Research Article

www.acsami.org

Manipulated Transformation of Filamentary and Homogeneous

Resistive Switching on ZnO Thin Film Memristor with Controllable
Multistate
Chi-Hsin Huang,†,§ Jian-Shiou Huang,†,§ Chih-Chung Lai,† Hsin-Wei Huang,† Su-Jien Lin,†
and Yu-Lun Chueh*,†
†
Department of Materials Science and Engineering, Center For Nanotechnology, Material Science, and Microsystem, National Tsing
Hua University, Hsinchu, Taiwan, Republic of China
*
S Supporting Information
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ABSTRACT: A bias polarity-manipulated transformation

Downloaded via CHUNGNAM NATL UNIV on May 22, 2020 at 08:59:32 (UTC).

from filamentary to homogeneous resistive switching was

demonstrated on a Pt/ZnO thin film/Pt device. Two types of
switching behaviors, exhibiting different resistive switching
characteristics and memory performances were investigated in
detail. The detailed transformation mechanisms are systemati-
cally proposed. By controlling different compliance currents
and RESET-stop voltages, controllable multistate resistances in
low resistance states and a high resistance states in the ZnO
thin film metal−insulator−metal structure under the homoge-
neous resistive switching were demonstrated. We believe that
findings would open up opportunities to explore the resistive switching mechanisms and performance memristor with multistate
storage.
KEYWORDS: resistive switching, ReRAM, ZnO, memristor, nonvolatile memory, multistate storage

1. INTRODUCTION addition to filamentary resistive switching, homogeneous

Resistive switching random access memory (ReRAM),
consisting of the simplest device configuration, namely
metal−insulator−metal (MIM), is the most promising
candidate as next generation nonvolatile memory. It has a
higher potential to replace current flash memory and dynamic
random access memory (DRAM) due to its faster switching
speed, higher stacking density, lower power consumption,
higher scalability, faster fabrication process, and stronger
potential for fabricating multistate memories.1−3 After
intensively investigating the switching mechanisms of ReRAM
in recent years, the migration of oxygen vacancies (or oxygen
ions) under an applied electrical field plays an important role in
the metal oxide materials, for which the resistive switching
behaviors are highly determined by the accumulation of oxygen
vacancies (or oxygen ions) either at the interface or in the bulk
of materials.1−4
Typically, two types of resistive switching mechanisms have
been proposed, namely, filamentary and homogeneous switch-
ing behaviors. The filamentary resistive switching, which is
described as fuse−antifuse switching, basically involves the
formation and rupture of conducting filaments by thermo-
chemical reactions inside the insulating metal oxide layer.3 The
electrochemical metallization mechanism (ECM) via formation
of metallic filaments by the accumulation of active metal
electrodes, such as Ag and Cu in the electrochemical
dissolution, is another case for filamentary switching.3,5 In
interface resistive switching, behaving as continuous switching,
is attributed to an asymmetric balance of Schottky-barrier
heights induced by the electric field via trapping and detrapping
processes of carriers through defective states at interfaces
between the functional material and electrode.6,7 The switching
current via homogeneous switching highly depends on the
device area, which guarantees a sufficient current to maintain
reliable operation. However, switching current localized by one
or several conducting filaments, namely filamentary resistive
switching, may result in the fluctuation of resistance states and a
larger current density, thus unstable operation in the scaled
device once the area of the device becomes smaller.
Some materials, such as WOx,8 Fe-doped SrTiO3,9,10 and
DyMn2O5,11 have been found that the coexistence of
filamentary and homogeneous resistive switching has been
illustrated, while how to control the transformation between
filamentary and homogeneous resistive switching has not been
reported yet. In addition, the transition from bipolar to unipolar
resistive switching relying on current compliance by an
electroforming process has been observed in TiO2,12 while
filamentary and homogeneous resistive switching exhibiting
different switching characteristics and device performances have
Received: February 26, 2013
Accepted: May 24, 2013
Published: May 24, 2013

