DSpace at VNU: Understanding Electrical Conduction States in WO3 Thin Films Applied for Resistive Random-Access Memory

Journal of ELECTRONIC MATERIALS

DOI: 10.1007/s11664-016-4361-4
Ó 2016 The Minerals, Metals & Materials Society

Understanding Electrical Conduction States in WO3 Thin Films
Applied for Resistive Random-Access Memory
THI KIEU HANH TA,1 KIM NGOC PHAM,1 THI BANG TAM DAO,1
DAI LAM TRAN,2 and BACH THANG PHAN1,3,4
1.—Faculty of Materials Science, University of Science, Vietnam National University, Ho Chi
Minh City, Viet Nam. 2.—Vietnam Academy of Science and Technology, Ha Noi, Viet Nam.
3.—Laboratory of Advanced Materials, University of Science, Vietnam National University, Ho
Chi Minh City, Viet Nam. 4.—e-mail:

The electrical conduction and associated resistance switching mechanism of
top electrode/WO3/bottom electrode devices [top electrode (TE): Ag, Ti; bottom
electrode (BE): Pt, fluorine-doped tin oxide] have been investigated. The
direction of switching and switching ability depended on both the top and
bottom electrode material. Multiple electrical conduction mechanisms control
the leakage current of such switching devices, including trap-controlled spacecharge, ballistic, Ohmic, and Fowler–Nordheim tunneling effects. The transition between electrical conduction states is also linked to the switching
(SET–RESET) process. This is the first report of ballistic conduction in research into resistive random-access memory. The associated resistive
switching mechanisms are also discussed.
Key words: Resistive switching, WO3 thin films, trap-controlled spacecharge limited conduction, Fowler–Nordheim tunneling,
ballistic conduction

INTRODUCTION
In terms of nonvolatile memory, it is generally
believed that transistor-based flash memory will
approach the end of scaling within about a decade.
As a result, non-field-effect transistor (FET)-based
devices and architectures will likely be needed to

satisfy growing demand for high-performance memory and logic electronics applications.1 Recent
research has demonstrated that nonvolatile resistance-switching resistive random-access memory
(ReRAM) is a promising alternative to floating-gate
technology beyond the 32-nm technology node. In
addition, transparent electronics is one of the most
important emerging technologies for next-generation electronics systems. Oxide-based ReRAM structures exploit the functionality of capacitor
structures in which oxide materials, for example,
perovskites (Cr-doped SrTiO3, Cr-doped SrZrO3,
Pr0.7Ca0.3MnO3, etc.),1–8 chalcogenide materials

(Received October 10, 2015; accepted January 16, 2016)

(GeSbTe),9 transition-metal oxides (TMOs), and
ordinary oxides (NiO, TiO2, CuOx, HfO2, ZrOx,
ZnO, Cr2O3, WO3),10–23 are sandwiched between
two metal electrodes. Choosing a material that is
compatible with complementary metal–oxide–semiconductor (CMOS) processes is currently a crucial
challenge in ReRAM research. Among the various
materials used, TMOs have the major advantages of
simple fabrication and compatibility with CMOS
processes. These TMOs are mostly transparent,
with wide bandgap energies, making ReRAM a
good candidate for enabling transparent memory.19
The switching phenomena in these material systems remain controversial. There are many publications addressing different switching behaviors,
even in the same system, making understanding
even more difficult.20,21 Understanding the electrical conduction behavior of such materials is an
important step in the design of applications using
these materials. Previously, we reported the correlation between the crystallinity and resistive
switching behavior of sputtered WO3 thin films.16
In this study, we investigated the electrical

Ta, Pham, Dao, Tran, and Phan

Fig. 1. I–V curves of (a) Ag/WO3/Pt, (b) Ag/WO3/FTO, (c) Ti/WO3/Pt, and (d) Ti/WO3/FTO devices.

conduction of WO3 thin films using different electrode materials purposely chosen to improve understanding of the switching mechanism.

