (±1.8 V) and V Read /3 (±0.6 V) in both the positive- and negative-voltage regions.
The memristor crossbar array can be operated with the V/2, V/3, and ground schemes [40].
The optimal bias scheme depends on the selectivity of the selector device and the operating
SET/RESET voltages of the memristor. Considering that we developed the a-ITZO-selector
device for TiO2 -based memristor array, the nonlinearity was evaluated using the V/3.
Among the times of oxygen plasma treatment processes, the a-ITZO-selector device with
the oxygen treatment time of 10 min exhibited the best nonlinearity value of 914 and 280 for
the positive voltage and negative voltage regions, respectively. The worst value of about 70
of the nonlinearity is attributed to the partial removal of SEAL by the insufficient oxygen
plasma treatment. On the other hand, the degraded nonlinearity caused by harsh oxygen
plasma treatment for more than 15 min results from a significant decrease in current level in
the high-voltage region, which is caused by the suppression of thermionic field emission by
the reduced completely oxygen vacancy. In addition, in our previous study, we analyzed the
chemical composition of a-ITZO film with oxygen plasma treatment using XPS [41]. It was
observed that the O 1s core level at 531.9 eV related to O-O bonding absorbed on the a-ITZO
surface increases by oxygen plasma treatment. Other research groups have also reported
that oxygen adsorbed on the oxide semiconductor surface eliminates the SEAL [37,42].
These findings indicate that the naturally formed SEAL can be eliminated using the optimal
oxygen plasma. An atomic force microscopy (AFM) investigation was carried out so that
the effect of the oxygen plasma treatment that was applied to the a-ITZO surface could be
investigated in further detail. The AFM topographic images of the a-ITZO film before and
after being treated with oxygen plasma are represented in Figure 2c, and d, respectively.
When comparing the surface roughness of 2.6 nm after oxygen plasma treatment for 10 min
with the surface roughness of 2.4 nm before oxygen treatment, it was confirmed that 10 min
of oxygen plasma did not significantly affect the microstructure/topology of the a-ITZO
material surface. In addition, the surface roughness is 2.6 nm and 2.8 nm for 15 min
and 20 min of oxygen plasma treatment at 100 W, respectively (Figure S3). The electrical
characteristics and AFM analyses indicate that the optimized oxygen plasma treatment
technique is capable of forming a Schottky barrier in the a-ITZO-selector device without
damaging the microstructure/topology of surface of a-ITZO material.
Micromachines 2023, 14, x FOR PEER REVIEW 6 of 11
Micromachines 2023, 14, 506 6 of 11

Figure
Figure2.2.(a)
(a)I-V
I-V characteristics of a-ITZO-selector
characteristics of a-ITZO-selectordevice
devicedepending
depending ononthethe process
process time
time of oxygen
of oxygen
plasma treatment. As shown in the red arrow, the current decreases as the O 2 plasma time increases.
plasma treatment. As shown in the red arrow, the current decreases as the O2 plasma time increases.
(b)
(b)Nonlinearity
Nonlinearity value of a-ITZO-selector
value of a-ITZO-selectordevice
deviceasasa afunction
function
ofof oxygen
oxygen plasma
plasma process
process time.
time. AMF AMF
image of a-ITZO surface (c) before and (d) after oxygen plasma
image of a-ITZO surface (c) before and (d) after oxygen plasma treatment. treatment.

ToToinvestigate
investigate the the electrical
electricalstability
stabilityofofthe thea-ITZO-selector
a-ITZO-selector device
device fabricated
fabricated through
through
oxygen
oxygen plasma treatmentfor
plasma treatment for1010
min min at 100
at 100 W, cycling
W, cycling endurance,
endurance, and constant
and constant bias sta-
bias stability
tests tests
bility were were
performed. FigureFigure
performed. 3a exhibits a representative
3a exhibits I-V characteristic
a representative of the a-ITZO-
I-V characteristic of the a-
selector device. As shown in the inset of Figure 3a, the linear I-V plot
ITZO-selector device. As shown in the inset of Figure 3a, the linear I-V plot indicates the indicates the apparent
rectifyingrectifying
apparent characteristics that result that
characteristics fromresult
the formation
from theofformation
a Schottkyofbarrier. The results
a Schottky barrier.ofThe
the cycle endurance test that was carried out under the repeated
results of the cycle endurance test that was carried out under the repeated voltage voltage pulses are shown
pulses
in Figure 3b. This test was carried out to validate the stability of the
are shown in Figure 3b. This test was carried out to validate the stability of the Schottky Schottky barrier that
was formed utilizing the oxygen plasma treatment. In this cycling endurance test, voltage
barrier that was formed utilizing the oxygen plasma treatment. In this cycling endurance
pulses with 100 ns width and ±3 V amplitude were applied over 107 cycles. While the
test, voltage pulses with 100 ns width and ± 3 V amplitude were applied over 107 cycles.
