Operation mechanism of Schottky barrier nonvolatile memory with high conductivity InGaZnO active layer

Article abstract of DOI:10.1063/1.3699221

Influence of Schottky contact between source/drain electrodes and high conductivity a-InGaZnO active layer to the performance of nonvolatile memory devices was first proposed. The Schottky barrier devices faced to the difficulty on electrical discharging process due to the energy barrier forming at the interface, which can be resolved by using Ohmic devices. A memory window of 2.83V at programming/erasing voltage of ± 13V for Ohmic and 5.58V at programming voltage of 13V and light assisted erasing at -7 V for Schottky devices was obtained. Both memory devices using SiO₂/SiO _x/SiO_xN_y stacks showed a retention exceeding 70% of trapped charges 10 yr with operation voltages of ±13V at an only programming duration of 1ms.

Full text of DOI:10.1063/1.3699221

Operation mechanism of Schottky barrier nonvolatile memory with high conductivity
InGaZnO active layer
Thanh Thuy Trinh, Van Duy Nguyen, Hong Hanh Nguyen, Jayapal Raja, Juyeon Jang, Kyungsoo Jang,
Kyunghyun Baek, Vinh Ai Dao, and Junsin Yi
Citation: Applied Physics Letters 100, 143502 (2012); doi: 10.1063/1.3699221
View online: http://dx.doi.org/10.1063/1.3699221
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/100/14?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Deep-level transient spectroscopy on an amorphous InGaZnO4 Schottky diode
Appl. Phys. Lett. 104, 082112 (2014); 10.1063/1.4867236
Origin of subgap states in amorphous In-Ga-Zn-O
J. Appl. Phys. 114, 163704 (2013); 10.1063/1.4826895
Quasi-reversible point defect relaxation in amorphous In-Ga-Zn-O thin films by in situ electrical measurements
Appl. Phys. Lett. 102, 122103 (2013); 10.1063/1.4796119
Electrical stability enhancement of the amorphous In-Ga-Zn-O thin film transistor by formation of Au
nanoparticles on the back-channel surface
Appl. Phys. Lett. 102, 102108 (2013); 10.1063/1.4795536
Impact of interface controlling layer of Al 2 O 3 for improving the retention behaviors of In–Ga–Zn oxide-based
ferroelectric memory transistor
Appl. Phys. Lett. 96, 232903 (2010); 10.1063/1.3452339
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
86.139.91.230 On: Tue, 20 May 2014 10:34:50

APPLIED PHYSICS LETTERS 100, 143502 (2012)
Operation mechanism of Schottky barrier nonvolatile memory with high
conductivity InGaZnO active layer
Thanh Thuy Trinh,1,2 Van Duy Nguyen,3,a) Hong Hanh Nguyen,1 Jayapal Raja,1
Juyeon Jang,1 Kyungsoo Jang,1 Kyunghyun Baek,1 Vinh Ai Dao,1,2 and Junsin Yi1,a)
1Information and Communication Device Laboratory, School of Information and Communication Engineering,
Sungkyunkwan University, South Korea
2Faculty of Materials Science, Vietnam National University, Ho Chi Minh City, Vietnam
3International Training Institute for Materials Science, Hanoi University of Science and Technology, Vietnam
(Received 28 November 2011; accepted 11 March 2012; published online 2 April 2012)
Influence of Schottky contact between source/drain electrodes and high conductivity a-InGaZnO
active layer to the performance of nonvolatile memory devices was first proposed. The Schottky
barrier devices faced to the difficulty on electrical discharging process due to the energy barrier
forming at the interface, which can be resolved by using Ohmic devices. A memory window of
2.83 V at programming/erasing voltage of 613 V for Ohmic and 5.58 V at programming voltage of
13 V and light assisted erasing at ꢀ7 V for Schottky devices was obtained. Both memory devices
using SiO2/SiOx/SiOxNy stacks showed a retention exceeding 70% of trapped charges 10 yr with
C
V
operation voltages of 613 V at an only programming duration of 1 ms. 2012 American Institute
of Physics. [http://dx.doi.org/10.1063/1.3699221]
In recent times, amorphous InGaZnO (a-IGZO) film has
been widely studied for using as an active layer in thin-film
transistors (TFTs) because of its inherent advantages, which
include high uniformity, transparence, and high mobility
compared to amorphous silicon.1,2 For next-generation appli-
cations, such as system-on-panel (SOP) displays, memory-
in-pixel, and transparent memory, an a-IGZO-based nonvo-
latile memory (NVM) is required, but high operating vol-
tages, retention loss, and cycling decay are challenges that
need to be addressed before arriving at the appropriate
material.3–5
deposition (IPCVD) process on a low-resistivity c-Si sub-
strate as gate electrode. An SiO2-blocking layer, with a
thickness of 20 nm, and an SiOx storage layer, with a thick-
ness of 20 nm at an SiH4:N2O ratio of 6:1, were sequentially
deposited at 170 ꢁC. Subsequently, an ultrathin amorphous
silicon (a-Si) film with an SiH4:H2 gas ratio of 5:20 was de-
posited with a RF power of 50 W and a temperature of
200 ꢁC for 2 min. Then, N2O plasma treatment was carried
out for the creation of the 3.2-nm-thick SiOxNy tunneling
layer. After the deposition of OOxOn memory stacks, an
active layer of 100-nm-thick a-IGZO film was deposited by
DC-pulsed magnetron sputter using a ceramic IGZO target
(Ga2O3:In2O3:ZnO ¼ 1:1:1) at room temperature. The base
vacuum level was maintained at a pressure lower than
5 ꢂ 10ꢀ5 Torr while the working pressure and DC power
were maintained at 5 ꢂ 10ꢀ3 Torr and 140 W, respectively,
during the sputtering. Before deposition, the presputtering
process was performed for 10 min to remove any contamina-
tion that may be present on the target surface. Then, the post-
annealing process was performed for an as-deposited sample
in an air atmosphere using rapid thermal annealing equip-
ment at a temperature of 250 ꢁC for 1 h. After the postanneal-
ing process, 150-nm-thick silver (Ag) or aluminum (Al)
films were vaporized to form Schottky or conventional
Ohmic devices, respectively. Finally, S/D regions were pat-
terned by the photolithography method using two mask-
patterning processes. The electrical and memory characteris-
tics of the devices were measured by Semiconductor Param-
eter Analyzer equipment (Model EL 420C).
The TFTs using Schottky barrier (SB) at source/drain
(S/D) electrodes (SBTFT) were introduced to improve the
off current in silicon TFTs.6–8 Hence, SBTFTs could be fab-
ricated on the high carrier concentration a-IGZO layer with-
out the problem of high off currents. Otherwise, the
increasing of mobility with carrier concentration is relied on
IGZO system.1,9 Due to these advantages, the IGZO SBTFTs
with high conductivity become potential candidates to
achieve higher field effect mobility (lFE).
In this study, the performance of NVM devices on
a-IGZO is investigated using the memory stack of
SiO2/SiOx/SiOxNy (OOxOn) that has previously been
reported to show some outstanding features, such as low
operating voltages and excellent retention.10,11 The devices
were fabricated in two types of TFT-NVM structures with
Schottky and Ohmic contacts between S/D electrodes and
active layer. The memory behavior of Schottky barrier NVM
(SBNVM) devices is explained in comparison to the
conventional Ohmic contact device (ONVM).
The bottom gate NVM structures were fabricated on two
kinds of a-IGZO films. The high conductivity layer was de-
posited in only Ar ambient condition and low conductivity
one was formed at Ar and O2 mixing gases with 80% O2
content. After heat treatment, conductivity of the two 100-
nm-thick films was found to be stable with values of about
101 and 10ꢀ4 S cmꢀ1, respectively. The SBNVM was fabri-
cated on high conductivity a-IGZO layer using 150-nm-thick
The multi-stack OOxOn-IGZO NVM devices were fab-
ricated in the following steps: OOxOn memory stacks were
deposited using inductive-coupled plasma chemical vapor
a)Authors to whom correspondence should be addressed. Electronic
addresses: nguyenvanduy@itims.edu.vn and yi@yurim.skku.ac.kr. Tel.:
þ82-31-290-7139. Fax: þ82-31-290-7159.
C
V
0003-6951/2012/100(14)/143502/4/$30.00
100, 143502-1
2012 American Institute of Physics
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
86.139.91.230 On: Tue, 20 May 2014 10:34:50