© 2013 American Chemical Society 6017 dx.doi.org/10.1021/am4007287 | ACS Appl. Mater. Interfaces 2013, 5, 6017−6023
ACS Applied Materials & Interfaces
Figure 1. (a) Schematic of a Pt/ZnO thin film/Pt resistive switching device. (b) I−V characteristics of a filamentary resistive switching for the Pt/
ZnO/Pt devices. (c) I−V characteristics of the transformation process. (d) I−V characteristics of the homogeneous resistive switching for the Pt/
ZnO thin film/Pt devices.
not yet been investigated systematically. Therefore, it is
important to investigate the relationship of two switching
mechanisms and the feasibility of controlling the trans-
formation between filamentary and homogeneous resistive
switching for further development of ReRAM.
In this regard, we demonstrate a controllable transformation
of filamentary to homogeneous interface resistive switching in a
ZnO thin film MIM structure via different bias polarities. These
two types of resistive switching behaviors, which exhibit
different resistive switching characteristics and memory
performances, were investigated in detail. The possible
transformation mechanisms are systematically proposed by
microstructural and chemical analyses obtained from trans-
mission electron microscopy (TEM). By controlling different
compliance currents (CC) and RESET-stop voltages, we are
able to demonstrate a controllable multilevel storage at a high
resistance state (HRS) and a low resistance state (LRS) in the
ZnO thin film MIM structure under homogeneous resistive
switching for high density with multistorage application.
2. EXPERIMENTAL SECTION
Materials and Methods. A 100-nm-thick ZnO thin film was
deposited on a Pt/Ti/SiO2/Si substrate at room temperature by radio
frequency (rf) magnetron sputtering using a ceramic ZnO target. The
100-nm-thick Pt top electrode was deposited on the ZnO thin film
patterned by a shadow mask via rf-magnetron sputtering at room
temperature. The diameters of the top electrodes from 200 to 50 μm
were also defined by metal shadow masks.
Characterization. Field-emission transmission electron micros-
copy (JEM-3000F, JEOL operated at 300 kV with a point-to-point
resolution of 0.17 nm) equipped with an energy dispersion
spectrometer (EDS) was used to obtain the microstructures and the
chemical compositions. The crystallinity of the ZnO film was
confirmed by X-ray diffraction (Shimadzu XRD 6000, Cu Kα radiation
with a wavelength of 0.154 nm). The electrical characterizations and
resistive switching characteristics were measured using a Keithley 4200
semiconductor parameter analyzer in voltage sweeping mode at room
6018
Research Article
temperature. All of the operation voltages were applied on the Pt top
electrode, with which the Pt bottom electrode was grounded.
3. RESULTS AND DISCUSSION
Figure 1a shows a schematic of the ZnO thin film MIM
structure, for which the stacking sequences of the entire
structure are Pt (100 nm)/ZnO (100 nm)/Pt (100 nm). The
X-ray spectrum confirms a wurtzite structure of the ZnO thin
film with a good crystallinity, for which a single peak at the
(002) plane could be obviously observed (Figure S1). By
applying a positive bias from 0 V to ∼2.8 V (forming voltage)
with a compliance current of 10 mA, an abrupt increase of
current could be observed, namely the “forming process,” for
which the resistance of the device is changed from the initial
resistance state to LRS owing to the formation of conducting
filaments in the ZnO thin film as shown in Figure 1b. When we
sweep the biases from 0 V to ∼0.7 V (RESET voltage, VRESET),
the resistance of the device changing to HRS was observed,
which is called the “RESET process” (Figure 1b). A change of
resistance from HRS to LRS, namely the “SET process,” could
be achieved after a positive bias > 1.5 V (SET voltage, VSET)
was applied (Figure 1b). The resistive switching in the ZnO
film is called unipolar switching, which is a kind of filamentary
resistive switching. To trigger the transformation of filamentary
to homogeneous resistive switching, a reverse sweep bias of
∼−1.75 V with a relatively high CC (∼ 100 mA) was applied,
with which an abrupt current change from HRS to LRS could
be found as shown in Figure 1c. Note that due to the higher
CC, a higher current/lower LRS could be observed, as shown
in Figure 1c. Subsequently, the bias sweep at a positive bias was
applied from 0 V to ∼2 V to conduct the RESET behavior,
resulting in the device in the HRS. Finally, SET and RESET
biases were applied from 0 to −2 V and 0 to 2 V with a CC of
100 mA, respectively, for which the transformation of the
filamentary to the homogeneous switching is accomplished, as
dx.doi.org/10.1021/am4007287 | ACS Appl. Mater. Interfaces 2013, 5, 6017−6023
ACS Applied Materials & Interfaces Research Article