0.02 V. The bottom electrode was biased, while the
top electrode was grounded.

EXPERIMENTAL PROCEDURES

RESULTS AND DISCUSSION

Ag, Ti, and tungsten oxide films were fabricated
on fluorine-doped tin oxide (FTO) and Pt/Ti/SiO2/Si
substrates using direct-current (DC) sputtering at
room temperature using metallic Ag, Ti, and W
targets. Deposition of 300-nm-thick WO3 thin films
was carried out under total pressure PAr+O2 of
7 9 10À3 Torr, 300°C, and mixture ratio of oxygen
to argon gas (PO2/PAr+O2) fixed at 90%. During the
deposition of the 100-nm-thick top electrode (Ag or
Ti) in argon environment at 7 9 10À3 Torr, a mask
was used for top electrode patterning. The crystalline phases of the thin films were characterized
in h–2h mode using a D8 Advance (Bruker) x-ray
diffractometer (XRD) with Cu Ka radiation
(k = 0.154 nm) and by Fourier-transform infrared
(FTIR) spectroscopy. The surface morphologies of

the films were obtained using field-emission scanning electron microscopy (FESEM). X-ray photoelectron spectroscopy (XPS) was used to investigate
the chemical state of the films. These characterizations were reported previously.16 Current–voltage
(I–V) measurements were carried out using a
semiconductor characterization system (Keithley
4200 SCS) and probe station. I–V curves were
obtained in voltage sweep mode using a
0 fi ±Vmax fi 0 fi ±Vmax fi 0 voltage profile,
sweep speed in normal mode, and step voltage of

Figure 1 shows the I–V characteristics of top
electrode/WO3/bottom electrode devices where the
top electrode (TE) was Ag or Ti and the
bottom electrode (BE) was Pt or FTO. I–V
hysteresis was observed for three devices,
viz. Ag/WO3/Pt, Ag/WO3/FTO, and Ti/WO3/FTO
(Fig. 1a, b and d), whereas no I–V hysteresis
was observed for the other device, i.e., Ti/WO3/Pt
(Fig. 1c). In these switching devices, the direction of
switching depended on both the top and bottom
electrode material. It was observed that, for WO3
devices with FTO bottom electrode, replacing the Ag
top electrode with a Ti electrode caused a change of
the switching direction, while for WO3 devices with
Pt bottom electrode, replacing the Ag top electrode
with a Ti electrode caused suppression of switching.
For both the Ag/WO3/Pt and Ag/WO3/FTO devices
(Fig. 1a and b), the initial high-resistance state
(HRS) was changed to a low-resistance state (LRS)
as negative bias (0 fi ÀVmax) was applied to the Pt
(FTO) bottom electrode. The device remained in the

LRS as the negative bias was decreased and progressively changed to the HRS only on voltage
sweeping
in
the
positive
voltage
range
(0 fi +Vmax).
For the Ti/WO3/FTO device (Fig. 1d), the initial
high-resistance state (HRS) was changed to a low-


Understanding Electrical Conduction States in WO3 Thin Films Applied for Resistive RandomAccess Memory

Fig. 2. I–V curve of HRS of Ag/WO3/Pt device under 0 fi À1.5 V process plotted in terms of: (a) SCLC, (b) BC, (c) SC, (d) PF, and (e) FN
conduction mechanisms.

resistance state (LRS) by application of positive bias
(0 fi +Vmax) to the FTO bottom electrode. The
device remained in the LRS as the positive bias was
decreased and progressively changed to the HRS
only on voltage sweeping in the negative voltage
range (0 fi ÀVmax).
To understand the switching mechanism in these
devices, the dominant leakage processes involved in
the Ag/WO3/Pt, Ag/WO3/FTO, and Ti/WO3/FTO
devices were determined from the measured I–V
data of the HRS and LRS in both polarity biases. All
the I–V curves were examined in terms of various
potential leakage mechanisms: space-charge-limited conduction (SCLC), ballistic conduction (BC),

interface-limited Schottky emission conduction
(SC), interface-limited Fowler–Nordheim (FN) tunneling, and bulk-limited Poole–Frenkel (PF)
emission.24–27