repeated pulse voltages were being applied, the currents at the read voltages V Read (1.8 V)
While the repeated pulse voltages were being applied, the currents at the read voltages
and V Read /3 (0.6 V) were being measured. The currents at V Read and V Read /3 was sustained
Vwithout
Read (1.8 V) and VRead/3 (0.6 V) were being measured. The currents at VRead and VRead/3 was
significant degradation, indicating that the nonlinearity of I-V characteristics was
sustained
preserved.without significant
The constant degradation,
bias stability test wasindicating
conducted thatfor 10the
4 s nonlinearity
under a constant of I-V charac-
voltage
teristics was preserved.
bias in order The constant
to further validate bias stability
the stability of Schottkytestbarrier
was conducted
(Figure 3c).forThe10constant
4 s under a
constant voltage bias in order to further validate the stability of Schottky
bias instability test was performed by applying a constant voltage of 3 V until a certain time, barrier (Figure
3c). Theisconstant
which the samebias
as the instability
general bias testinstability
was performed by applying
test of tin-film a constant
transistor. Becausevoltage of 3 V
of electron
until a certain
trapping at thetime, which is
Pd/a-ITZO the same
interface, theasI-V
the general
curves biasa-ITZO-selector
of the instability testdevice
of tin-film
shiftedtran-
toward
sistor. the positive
Because voltagetrapping
of electron region byat300 themV when subjected
Pd/a-ITZO interface,to athe
constant bias stress,
I-V curves of the a-
as shown in Figure
ITZO-selector device3c.shifted
However,towardthe nonlinearity
the positiveofvoltageI-V characteristics
region by 300 wasmVmaintained
when sub-
without significant degradation, indicating the stable formation
jected to a constant bias stress, as shown in Figure 3c. However, the nonlinearity of the Schottky barrier.ofIn I-V
addition, in order to investigate the temperature stability of the formed
characteristics was maintained without significant ◦degradation, indicating the stable for- Schottky barrier, we
performed
mation of thetheSchottky
stability barrier.
test at theIntemperature
addition, in of 85 C.
order to As shown inthe
investigate Figure 3d, although
temperature stabil-
the a-ITZO-selector device was tested for 1000 s at the temperature of 85 ◦ C, the selector
ity of the formed Schottky barrier, we performed the stability test at the temperature of
device did not experience a significant degradation. As electrical and temperature stability
85 ℃. As shown in Figure 3d, although the a-ITZO-selector device was tested for 1000 s at
of the selector device are critical requirements, these results indicate that the oxygen plasma
the temperature of 85 ℃, the selector device did not experience a significant degradation.
treated a-ITZO-selector device can be suitable for a high-density memristor application.
As electrical and temperature stability of the selector device are critical requirements,
these results indicate that the oxygen plasma treated a-ITZO-selector device can be suita-
ble for a high-density memristor application.
Micromachines 2023, 14, x FOR PEER REVIEW 7 of 11
Micromachines 2023, 14, 506 7 of 11

Figure
Figure 3.
3. (a) I-V characteristics
(a) I-V characteristicsofofoxygen
oxygenplasma
plasma treated
treated a-ITZO-selector
a-ITZO-selector device.
device. TheThe
insetinset shows
shows
the linear I-V plot. Electrical stability tests under (b) cycling endurance and (c) constant bias stress.
the linear I-V plot. Electrical stability tests under (b) cycling endurance and (c) constant bias stress.
(d) Temperature stability test.
(d) Temperature stability test.