Figure 2. (a) Cumulative distribution of the HRS/LRS for the Pt/ZnO/Pt devices. (b) Retention tests of the filamentary and homogeneous resistive
switching under a continuous read voltage of 0.1 V. (c) Currents as the function of device sizes at HRS/LRS of the filamentary and homogeneous
resistive switching. The error bar at each cell size is calculated by the standard deviation of 20 devices. (d) Temperature dependences at HRS/LRS
for filamentary and homogeneous resistive switching. (e) ln (I)−ln (V) plots for homogeneous resistive switching. (f) ln (J/E) − E1/2 and inset J−E
plots for filamentary resistive switching.

shown in Figure 1d. In addition, an opposite operation can be
obtained if we switch the bias polarity, confirming the stable
and controllable process, as shown in Figure S2.
The different resistive switching modes (“fuse−antifuse
switching by conducting filaments” versus “continuous switch-
ing by trapping−detrapping behaviors via the migration of
defects/ions”)2,3 for both switching behaviors could be
achieved. The fuse−antifuse switching is a typical feature of
filamentary switching, which usually manifests in the so-called
unipolar resistive switching mode, for which the switching of
LRS and HRS occurs owing to the formation and rupture of
local metallic paths by Joule heating at the same bias polarity
while it can be also operated at any polarity, namely a nonpolar
resistive switching mode.3 In contrast, homogeneous resistive
switching involves continuous switching by the trapping−
detrapping effect via intrinsic defects at the metal/insulator
interface or the migrated ions/defects to the interface induced
by the electrical field, belonging to bipolar switching.6,7
Notably, the abrupt current increases in the SET process and
sudden decreases in the RESET process are a particular feature
of filamentary resistive switching (Figure 1b). After the
transformation from filamentary to homogeneous resistive
switching, a gradually continuous increase and decrease in
current from the current−voltage (I−V) curve could be found
during the SET and RESET processes (Figure 1d). In addition,
6019
a memristive effect, namely different hysteresis loops due to
changes of Schottky-barrier depletion width via the charge
accumulation and trapping/detrapping processes could be
obviously obtained by different positive/negative biases and
sweep cycles as shown in Figure S3, evidently indicating the
replacement of the filamentary resistive switching by homoge-
neous resistive switching.13
Figure 2a shows cumulative probability distributions of
resistances at HRS/LRS by filamentary and homogeneous
resistive switching. The corresponding cumulative probability
distributions of resistances at VSET/VRESET voltages are shown
in Figure S3. The statistic distributions of the resistances (HRS
and LRS) and switching threshold voltages (VSET and VRESET)
of 100 cycles in the homogeneous resistive switching are nearly
vertical with a slight fluctuation, confirming that the switching
behavior of homogeneous resistive switching is very uniform.
The nonuniform performances for filamentary resistive switch-
ing mainly result from variable locations, orientations, and sizes
of conducting filaments during SET and RESET processes. The
retention test was conducted as shown in Figure 2b to evaluate
an ability of nonvolatile data storage. Notably, filamentary
resistive switching shows good retention characteristics up to
104 s without any degradation in both HRS and LRS. The
electrical mechanisms involving the trapping and detrapping of
carriers within the interface states for homogeneous switching
dx.doi.org/10.1021/am4007287 | ACS Appl. Mater. Interfaces 2013, 5, 6017−6023
ACS Applied Materials & Interfaces