JSCLC ¼ JOhm þ JTFL þ JChild
JOhm $ V
JTrapfilled $ V m

ðm > 2Þ

ð1Þ

JChild $ V 2

JSC

rffiffiffiffiffiffi
2q 3=2
E ;
JBC $
d
0 
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1
Àq ub À qE=4per e0
A;
$ T 2 exp@
kB T

ð3Þ

0

JPF

pffiffiffiffiffiffiffiffiffi 3=2 1
À4
2mà ub A
;
$ E2 exp@
3qhE

ð2Þ

ð4Þ


Ta, Pham, Dao, Tran, and Phan

Fig. 3. I–V curve of LRS of Ag/WO3/Pt device under À1.5 V fi 0 process plotted in terms of: (a) SCLC, (b) BC, (c) SC, (d) PF, and (e) FN
conduction mechanisms.


pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1
Àq ut À qE=per e0
A;
$ E exp@
kB T
0

JFN

ð5Þ

where er is the relative dielectric constant, e0 is the
permittivity of free space, d is the film thickness, and A
is a constant; ub and ut are the height of the Schottky
barrier and trap ionization energy, respectively.
Electrical Conduction Mechanisms
in Switching Devices
For the Ag/WO3/Pt device, the I–V curve obtained
under the 0 fi À1.5 V sweeping process (Fig. 1a)
showed that the leakage current of the HRS followed a nonlinear I–V dependence where one or
more conduction processes may be involved. The
leakage current depended linearly on voltage (I–V),
then increased steeply with voltage (I–V12), followed

by an I–V1.5 dependence (Fig. 2a). This behavior
suggests trap-controlled SCLC as the dominant
leakage process. However, for SCLC (Eq. 1), Ohmic
conduction (OC) at low voltages is followed by a
trap-filled limit region with I–Vm dependence
(m > 2) and trap-free SCLC (Child’s law, I–V2).
The measured I–V data therefore do not completely
fit with SCLC at high electric fields. We carried out
further analysis and found that the I–V1.5 dependence can be classified as ballistic conduction
(Fig. 2b, Eq. 2). Therefore, SCLC is largely responsible for the leakage behavior of the Ag/WO3/Pt
device at low electric fields whereas BC dominates
at high electric fields. It is also noticed that ballistic
conduction is rarely reported in published ReRAM
papers. However, in this study, we found ballistic

conduction.
To further determine whether Schottky barriers,
PF emission, or interface-limited FN tunneling were
largely involved, similar analyses were conducted


Understanding Electrical Conduction States in WO3 Thin Films Applied for Resistive RandomAccess Memory

Fig. 4. I–V curve of LRS of Ag/WO3/Pt device under 0 fi +2 V process plotted in terms of: (a) SCLC, (b) BC, (c) SC, (d) PF, and (e, f) FN
conduction mechanisms.

according to Eqs. 3–5. For a linear fit with Eqs. 3
and 4, an appropriate er value is necessary to extract
the optical dielectric permittivity of WO3 from the
slopes of these plots. The refractive index of WO3 is
n % 2.1,28 thus an optical dielectric permittivity of
er = n2 % 4.41 can be reasonably assumed. From the
Schottky and PF plots in Fig. 2c and d, the er value
differs greatly from the value of 4.41 for WO3,
indicating that involvement of Schottky barriers or
PF emission is highly unlikely. Figure 2e shows a
plot of ln J/E2 as a function of 1/E without a
negative slope at high electric fields (marked by
red rectangle), suggesting that interface-limited FN
tunneling is not a dominant process.
Similar analyses were conducted for the sequence
of the sweeping process, À1.5 V fi 0 fi +2 V
and +2 V fi 0, as shown in Figs. 3, 4, and 5. The
I–V curve for the LRS obtained for the À1.5 V to 0
sweeping process shows a linear dependence, a