To evaluate the
To evaluate the performance
performanceofofthe thememristor
memristor crossbar
crossbar integrated
integrated a-ITZO
a-ITZO selector
selector
device,
device, the read margin,which
read margin, whichisisrelated
relatedtotothe
themaximum
maximum size
size of of
thethe memristor
memristor array,
array, waswas
evaluated. To
evaluated. To investigate
investigatethe theread
readmargin,
margin,we we utilized
utilized thetheTiO 2 -based
TiO 2-basedmemristor
memristor with a a
with
structure of Al/TiO /AI which shows bipolar resistive switching
structure of Al/TiO22/AI which shows bipolar resistive switching [39] (Figure 4a). Due to [39] (Figure 4a). Due
to the
the modification
modification of the
of the interfacial
interfacial AI-Ti-O
AI-Ti-O layerlayer
by by oxygen
oxygen ionion migration,
migration, thethe
TiO TiO 2-
2-based
based memristor operates with the positive-voltage region with the
memristor operates with the positive-voltage region with the SET process and in the neg- SET process and in the
negative-voltage
ative-voltage region
region with
with theRESET
the RESETprocess.
process. TheThe TiO TiO2 -based memristor is suitable for
2-based memristor is suitable for
calculating the read margin, which can be improved by a-ITZO-selector, as the SET and
calculating the read margin, which can be improved by a-ITZO-selector, as the SET and
RESET voltages of TiO -meristor are in the voltage region where the high current flows
RESET voltages of TiO22-meristor are in the voltage region where the high current flows in
in nonlinearity of a-ITZO-selector. For the reading operation of the memristor device in
nonlinearity of a-ITZO-selector. For the reading operation of the memristor device in the
the crossbar array, ground, V/2, and V/3 reading bias schemes were generally utilized.
crossbar array, ground,allV/2,
In the ground scheme, the and V/3 reading
unselected top and bias schemes
bottom lineswere
weregenerally
grounded.utilized. In the
The V Read
ground scheme, all the unselected top and bottom lines were grounded.
voltages are only applied to the devices with a selected top line, where most sneak currents The V Read voltages

are
are only applied
generated. Alltounselected
the devices topwith
and abottom
selected topwere
lines line,biased
whereatmost sneak
half the readcurrents
voltage are
generated. All unselected top and bottom lines were biased
(V Read /2) in the V/2 scheme, permitting sneak currents to flow through the half-selected at half the read voltage
(V Read/2) in
devices in the
the selected
V/2 scheme,
top andpermitting sneakIncurrents
bottom lines. the V/3 to flow through
scheme, the half-selected
all the unselected top
devices in thelines
and bottom selected
weretop and at
biased bottom lines.(VIn
one-third the
Read /3) V/3 andscheme, all the(2V
two-thirds unselected
Read /3) oftoptheand
read voltage,
bottom generating
lines were biasedsneak currents (V
at one-third across
Read/3)theand
selected
two-thirdsdevices (2Vwith the of
Read/3) voltage dropvolt-
the read
magnitude
age, generatingof V Read
sneak/3,currents
with theacross
exception of the selected
the selected devices cells
with in the
the voltage
crossbardroparray,magni-
as
depicted in Figure 4b. In the numerical calculation, the worst-case scenario
tude of VRead/3, with the exception of the selected cells in the crossbar array, as depicted in (all unselected
cells with
Figure 4b. low resistance
In the numericalstate) was considered,
calculation, and the three
the worst-case reading
scenario (all bias schemescells
unselected werewith
assessed
low to determine
resistance state) wastheconsidered,
optimal reading biasthree
and the scheme.reading MATLAB was utilized
bias schemes weretoassessed
obtain to
the numerical solution to the reading margin computation. The maximum crossbar array
determine the optimal reading bias scheme. MATLAB was utilized to obtain the numeri-
size was determined using a 10% reading margin as the criterion. To access the information
cal solution to the reading margin computation. The maximum crossbar array size was
stored in the selected device, a pull-up voltage (Vpull ) through pull-up resistance (Rpull ) is
determined
applied to the using a 10%
selected topreading margin
line, while as the criterion.