Figure 3. Schematics of transformation mechanisms of filamentary into homogeneous resistive switching.
face a problem of intrinsically low retention time.14,15 However,
the homogeneous resistive switching still maintains a
reasonable retention characteristic of ∼103 s in our device.
To shed light on filamentary and homogeneous resistive
switching, the current at HRS and LRS with different electrode
sizes was measured, as shown in Figure 2c. Current at HRS and
LRS in the homogeneous resistive switching increases as
electrode sizes increase, namely slightly linear dependence,
indicating that the current is homogeneous throughout the
device and the switching occurs on the entire thin film. Notice
that the lesser decrease of current trend as the device area
decreases is probably due to a lesser resistivity (ρ) modulation
at the bigger cell size during the homogeneous operation. The
smaller electrodes contain higher resistivity due to a lesser
amount of grain boundaries and leakage current. Moreover, the
polarity-dependent migration via oxygen vacancies induces
oxygen-deficient/oxygen-rich region modulation, resulting in
different resistivities. However, the current at the HRS in
filamentary resistive switching is still dependent on the device
sizes while the current at the LRS was found to be independent
of the device sizes resulting from the formation of localized
conducting filaments.2 Figure 2d shows temperature depend-
ence studies at changes of resistances from HRS and LRS in
filamentary and homogeneous resistive switching. The
resistances at HRS in both types decrease with an increase in
temperature, which originates from an intrinsic behavior of the
ZnO thin film, indicating a semiconducting feature. Interest-
ingly, the resistance at LRS increases with an increase in
temperature, indicating a metallic characteristic of filamentary
6020
Research Article
resistive switching while the resistance at LRS in homogeneous
resistive switching exhibits an opposite trend, namely a decrease
of resistance with an increase of temperature, again confirming
the semiconducting behavior. An increase of the resistance at
LRS in the homogeneous resistive switching beyond 360 K is
most likely from the metallic filaments embedded in the
semiconducting ZnO film during the initial unipolar operation.
It is because such behavior could be related to either the
diffusion of oxygen vacancies in a thin film or the trapped
electronic charges at the metal/semiconductor interface or both
of them. Figure 2e shows the double-logarithmic plots of the I−
V curves for negative voltage regions of homogeneous resistive
switching. The log I−V plot at HRS shows an Ohmic
conduction behavior, namely a linear proportion for current
to voltage at a low voltage region (<2.5 V) gradually changes to
the Child’s law region, for which the current is proportional to
the square of the voltage.16 This behavior is qualitatively
interpreted to follow the shallow trap-associated SCLC theory
with a simple equation of I(V) = aV + bV2.16 The trap-
controlled SCLC describes the charge conduction via charge
trapping and detrapping, which is consistent with previous
reports on the Schottky effect, where the migration of oxygen
vacancies in the vicinity of the interface drives resistive
switching. Under a positive electric field, oxygen vacancies
migrate away from the Schottky interface between ZnO and Pt,
which changes the Schottky-barrier widths and widens the
depletion layer, resulting in HRS and vice versa. On the other
hand, Poole−Frenkel emission was found to follow typical I−V
curves of the filamentary resistive switching in the HRS.17 A
dx.doi.org/10.1021/am4007287 | ACS Appl. Mater. Interfaces 2013, 5, 6017−6023
ACS Applied Materials & Interfaces Research Article