behavior corresponding to an Ohmic-conductionbased filament path (Fig. 3a). The I–V curve for the
LRS obtained for the 0 fi +2 V sweeping process
shows nonlinear dependence, with electrical conduction corresponding to an Ohmic-conductionbased filament path and FN tunneling, as shown
in Fig. 4a, e and f. The I–V curve is well fit by FN
tunneling conduction with a negative slope (marked
in red rectangle in Fig. 4e and f). It is well known
that FN tunneling conduction is dominant in thin
dielectric films at high electric field, where charge
carriers are injected from the electrode to the
insulator by tunneling through a high potential
barrier. In contrast to the abrupt change of
current with voltage of the HRS under negative
bias, the I–V curve for the HRS under positive
bias (+2 V fi 0) showed nonlinear dependence,
and its electrical conduction follows FN tunneling
(Fig. 5e).


Ta, Pham, Dao, Tran, and Phan

Fig. 5. I–V curve of HRS of Ag/WO3/Pt device under +2 V fi 0 process plotted in terms of: (a) SCLC, (b) BC, (c) SC, (d) PF, and (e) FN
conduction mechanisms.

In summary, the dominant electrical conduction
mechanisms in the Ag/WO3/Pt device under the
0 fi À1.5 V fi 0 fi +2 V fi 0 sweeping process are shown in Fig. 6 and also listed in Table I.
Similar analyses were carried out for the Ag/WO3/
FTO and Ti/WO3/FTO devices, whose electrical
conduction mechanisms are shown in Figs. 7 and 8

and listed in Table I.
All the switching devices showed a transition
from SCLC to BC along with the HRS to LRS
switching. Note that SCLC implies trap levels
within the bandgap. Our previously reported XPS
results show that the WO3 film contains defects
such as nonlattice oxygen ions.16 Defects in the WO3
film could form trap sites in the bandgap below the
conduction band, where injected charge carriers can
be trapped. After saturation of all such defect levels
by injected carriers, additional excess charges

appear in the conduction band, resulting in a
sudden increase of current. With increasingly negative applied voltage, the I–V curve was well fit by
BC. Ballistic transport refers to carrier transport
without scattering. Accordingly, the conduction
mechanism changes from SCLC to BC and the
resistance switches from HRS to LRS. This analysis
suggests that the mechanism of the SET process
(HRS fi LRS) is due to the presence of scatteringfree regions in these switching devices. For example, the change from SCLC in the HRS to BC and
then Ohmic conduction in the LRS for both the Ag/
WO3/Pt and Ag/WO3/FTO devices can be ascribed to
resistive switching controlled by metallic filaments,
as described below.
For the Ti/WO3/FTO device, the electrical conduction of the LRS before the RESET (LRS fi
HRS) process follows ballistic conduction instead


ÀVmax fi 0
HRS

Fowler–Nordheim
tunneling conduction

HRS
Fowler–Nordheim
tunneling conduction
HRS
Fowler–Nordheim
tunneling conduction

LRS fi HRS
Ohmic conduction fi Fowler–Nordheim
tunneling conduction
LRS
LRS fi HRS
Ohmic conduction
Ohmic conduction fi Fowler–Nordheim
tunneling conduction
Sweeping process: 0 fi +Vmax fi 0
fi ÀVmax fi 0
+Vmax fi 0
0 fi ÀVmax
LRS
LRS fi HRS
Ballistic conduction
Ballistic conduction fi Fowler–Nordheim
tunneling conduction
LRS
Ohmic conduction

0 fi +Vmax
HRS fi LRS
Trap-controlled space-charge-limited
conduction fi ballistic conduction
Ti/WO3/FTO

Device

Ag/WO3/FTO

0 fi +Vmax
2Vmax fi 0
0 fi 2Vmax

HRS fi LRS
Trap-controlled space-charge-limited
conduction fi ballistic conduction
HRS fi LRS
Trap-controlled space-charge-limited
conduction fi ballistic conduction