the selected bottom line Toisaccess the information
grounded. Depending stored
on
in the selected device, a pull-up voltage (V pull) through pull-up resistance (Rpull) is applied
the reading scheme, the regions of the unselected devices can be categorized as Region 1,
to the selected
Region top line,
2, and Region whilecan
3, which thebeselected
expressed bottom
by three lineresistor
is grounded.
networks Depending
as depictedon in the
reading
Figure 4c. For ground scheme, the applied voltages of Region 1 and Region 2 are ground,Re-
scheme, the regions of the unselected devices can be categorized as Region 1,
gion 2, and Region 3, which can be expressed by three resistor networks as depicted in
Figure 4c. For ground scheme, the applied voltages of Region 1 and Region 2 are ground,
Micromachines 2023, 14, x FOR PEER REVIEW 8 of 11

Micromachines 2023, 14, 506 whereas the applied voltage of Region 3 is VRead. For V/2 scheme, the applied voltages
8 of 11for
Region 1 and Region 3 are VRead/2, while for V/3 scheme, the applied voltages for all re-
gions are VRead/3. We determined the reading margin based on the above three resistor
networks by solving
whereas the applied the following
voltage equation:
of Region 3 is V Read . For V/2 scheme, the applied voltages
for Region 𝑉𝑜𝑢𝑡,𝐿𝑅𝑆 3 are V Read /2, while for V/3 scheme, the applied voltages for all
1 and−Region
𝑉𝑜𝑢𝑡,𝐻𝑅𝑆
R. M. = are V Read /3. We determined the reading margin based on the above three resistor
regions 𝑉𝑝𝑢𝑙𝑙
networks by solving the 𝑅(𝐻𝑅𝑆)
following equation:
∥ (𝑅 1 + 𝑅2 + 𝑅3 ) 𝑅(𝐿𝑅𝑆) ∥ (𝑅1 + 𝑅2 + 𝑅3 )
= −
𝑅
𝑝𝑢𝑙𝑙 + 𝑅(𝐻𝑅𝑆) ∥ (𝑅 + 𝑅 + 𝑅 ) 𝑅 + 𝑅(𝐿𝑅𝑆) ∥ (𝑅 + 𝑅 + 𝑅 )
Vout,HRS − Vout,LRS R( HRS) k1( R1 + 2
R2 +3 R3 ) 𝑝𝑢𝑙𝑙 R( LRS) k ( R11 + R22 + R33 )
R. M. = = −
where 𝑉 V R
ispullthe voltagepull + R ( HRS ) k (
drop at the output R 1 + R + R
2node ) R pull +
3 caused byR(the ) k ( R1 + Rdevice
LRSselected 2 + R3 ) in a
𝑜𝑢𝑡,𝐻𝑅𝑆
high-resistance state (HRS), whereas 𝑉𝑜𝑢𝑡,𝐿𝑅𝑆 is the voltage drop at the output node
where Vout,HRS is the voltage drop at the output node caused by the selected device in
caused by the selected
a high-resistance state device
(HRS), in a low-resistance
whereas statevoltage
Vout,LRS is the (LRS). drop
We calculated
at the output the node
reading
margin using the LRS/HRS resistances and SET/RESET voltages of
caused by the selected device in a low-resistance state (LRS). We calculated the readingTiO 2-memristor and

the resistance
margin of a-ITZO
using the LRS/HRS -selector. It was
resistances andconfirmed
SET/RESET that the ground
voltages of TiO2scheme
-memristor hasand
thethe
max-
imum reading margin (Figure 4d). In contrast, the V/2 and V/3 schemes
resistance of a-ITZO -selector. It was confirmed that the ground scheme has the maximum show the reading
margin
readingdegradation
margin (Figure as the
4d).array size increases.