Figure 4. (a) A TEM image of the pristine device. (b) The corresponding EDS elemental profiles. (c) A TEM image of the device after
transformation of filamentary into homogeneous resistive switching. (d) The corresponding EDS elemental profiles.

Figure 5. (a) I−V characteristics of homogeneous resistive switching when varying the compliance currents (CC). (b) Endurance tests of 100 cycles
for the homogeneous resistive switching when varying the CC at a read bias of 0.1 V. (c) I−V characteristics of the homogeneous resistive switching
when varying the RESET-stop voltage. (d) Endurance tests of 100 cycles for the homogeneous resistive switching when varying the RESET-stop
voltage at a read bias of 0.1 V.

linear curve with a slope of ∼1 was observed in the LRS, which
suggests Ohmic conduction, indicating the existence of
conducting filaments as shown in inset of Figure 2f. (More
detailed modeling of transport mechanisms from the fitting of
I−V curves is discussed in the Supporting Information file.)
Therefore, detailed mechanisms on how the transformation
of filamentary to homogeneous resistive switching behaviors
with controllable biases and polarities were proposed as shown
in Figure 3a−f. By applying a large voltage, namely the forming
process, the oxygen vacancies are migrated and gathered
6021
together to generate conducting filaments located inside the
ZnO film to result in the device from HRS to LRS (Figure
3a).18 Subsequently, the rupture of filaments via Joule heating
due to the high current density occurred, turning the device
from LRS to HRS in the RESET process (Figure 3b).18,19
Notice that the positions for the rupture of the filaments would
normally occur near the top electrode because of much finer
filament sizes.19 When we sweep a large enough negative bias at
the top electrode, an abrupt change in current from HRS to
LRS could be observed owing to the formation of filament
dx.doi.org/10.1021/am4007287 | ACS Appl. Mater. Interfaces 2013, 5, 6017−6023
ACS Applied Materials & Interfaces
paths near the top electrode caused by the migration of oxygen
vacancies while the compensation of filament paths near the
bottom electrode caused by the migration of oxygen ions can
be achieved, resulting in the formation of the ZnO1−x (ZnO
layer with oxygen-deficient region) layer as shown in Figure 3c.
A transition of filamentary to homogeneous resistive switching
can be triggered at this stage but is not completed yet until a
further positive bias is applied to finish the homogeneous
switching. When the positive bias was applied on the top
electrode, the O2− would be attracted toward the top electrode
from the ZnO1−x layer, resulting in the formation of an oxygen-
rich (ZnO) layer near the top electrode, leading to HRS.
Consequently, the transformation of the filamentary to the
homogeneous resistance switching could be completed (Figure
3d). The following device switching from HRS to LRS is caused
by trapping and detrapping processes owing to the asymmetric
Schottky-barrier height resulting from stoichiometric modu-
lation of the oxygen-rich layer (e.g., oxygen ion/vacancy
electromigration in Figure 3e and f).
To corroborate the existence of the oxygen-defective region
nearby the bottom Pt electrode, the energy dispersive spectrum
(EDS) was used to obtain the compositional difference in two
kinds of samples, namely a pristine sample without any
operation and a sample after the transformation of filamentary
to homogeneous resistive switching, as shown in Figure 4a and
b, respectively. The uniform atomic concentrations of ∼51
atom % Zn and ∼49 atom % O through the entire ZnO thin
film for the pristine sample could be observed (Figures 4a)
while a Zn rich region (∼25−30-nm-thick), namely the oxygen-
defective region with atomic concentrations of ∼60 atom % Zn
and ∼40 atom % O could be detected (Figure 4b), distinctly
confirming the existence of the oxygen-deficient regions nearby
the bottom Pt electrode.
Furthermore, by controlling the different compliance
currents (CC) and RESET-stop voltages, we are able to
enlarge or deplete the oxygen-defective region, leading to the
multistate resistances at LRS and HRS under homogeneous
resistive switching, respectively. Figure 5a shows the I−V