On the basis of the nature of the electrode, the
direction of switching, and the analyzed electrical
conduction mechanisms, it seems that the mechanism of resistive switching in both the Ag/WO3/Pt
and Ag/WO3/FTO devices is controlled by electrochemical redox reactions.16,20 The resistive switching mechanism of the Ag/WO3/Pt (FTO) device is
modeled in Fig. 9 and explained as follows: On
application of a negative voltage to the Pt (FTO)

Ag/WO3/Pt

Resistance Switching Mechanism of Ag/WO3/
Pt and Ag/WO3/FTO Devices

Device

of Ohmic conduction as obtained for both the Ag/
WO3/Pt and Ag/WO3/FTO devices. Therefore, the
switching mechanism in the Ti/WO3/FTO device is
controlled by another factor, as also discussed
below.
After the RESET process, the electrical conduction in the HRS for all devices can be well described
by FN tunneling. These above-classified electrical
conduction mechanisms help us to understand the
resistive switching behavior of these switching
devices.

Sweeping Process: 0 fi 2Vmax fi 0 fi +Vmax fi 0

Fig. 6. Dominant electrical conduction mechanisms of Ag/WO3/Pt
device under (a) negative bias (SCLC fi BC fi OC) and (b, c)
positive bias (OC fi FN).

Table I. Dominant electrical conduction mechanisms in Ag/WO3/Pt, Ag/WO3/FTO, and Ti/WO3/FTO devices

+Vmax fi 0

Understanding Electrical Conduction States in WO3 Thin Films Applied for Resistive RandomAccess Memory


Ta, Pham, Dao, Tran, and Phan

Fig. 7. Dominant electrical conduction mechanisms of Ag/WO3/FTO
device under (a) negative bias (SCLC fi BC fi OC) and (b, c)
positive bias (OC fi FN).

bottom electrode (positive voltage to the Ag top
electrode), an electrochemical reaction occurs at the
anode (Ag), which oxidizes Ag metal atoms to Ag
ions. These Ag+ ions start from the top interface and
drift through the WO3 films to connect with the
bottom electrode. At the Pt (FTO) cathode, electrochemical reduction and electrocrystallization of Ag
occur. This electrocrystallization process results in
formation of an Ag filament, which grows toward
the Ag electrode. As a result, Ag filaments grow and
connect to the Ag top electrode, leading to HRS to
LRS switching. To RESET the cell, a positive
voltage is applied to the Pt bottom electrode (negative switching voltage to the Ag top electrode),
which leads to dissolution of the Ag filament, and
LRS to HRS switching occurs.
In our previous publication,16 XPS analysis
showed the presence of oxygen vacancies V2+
O in
+
ions have
WO3 thin films. As both V2+
O and Ag
positive charge, they will drift in the same direction
under the bias processes. Such oxygen vacancies
can affect the resistive switching of the WO3 thin
films. A postannealing process was carried out in air

at 600°C for 2 h to check the effect of these oxygen

Fig. 8. Dominant electrical conduction mechanisms of Ti/WO3/FTO
device under (a) positive bias (SCLC fi BC) and (b, c) negative
bias (BC fi FN).

vacancies; the results showed that both the postannealed Ag/WO3/Pt and Ag/WO3/FTO devices
retained the same switching behavior as the asgrown devices. Therefore, oxygen vacancies may be
involved in the electrical conduction but do not
make a major contribution, if any, to the switching
mechanism. In addition, the Ag filament was confirmed through the replacement of Ag by Ti as top
electrode. The Ti/WO3/Pt device exhibited no resistive switching behavior. It can again be concluded
that Ag filament paths mediated by electrochemical
redox reactions are responsible for resistive switching in the Ag/WO3/Pt and Ag/WO3/FTO devices.
Resistance Switching Mechanism
of the Ti/WO3/FTO Device
For the Ti/WO3/FTO device, the resistance
switching mechanism is modeled in Fig. 10 and
explained as follows: Resistance switching involves
back and forth drift of O2À ions through the bottom
interface. It is suggested that this occurs because
the polycrystalline phase of the FTO electrode acts
as a reservoir for defects such as V2+
O sites and,
simultaneously, O2À ions. The FTO electrode


Understanding Electrical Conduction States in WO3 Thin Films Applied for Resistive RandomAccess Memory

Fig. 9. Switching model for clockwise I–V hysteresis devices: Ag filament paths mediated by electrochemical redox reactions.