In contrast, the V/2 andTheV/3
obvious
schemesdifferences
show thein reading
reading margin
margin
among the three
degradation reading
as the schemes
array size are attributed
increases. The obvious todifferences
the sourceinofreading
the sneak currents.
margin amongThe
the three
ground reading
scheme schemesleakage
generates are attributed
currenttoonly
the source
from the of devices
the sneak currents.
with The ground
the selected top line,
scheme generates leakage current only from the devices with the selected
while the other two reading schemes induce leakage currents from half and one-third top line, while the of
other two reading schemes induce leakage currents from half and one-third
the selected devices on the memristor crossbar array. In addition, the sneak current de- of the selected
devices
grades onSET
the the memristor crossbar
voltage window array.isIn
which addition,
defined by the
thesneak current
difference degrades
between thethe
SETSET
volt-
voltage window which is defined by the difference between the SET
age delivered to the selected cell and the maximum voltage applied to the most disturbed voltage delivered
to the selected cell and the maximum voltage applied to the most disturbed unselected
unselected cell [43]. Thanks to the selector device, the TiO2-memristor shows the only
cell [43]. Thanks to the selector device, the TiO2 -memristor shows the only about 25%
about 25% degradation of the SET voltage window, which is calculated using numerical
degradation of the SET voltage window, which is calculated using numerical simulation
simulation through MATLAB [43] (Figure S4). It is noteworthy that the size of the memris-
through MATLAB [43] (Figure S4). It is noteworthy that the size of the memristor array
tor
witharray with integrated
integrated a-ITZO-selector
a-ITZO-selector device candevice can be to
be expanded expanded
1 Mbit bytoutilizing
1 Mbit by twoutilizing
V/2
two V/2 and V/3
and V/3 methods. methods.

(a) I-V
Figure4.4.(a)
Figure I-V characteristics
characteristics of
ofTiO
TiO2 -memristor and a-ITZO-selector device. (b) V/3 scheme for
2-memristor and a-ITZO-selector device. (b) V/3 scheme for
operationofofthe
operation thememristor
memristor crossbar
crossbar array.
array. (c)
(c)Equivalent
Equivalentcircuit
circuitofofanan
NNXNXNmemristor crossbar
memristor crossbar
array to calculate the reading margin. (d) Calculated reading margin according to ground,
array to calculate the reading margin. (d) Calculated reading margin according to ground, V/2, V/2, andand
V/3 schemes as a function of array
V/3 schemes as a function of array size. size.
Micromachines 2023, 14, 506 9 of 11

4. Conclusions
In conclusion, we have demonstrated the feasibility of a-ITZO-selector device for a
memristor crossbar array. To develop the back-to-back Schottky diode type selector device
based on a-ITZO material, we employed an oxygen plasma treatment prior to Pd top
electrode deposition in order to eliminate SEAL. The effect of time condition on the oxygen
plasma treatment process for the a-ITZO-selector was explored, revealing that optimizing
the process time is crucial for attaining nonlinear I-V characteristics. The a-ITZO-selector
device that has been treated with oxygen plasma exhibits nonlinear I-V characteristics,
stable cycling endurance, constant bias stability, and temperature stability. In addition,
to prove the practical application of selector devices, the integrated TiO2 -memristor and
a-ITZO-selector device array demonstrates that its memristor array size can rise to 1 Mbit
in density. Compared to the state-of-the-art selector device for memristor crossbar array,
the a-ITZO-selector device showed the comparable reliability against electrical stress. It is
noted that the a-ITZO-selector device exhibited temperature stability at 85 ◦ C, thanks to
the intrinsic material property of an oxide semiconductor. Other selector devices including
threshold switching based on MIT transition and the fast relaxation of Ag filament suffer
from the poor temperature stability by their intrinsic material property. Therefore, we
believe that these results on the a-ITZO-selector device can give a viable approach for
realizing a high-density, reliable memristor array.

Supplementary Materials: The following supporting information can be downloaded at: https://
www.mdpi.com/article/10.3390/mi14030506/s1, Figure S1: XPS depth profile result of the deposited
a-ITZO film; Figure S2: Linear plot of I-V characteristics of a-ITZO-selector with and without oxygen
plasma treatment; Figure S3: AMF image of oxygen plasma traeted a-ITZO film with (a) 15 min and
(d) 20 min at 100 W; Figure S4: Calculated SET voltage window with and without selector device as a
function of array size.
Author Contributions: Conceptualization, B.C.J. and M.-J.S.; validation, K.H.K., M.-J.S.; writing—
review and editing, Data curation, K.H.K. and M.-J.S.; Supervision, B.C.J., M.-J.S. All authors have
read and agreed to the published version of the manuscript.