characteristics of SET processes after a constant SET bias of −3
V was applied with different CCs of 0.04, 0.05, and 0.075 A,
respectively, with which a RESET bias of 3 V was applied for
each measurement. It could be clearly seen that varying the
compliance currents could result in different resistances at LRS.
The higher the CC that was applied, the smaller the resistance
at LRS that was obtained, while the resistances at HRS are
almost identical (Figure S5a). An endurance test of ∼100 cycles
as shown in Figure 5b indicates that resistive switching for each
memory state is very stable. On the other hand, by controlling
the RESET-stop voltages of 2, 2.5, and 2.7 V, the I−V
characteristics with different resistances at HRS could be
achieved as shown in Figure 5c, for which a constant SET
voltage of 3 V with a CC of 10 mA was applied for each
measurement. Again, the higher resistance at HRS was achieved
as the larger RESET-stop voltage was applied (Figure S5b). We
found that resistive switching is also very stable at each memory
state with an endurance test of ∼100 cycles as shown in Figure
5d. Most importantly, all resistances at HRS and LRS are
reversible, depending on the different CC and RESET-stop
voltages.
4. CONCLUSIONS
We present a method for the transformation of the filamentary
resistive switching into the homogeneous resistive switching for
6022
Research Article
the Pt/ZnO thin films/Pt MIM structure memory device. We
believe the critical role of the transformation between
filamentary and homogeneous resistive switching highly relies
on the distribution of oxygen vacancies. Detailed trans-
formation mechanisms are systematically proposed. By
controlling different compliance currents and RESET-stop
voltages, we are able to demonstrate controllable multistate
resistances at LRS and HRS in the ZnO thin film MIM
structure under homogeneous resistive switching. Our method
could open up an opportunity to explore not only the resistive
switch mechanism but also develop next generation multistate
high performance memory.
■
*
ASSOCIATED CONTENT
S Supporting Information
XRD spectrum of ZnO film; I−V characteristics of the
memristive effects; transformation of filamentary to homoge-
neous switching under an opposite operation; transport
mechanism for filamentary and homogeneous resistive switch-
ing; demonstration of multistate resistances at LRS and HRS
under homogeneous resistive switching. These materials are
available free of charge via the Internet http://pubs.acs.org
■
AUTHOR INFORMATION
Corresponding Author
*E-mail: ylchueh@mx.nthu.edu.tw.
Author Contributions
§
These two authors contributed equally to this work.
Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS
The research was supported by the National Science Council
through grant nos. NSC 101-2112-M-007-015-MY3 and NSC
100-2120-M-007-008 and National Tsing Hua University
through grant no. 102N2022E1. Y.L.C. greatly appreciates the
use of the facility at CNMM, the National Tsing Hua
University through grant no. 101N2744E1.
■ REFERENCES
(1) Waser, R.; Aono, M. Nat. Mater. 2007, 6, 833−840.
(2) Sawa, A. Mater. Today 2008, 11, 28−36.
(3) Waser, R.; Dittmann, R.; Staikov, G.; Szot, K. Adv. Mater. 2009,
21, 2632−2663.
(4) Huang, C. H.; Huang, J. S.; Lin, S. M.; Chang, W. Y.; He, J. H.;
Chueh, Y. L. ACS Nano 2012, 6, 8407−8414.
(5) Yang, Y. C.; Pan, F.; Liu, Q.; Liu, M.; Zeng, F. Nano Lett. 2009, 9,
1636−1643.
(6) Sawa, A.; Fujii, T.; Kawasaki, M.; Tokura, Y. Appl. Phys. Lett.
2004, 85, 4073−4075.
(7) Yang, J. J.; Pickett, M. D.; Li, X.; Ohlberg, D. A. A.; Stewart, D.
R.; Williams, R. S. Nat. Nanotechnol. 2008, 3, 429−433.
(8) Biju, K. P.; Liu, X.; Kim, S.; Jung, S.; Park, J.; Hwang, H. Phys.
Status Solidi RRL 2011, 5, 89−91.
(9) Shibuya, K.; Dittmann, R.; Mi, S.; Waser, R. Adv. Mater. 2010, 22,
411−414.
(10) Muenstermann, R.; Menke, T.; Dittmann, R.; Waser, R. Adv.
Mater. 2010, 22, 4819−4822.
(11) Tsai, Y. T.; Chang, T. C.; Huang, W. L.; Huang, C. W.; Syu, Y.
E.; Chen, S. C.; Sze, S. M.; Tsai, M. J.; Tseng, T. Y. Appl. Phys. Lett.
2011, 99, 092106−3.
(12) Jeong, D. S.; Schroeder, H.; Waser, R. Electrochem. Solid-State
Lett. 2007, 10, G51−G53.
dx.doi.org/10.1021/am4007287 | ACS Appl. Mater. Interfaces 2013, 5, 6017−6023
ACS Applied Materials & Interfaces Research Article