Fig. 10. Switching model for anticlockwise I–V hysteresis devices: negative O2À ions migrate under polarity biases.

contains enough vacancy sites near the interface to
readily provide resting sites for migrating O2À ions.
This implies that the FTO electrode can be treated
as an O2À ion source. The O2À ions move a short
distance from the bulk oxide films, through the
bottom interface, into the FTO bottom electrode
during the positive bias process. This movement of
O2À ions is associated with formation of oxygen
vacancies V2+
O in the bulk oxide films near the
bottom interface; consequently, the SET process
(HRS to LRS) occurs. During the negative bias
process, the O2À ions move back from the FTO
bottom electrode to the bulk oxide films, eliminating
oxygen vacancies V2+
O at the interface, resulting in
the RESET process (LRS to HRS).
It is also noted that the postannealed Ti/WO3/
FTO device did not show any switching behavior. It

can be concluded that oxygen ions are mainly
involved in the switching mechanism of the asgrown Ti/WO3/FTO device.
CONCLUSIONS
The electrical conduction and associated resistance switching mechanism of TE/WO3/BE devices
(TE: Ag, Ti; BE: Pt, FTO) were investigated. In the
switching devices, the direction of switching
depended on both the top and bottom electrode

material. It was observed that, for WO3 devices with
FTO bottom electrode, replacing the Ag top electrode with a Ti electrode caused a change of the
switching direction, while for WO3 devices with Pt
bottom electrode, replacing the Ag top electrode
with a Ti electrode caused suppression of switching.


Ta, Pham, Dao, Tran, and Phan

For devices with the same switching direction, i.e.,
Ag/WO3/Pt and Ag/WO3/FTO devices, the resistance
switching is due to Ag filament paths mediated by
electrochemical redox reactions. The governing
electrical conduction mechanisms are FN tunneling
for the HRS under positive bias, trap-controlled
SCLC for the HRS and ballistic and Ohmic conduction for the LRS under negative bias, and Ohmic
conduction before the RESET process under positive
bias.
For the device exhibiting the opposite switching
direction, i.e., Ti/WO3/FTO, the resistance switching
involves back and forth drift of oxygen ions through
the bottom interface. The governing electrical conduction mechanisms are FN tunneling for the HRS
under positive bias, trap-controlled SCLC for the
HRS, and ballistic conduction for the LRS under
both negative and positive biases.
Both the ballistic and Ohmic conduction mechanisms suggest that the SET process (HRS fi LRS)
occurs due to the presence of scattering-free regions
in the switching devices.
ACKNOWLEDGEMENTS
This work is financially supported by Vietnam

National University in Ho Chi Minh City under
Grant HS2015-18-02.
REFERENCES
1. S.H. Jo, Doctoral Thesis, The University of Michigan, 2010.
2. S.Q. Liu, N.J. Wii, and A. Ignatiev, Appl. Phys. Lett. 76,
2749 (2000).
3. Y. Watanabe, J.G. Bednorz, A. Bietsch, Ch Gerber, D.
Widmer, A. Beck, and S.J. Wind, Appl. Phys. Lett. 78, 3738
(2001).
4. R. Waser and M. Aono, Nat. Mater. 6, 833 (2007).
5. B.T. Phan and J. Lee, Appl. Phys. Lett. 93, 222906 (2008).
6. B.T. Phan and J. Lee, Appl. Phys. Lett. 94, 232102 (2009).
7. B.T. Phan, N.C. Kim, and J. Lee, J. Korean Phys. Soc. 54,
873 (2009).