Funding: This research was supported by Kyungpook National University Research Fund, 2021.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Waser, R.; Aono, M. Nanoionics-based resistive switching memories. Nat. Mater. 2007, 6, 833–840. [CrossRef]
2. Lee, M.J.; Lee, C.B.; Lee, D.; Lee, S.R.; Chang, M.; Hur, J.H.; Kim, Y.B.; Kim, C.J.; Seo, D.H.; Seo, S.; et al. A fast, high-endurance
and scalable non-volatile memory device made from asymmetric Ta2 O5-x /TaO2-x bilayer structures. Nat. Mater. 2011, 10, 625–630.
[CrossRef]
3. Li, Y.; Loh, L.; Li, S.; Chen, L.; Li, B.; Bosman, M.; Ang, K.-W. Anomalous resistive switching in memristors based on two-
dimensional palladium diselenide using heterophase grain boundaries. Nat. Electron. 2021, 4, 348–356. [CrossRef]
4. Simanjuntak, F.M.; Hsu, C.-L.; Abbey, T.; Chang, L.-Y.; Rajasekaran, S.; Prodromakis, T.; Tseng, T.-Y. Conduction channel
configuration controlled digital and analog response in TiO2 -based inorganic memristive artificial synapses. APL Mater. 2021, 9,
121103. [CrossRef]
5. Wang, H.; Yan, X.; Wang, S.; Lu, N. High-Stability Memristive Devices Based on Pd Conductive Filaments and Its Applications in
Neuromorphic Computing. ACS Appl. Mater. Interfaces 2021, 13, 17844–17851. [CrossRef]
6. Balatti, S.; Ambrogio, S.; Ielmini, D. Normally-off Logic Based on Resistive Switches—Part I: Logic Gates. IEEE Trans. Electr. Dev.
2015, 62, 1831–1838. [CrossRef]
7. Cassinerio, M.; Ciocchini, N.; Ielmini, D. Logic computation in phase change materials by threshold and memory switching. Adv.
Mater. 2013, 25, 5975–5980. [CrossRef] [PubMed]
8. Jang, B.C.; Yang, S.Y.; Seong, H.; Kim, S.Y.; Choi, J.; Im, S.G.; Choi, S.-Y. Zero-Static-Power Nonvolatile Logic-in-Memory Circuits
for Flexible Electronics. Nano Res. 2017, 10, 2459–2740. [CrossRef]
9. Borghetti, J.; Snider, G.S.; Kuekes, P.J.; Yang, J.J.; Stewart, D.R.; Williams, R.S. ‘Memristive’ switches enable ‘stateful’ logic
operations via material implication. Nature 2010, 464, 873–876. [CrossRef] [PubMed]
Micromachines 2023, 14, 506 10 of 11

10. Strukov, D.B.; Likharev, K.K. Defect-Tolerant Architectures for Nanoelectronic Crossbar Memories. J. Nanosci. Nanotechnol. 2007,
7, 151–167. [CrossRef]
11. Snider, G. Computing with hysteretic resistor crossbars. Appl. Phys. A 2005, 80, 1165–1172. [CrossRef]
12. Yao, P.; Wu, H.; Gao, B.; Tang, J.; Zhang, Q.; Zhang, W.; Yang, J.J.; Qian, H. Fully hardware-implemented memristor convolutional
neural network. Nature 2020, 577, 641–646. [CrossRef]
13. Li, H.; Gao, B.; Chen, Z.; Zhao, Y.; Huang, P.; Ye, H.; Liu, L.; Liu, X.; Kang, J. A learnable parallel processing architecture towards
unity of memory and computing. Sci. Rep. 2015, 5, 13330. [CrossRef]
14. Choi, B.J.; Zhang, J.; Norris, K.; Gibson, G.; Kim, K.M.; Jackson, W.; Zhang, M.X.; Li, Z.; Yang, J.J.; Williams, R.S. Trilayer Tunnel
Selectors for Memristor Memory Cells. Adv. Mater. 2016, 28, 356–862. [CrossRef]