(13) Strukov, D. B.; Snider, G. S.; Stewart, D. R.; Williams, R. S.

Nature 2008, 453, 80−83.
(14) Peng, H. Y.; Li, G. P.; Ye, J. Y.; Wei, Z. P.; Zhang, Z.; Wang, D.
D.; Xing, G. Z.; Wu, T. Appl. Phys. Lett. 2010, 96, 192113−3.
(15) Schroeder, H.; Zhirnov, V. V.; Cavin, R. K.; Waser, R. J. Appl.
Phys. 2010, 107, 054517−8.
(16) Lampert, M. A. Phys. Rev. 1956, 103, 1648−1656.
(17) Yeargan, J. R.; Taylor, H. L. J. Appl. Phys. 1968, 39, 5600−5604.
(18) Kim, M. K.; Jeong, D. S..; Hwang, C. S. Nanotechnol. 2011, 22,
254002−17.
(19) Chang, S. H.; Lee, J. S.; Chae, S. C.; Lee, S. B.; Liu, C.; Kahng,
B.; Kim, D. W.; Noh, T. W. Phys. Rev. Lett. 2009, 102, 026801−4.

6023 dx.doi.org/10.1021/am4007287 | ACS Appl. Mater. Interfaces 2013, 5, 6017−6023

 Supplementary Information

Manipulated transformation of filamentary and homogeneous resistive switching on

ZnO thin film memristor with controllable multistate

Chi-Hsin Huang1, 2†, Jian-Shiou Huang1, 2†, Chih-Chung Lai1, 2, Hsin-Wei Huang1, 2, Su-Jien
Lin1, and Yu-Lun Chueh1, 2*
1
Department of Materials Science and Engineering, Center For Nanotechnology, Material
Science, and Microsystem, National Tsing Hua University, Hsinchu , Taiwan, ROC
2
Center For Nanotechnology, Material Science, and Microsystem, National Tsing Hua
University, Hsinchu 30013, Taiwan, R.O.C.

†
These two authors contribute equally in this work.
*
Corresponding author: ylchueh@mx.nthu.edu.tw
(002)

ZnO
Intensity (a.u.)

20 30 40 50 60
2θ (degree)
Figure S1 XRD spectra of the ZnO thin film.

1
 (a) -1
(b) -1
10 10
-2
RESET
10 -2
10
-3
10 SET

Current (A)

Current (A)
-4
-3
10 SET
10
-5
Forming
10 -4
10
-6
10
-5
-7 10
10 Fila.-RS
-8 -6
10 10
-4 -3 -2 -1 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Voltage (V) Voltage (V)
(c)
-1
10

10
-2
RESET SET
Current (A)
10
-3

-4
10
-5
10
-6 Homo.-RS
10
-3 -2 -1 0 1 2 3
Voltage (V)

Figure S2 I-V curves of the transformation from filamentary to homogeneous RS with a negative
forming process.

ffect
I–V characteristics of memristive eff

A memristive effect, namely different hysteresis loops, due to changes of Schottky-barrier

heights via accumulation of charges by trapping and detrapping processes could be obviously obtained

at different positive/negative biases with different sweep cycles as shown in Figure S3. It evidently

indicates the transformation of the filamentary resistive switching by the homogeneous resistive

switching. In Figure S2 (a), hysteresis loops of the I-V curves could be obtained by several minimum

voltages (0→Vmin→0), switching the device from HRS to LRS and a constant reset voltage to retain

the same HRS for every loop. Figure S2 (b) presents the I-V curves with several maximum voltages,

namely, 0→Vmax→0, switching LRS to HRS and a constant set voltage to retain the same LRS for

every loop. All the results obviously indicate the formation of hysteresis loops when voltages reach the

threshold values. As shown in Figures S2(c) and S2(d), when the four positive/negative sweeps

( 0→Vmin / Vmax→0 ) are continuously and gradually applied, the increasing (decreasing) conductivity

during the positive negative (positive) voltage sweeps could be observed, respectively.

2
 (a) (b)
-1 -1
10 10

Current (A)

Current (A)
-2 -2
10 10
Vmin = -2.5 V
Vmax = 3 V
-3
Vmin = -1.5 V -3
10 10 Vmax = 2 V
Vmin = -1.25 V
Vmax = 1.7 V
-4 -4
10 10
-3 -2 -1 0 0 1 2 3
Voltage (V) Voltage (V)
(c) (d)
-1 -1
10 10

Current (A)
Current (A)

-2 -2
10 10 1st sweep
1st sweep
2nd sweep 2nd sweep
-3
10 3rd sweep
-3
10 3rd sweep
4th sweep 4th sweep
-4 -4
10 10
-2 -1 0 0 1 2 3
Voltage (V) Voltage (V)

Figure S3 I-V characteristics of the homogeneous resistive switching at (a) different Vmin and (b) Vmax

voltages, respectively. (c) and (d) I-V characteristics of the memristive effect at continuous positive

and negative bias voltages, respectively.