8. B.T. Phan, T. Choi, A. Romanenko, and J. Lee, Solid-State
Electron. 75, 43 (2012).
9. Y.C. Chen, C.F. Chen, C.T. Chen, J.Y. Yu, S. Wu, S.L. Lung,
R. Liu, and C.Y. Lu, IEDM Tech. Dig. (2003), pp. 905–908.
10. B.J. Choi, D.S. Jeong, S.K. Kim, S. Choi, J.H. Oh, C. Rohde,
H.J. Kim, C.S. Hwang, K. Szot, R. Waser, B. Reichenberg,
and S. Tiedke, J. Appl. Phys. 98, 033715 (2005).
11. K. Jung, H. Seo, Y. Kim, H. Im, J.P. Hong, J.W. Park, and
J.K. Lee, Appl. Phys. Lett. 90, 052104 (2007).
12. A. Chen, S. Haddad, Y.C. Wu, Z. Lan, T.N. Fang, and S.
Kaza, Appl. Phys. Lett. 91, 123517 (2007).
13. C.Y. Lin, C.Y. Wu, C. Hu, and T.Y. Tseng, J. Electrochem.
Soc. 154, G189 (2007).
14. T. Le, H.C.S. Tran, V.H. Le, T. Tran, C.V. Tran, T.T. Vo,
M.C. Dang, S.S. Kim, J. Lee, and B.T. Phan, J. Korean

Phys. Soc. 60, 1087 (2012).
15. N.K. Pham, D.T. Nguyen, B.T.T. Dao, K.H.T. Ta, V.C.
Tran, V.H. Nguyen, S.S. Kim, S. Maenosono, and B.T.
Phan, J. Electron. Mater. 43, 2747 (2014).
16. T.B.T. Dao, K.N. Pham, Y.L. Cheng, S.S. Kim, and B.T.
Phan, Curr. Appl. Phys. 14, 1707 (2014).
17. K.N. Pham, M.S. Choi, C.V. Tran, T.D. Nguyen, V.H. Le, T.
Choi, J. Lee, and B.T. Phan, J. Electron. Mater. 44, 3395
(2015).
18. J.B. Park, K.P. Biju, S.J. Jung, W.T. Lee, J.M. Lee, S.H.
Kim, S.S. Park, J.H. Shin, and H.S. Hwang, IEEE Electron
Device Lett. 32, 476 (2011).
19. J.W. Seo, Doctoral Thesis, Korean Advanced Institute of
Science and Technology, 2011.
20. D.S. Jeong, R. Thomas, R. Katiyar, J.F. Scott, H. Kohlstedt,
A. Petraru, and C.S. Hwang, Rep. Prog. Phys. 75, 076502
(2012).
21. Y.E. Syu, T.C. Chang, T.M. Tsa, G.W. Chang, K.C. Chang,
Y.H. Tai, M.J. Tsai, Y.L. Wang, and S.M. Sze, Appl. Phys.
Lett. 100, 022904 (2012).
22. B.U. Jang, A.I. Inamdar, J.M. Kim, W. Jung, H.S. Im, H.S.
Kim, and J.P. Hong, Thin Solid Films 520, 5451 (2012).
23. D.S. Hong, Y.S. Chen, Y. Li, H.W. Yang, L.L. Wei, B.G.
Shen, and J.R. Sun, Sci. Rep. 4, 4058 (2014).
24. J. Wua and J. Wang, J. Appl. Phys. 108, 034102 (2010).
25. W.Y. Yang and S.W. Rhee, Appl. Phys. Lett. 91, 232907
(2007).
26. N.F. Mott and R.W. Gurney, Electronic Processes in Ionic
Crystals (Oxford: Clarendon, 1940).
27. S.M. Sze, Physics of Semiconductor Devices, 2nd ed., Vol. 1

(New York: Wiley, 1981), pp. 72–75.
28. C. Charles, N. Martin, M. Devel, J. Ollitrault, and A. Billard, Thin Solid Films 534, 275 (2013).