15. Wang, Z.; Kang, J.; Bai, G.; Zhong, G.; Wang, B.; Ling, Y.; Chen, Q.; Bao, L.; Wu, L.; Cai, Y.; et al. Self-Selective Resistive Device
With Hybrid Switching Mode for Passive Crossbar Memory Application. IEEE Electr. Dev. Lett. 2020, 41, 1009–1012. [CrossRef]
16. Jo, S.H.; Kumar, T.; Narayanan, S.; Nazarian, H. Cross-Point Resistive RAM Based on Field-Assisted Superlinear Threshold
Selector. IEEE Trans. Electr. Dev. 2015, 62, 3477–3481. [CrossRef]
17. Li, H.; Wang, S.; Zhang, X.; Wang, W.; Yang, R.; Sun, Z.; Feng, W.; Lin, P.; Wang, Z.; Sun, L.; et al. Memristive Crossbar Arrays for
Storage and Computing Applications. Adv. Intell. Syst. 2021, 3, 2100017. [CrossRef]
18. Youn, S.; Kim, S.; Kim, T.-H.; Park, J.; Kim, H. Memristor Crossbar Circuit for Ternary Content-Addressable Memory with
Fine-Tuning Operation. Adv. Intell. Syst. 2023, 220325. [CrossRef]
19. Lee, M.J.; Lee, D.; Cho, S.H.; Hur, J.H.; Lee, S.M.; Seo, D.H.; Kim, D.S.; Yang, M.S.; Lee, S.; Hwang, E.; et al. A plasma-treated
chalcogenide switch device for stackable scalable 3D nanoscale memory. Nat. Commun. 2013, 4, 2629. [CrossRef]
20. Huang, J.-J.; Tseng, Y.-M.; Hsu, C.-W.; Hou, T.-H. Bipolar Nonlinear Ni/TiO2 /Ni Selector for 1S1R Crossbar Array Application.
IEEE Electr. Dev. Lett. 2011, 32, 1427–1429. [CrossRef]
21. Hsieh, C.-C.; Chang, Y.-F.; Chen, Y.-C.; Shahrjerdi, D.; Banerjee, S.K. Highly Non-linear and Reliable Amorphous Silicon Based
Back-to-Back Schottky Diode as Selector Device for Large Scale RRAM Arrays. ECS J. Solid State Sci. Technol. 2017, 6, N143–N147.
[CrossRef]
22. Shuang, Y.; Hatayama, S.; An, J.; Hong, J.; Ando, D.; Song, Y.; Sutou, Y. Bidirectional Selector Utilizing Hybrid Diodes for PCRAM
Applications. Sci. Rep. 2019, 9, 20209. [CrossRef] [PubMed]
23. Wu, C.; Feng, F.; Xie, Y. Design of vanadium oxide structures with controllable electrical properties for energy applications. Chem.
Soc. Rev. 2013, 42, 5157–5183. [CrossRef]
24. Kim, S.; Son, J.H.; Lee, S.H.; You, B.K.; Park, K.I.; Lee, H.K.; Byun, M.; Lee, K.J. Flexible Crossbar-Structured Resistive Memory
Arrays on Plastic Substrates via Inorganic-Based Laser Lift-Off. Adv. Mater. 2014, 26, 7480–7487. [CrossRef]
25. Huang, J.-J.; Tseng, Y.-M.; Luo, W.-C.; Hsu, C.-W.; Hou, T.-H. One Selector-One Resistor (1S1R) Crossbar Array for High-density
Flexible Memory Applications. In Proceedings of the 2011 International Electron Devices Meeting, Washington, DC, USA, 5–7
December 2011; pp. 733–736.
26. Hosono, H. Ionic amorphous oxide semiconductors: Material design, carrier transport, and device application. J. Non-Cryst.
Solids 2006, 352, 851–858. [CrossRef]
27. Chiang, H.Q.; Wager, J.F.; Hoffman, R.L.; Jeong, J.; Keszler, D.A. High mobility transparent thin-film transistors with amorphous
zinc tin oxide channel layer. Appl. Phys. Lett. 2005, 86, 013503. [CrossRef]
28. Park, J.-S.; Kim, H.; Kim, I.-D. Overview of electroceramic materials for oxide semiconductor thin film transistors. J. Electroceram.