Cumulative Probility (%)

100

80
Homo. Vset
60 Homo. Vreset
40 Fila. Vset
Fila. Vreset
20

-1 0 1 2
Voltage (V)

Figure S4 Cumulative distribution of the set/reset voltages for the Pt/ZnO/Pt devices.

3
Transport mechanisms for filamentary and homogeneous resistive switching

To investigate the different origins of these two resistive switching behaviors, the current

conduction mechanism of HRS and LRS are investigated in detail. The classical nonlinear leakage

conduction mechanisms, including Schottky emission1, Poole-Frenkel (PF) emission2, and

space-charge-limited-current (SCLC)3 are adopted to fit the nonlinearity of the measured J-E relation.

Figure 2(e) shows a double-logarithmic plot of the I-V curves for the negative voltage region of

homogeneous resistive switching. The log I-V plot at HRS shows an ohmic conduction behavior,

namely the linear proportion of current to voltage at the low voltage region (<2.5 V), gradually

changes to the Child’s law region, for which the current is proportional to square of the voltage.3 This

behavior is qualitatively interpreted to follow the shallow trap-associated SCLC theory with a simple

equation of I(V) = aV + bV2.3 The trap-controlled SCLC describes the charge conduction via the

charge trapping and detrapping, which is consistent with previous reports on the Schottky effect that

the migration of oxygen vacancies in the vicinity of the interface drives resistive switching. Under a

positive electric field, oxygen vacancies migrate away from the Schottky interface between ZnO and

Pt, which changes the Schottky-barrier height and widens the depletion layer, resulting in HRS and

vice versa.

On the other hand, I-V curve of the filamentary resistive switching was found to follow

Poole-Frenkel (PF) emission, of which the J-E relation can be given,2

 q 3 E / πε ε 
J ∝ E exp r 0 
 rk BT 
 

where q is the electric charge, ɛr is the dynamic dielectric constant, ɛo is the vacuum permittivity, kB

is the Boltzmann’s constant, T is the temperature, and r is the coefficient ranged between 1 and 2.4 The

conduction mechanism is called the normal PF effect as r=1. However, the conduction is called the

modified PF emission as r=2 due to containing non-negligible traps. Therefore, the calculated slope of

the curve from ln(J/E) vs. E1/2 could be obtained to be ~14.91 with the r = 1 as shown in Figure 2(f).

From the slope of the linear ln(J/E) vs. E1/2 relation, the refractive index n ( n = ε r ) can be

evaluated. The refractive index obtained from the fitting of the PF emission is 1.97, which is

consistent with the value reported in the literature.4 The result confirms that for the filamentary
4
resistive switching in the HRS, electrons transport through trap sites via the PF emission. In our case,

we believe the Schottky and P-F emission should be co-existed together while P-F emission dominates

the emission process. In the LRS, a linear curve with a slope of ~1 was observed, which suggests the

Ohmic conduction owing to existence of conducting filaments in LRS as shown in inset of Figure 2(f).

Demonstration of multistate resistances at LRS and HRS under the homogeneous resistive
switching

(a) (b)
80 350
LRS HRS
Resistance (Ω)

Resistance (Ω)
300
60 250

200
40 150

100
20 50
0.03 0.04 0.05 0.06 0.07 0.08 1.8 2.0 2.2 2.4 2.6 2.8 3.0
Compliance Current (A) Reset Voltage (V)
Figure S5 (a) Decreasing of resistances with increasing of compliance currents at LRS (b) Increasing
of resistances with increasing of reset-stop voltages at HRS.

REFERENCES
1. P. R. Emtage and W. Tantraporn, Physical Review Letters, 1962, 8, 267-268.
2. J. R. Yeargan and H. L. Taylor, Journal of Applied Physics, 1968, 39, 5600-5604.
3. M. A. Lampert, Physical Review, 1956, 103, 1648-1656.
4. W. Y. Chang, Y. C. Lai, T. B. Wu, S. F. Wang, F. Chen and M. J. Tsai, Applied Physics
Letters, 2008, 92, 022110-022113.