2013, 32, 117–140. [CrossRef]
29. Nomura, K.; Kamiya, T.; Ikenaga, E.; Yanagi, H.; Kobayashi, K.; Hosono, H. Depth analysis of subgap electronic states in
amorphous oxide semiconductor, a-In-Ga-Zn-O, studied by hard x-ray photoelectron spectroscopy. J. Appl. Phys. 2011, 109,
073726. [CrossRef]
30. Walsh, A. Surface oxygen vacancy origin of electron accumulation in indium oxide. Appl. Phys. Lett. 2011, 98, 261910. [CrossRef]
31. Bierwagen, O.; Speck, J.S.; Nagata, T.; Chikyow, T.; Yamashita, Y.; Yoshikawa, H.; Kobayashi, K. Depletion of the In2 O3 (001) and
(111) surface electron accumulation by an oxygen plasma surface treatment. Appl. Phys. Lett. 2011, 98, 172101. [CrossRef]
32. von Wenckstern, H.; Splith, D.; Schmidt, F.; Grundmann, M.; Bierwagen, O.; Speck, J.S. Schottky contacts to In2 O3 . APL Mater.
2014, 2, 046104. [CrossRef]
33. Nakano, M.; Tsukazaki, A.; Gunji, R.Y.; Ueno, K.; Ohtomo, A.; Fukumura, T.; Kawasaki, M. Schottky contact on a ZnO (0001)
single crystal with conducting polymer. Appl. Phys. Lett. 2007, 91, 142113. [CrossRef]
34. Schifano, R.; Monakhov, E.V.; Grossner, U.; Svensson, B.G. Electrical characteristics of palladium Schottky contacts to hydrogen
peroxide treated hydrothermally grown ZnO. Appl. Phys. Lett. 2007, 91, 193507. [CrossRef]
35. Schifano, R.; Monakhov, E.V.; Svensson, B.G.; Diplas, S. Surface passivation and interface reactions induced by hydrogen peroxide
treatment of n-type ZnO (0001). Appl. Phys. Lett. 2009, 94, 132101. [CrossRef]
36. Mosbacker, H.L.; Strzhemechny, Y.M.; White, B.D.; Smith, P.E.; Look, D.C.; Reynolds, D.C.; Litton, C.W.; Brillson, L.J. Role of
near-surface states in ohmic-Schottky conversion of Au contacts to ZnO. Appl. Phys. Lett. 2005, 87, 012102. [CrossRef]
37. Nagata, T.; Bierwagen, O.; White, M.E.; Tsai, M.-Y.; Speck, J.S. Study of the Au Schottky contact formation on oxygen plasma
treated n-type SnO2 (101) thin films. J. Appl. Phys. 2010, 107, 033707. [CrossRef]
38. Migani, A.; Sousa, C.; Illas, F. Chemisorption of atomic chlorine on metal surfaces and the interpretation of the induced work
function changes. Surf. Sci. 2005, 574, 297–305. [CrossRef]
Micromachines 2023, 14, 506 11 of 11

39. Jeong, H.Y.; Kim, S.K.; Lee, J.Y.; Choi, S.-Y. Impact of amorphous titanium oxide film on the device stability of Al/TiO2/Al
resistive memory. Appl. Phys. A 2011, 102, 967–972. [CrossRef]
40. Kim, S.; Kim, H.-D.; Choi, S.-J. Numerical study of read scheme in one-selector one-resistor crossbar array. Solid State Electron.
2015, 114, 80–86. [CrossRef]
41. Jang, B.C.; Nam, Y.; Koo, B.J.; Choi, J.; Im, S.G.; Park, S.-H.K.; Choi, S.-Y. Memristive Logic-in-Memory Integrated Circuits for
Energy-Efficient Flexible Electronics. Adv. Funt. Mater. 2018, 28, 1704725. [CrossRef]
42. Kim, K.-P.; Hussain, A.M.; Hwang, D.-K.; Woo, S.-H.; Lyu, H.-K.; Baek, S.-H.; Jang, Y.; Kim, J.-H. Work Function Modification of
Indium–Tin Oxide by Surface Plasma Treatments Using Different Gases. Jpn. J. Appl. Phys. 2009, 48, 021601. [CrossRef]
43. Sungho, K.; Jiantao, Z.; Lu, W.D. Crossbar RRAM Arrays: Selector Device Requirements During Write Operation. IEEE Trans.
Electr. Dev. 2014, 61, 2820–2826. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.