143502-2
Trinh et al.
Appl. Phys. Lett. 100, 143502 (2012)
FIG. 1. (a) Initial transfer characteristics
and lFE of SBNVM and ONVM devices,
(b) energy band diagram of IGZO
Ohmic contact, and (c) Schottky contact.
Work functions of materials are referred
from Refs. 12–14.
Ag electrodes, with a work function estimated to be about
4.74 eV,12 which has already been confirmed by structure an-
alyzed (result is not shown here). The ONVM used a-low-
conductivity active layer with 150-nm-thick Al electrodes
with a work function of about 4.26 eV.12 TFT characteristics
of the two types were measured and compared as in Fig. 1.
Note that the high-conductivity IGZO-TFT using Al electro-
des could not be modulated by the gate voltage like other
conventional TFTs (not shown here). The advantages of the
SB structure in the on current (ION) and lFE values are
assumed to the high carrier concentration of a-IGZO layer.
The unusual finding of mobility increases when carrier con-
centration increases up to 1020 cmꢀ3 in a-IGZO system,9
which suggests the possible explanation for the earlier
results. While high-conductivity a-IGZO film showed a car-
rier concentration and Hall mobility of about 1019 cmꢀ3 and
15 cm2 Vꢀ1 sꢀ1, the mobility in low-conductivity film could
not be measured using Hall equipment. When study with the
same device dimensions and measurement, SB structure pro-
vides an ION approximately 2 orders higher in magnitude
than that of the Ohmic contact structure. Field effect mobil-
ity of SBTFT reaches a value of about 6 cm2 Vꢀ1 sꢀ1 in
comparison to that of about 0.1 cm2 Vꢀ1 sꢀ1 for OTFT.
SBTFT operation mechanism has been proposed, in which
the gate voltage modulated the SB height in order to make it
possible to operate the devices (Fig. 1(c)). The OTFT oper-
ates using the conventional accumulation process, with
behavior very similar to those of the Si devices, because the
contact at the active and electrodes does not prevent the elec-
tron from moving (Fig. 1(b)).
Although the programming of the SBNVM showed
excellent performance, these devices showed difficulty in the
erasing process. As seen in Fig. 2(a), even negative biasing at
ꢀ15V for 10 s did not shift the threshold voltage to the nega-
tive region corresponding to the hole-ejection process. Mean-
while, the electron ejection process occurred at the positive
bias voltage of 10 V for 1 ms. In order to discharge the trapped
electron, fluorescent light with an intensity of 10mW cmꢀ2
was introduced during the negative-biasing process, called
light-assisted erasing. The erasing process can be successfully
FIG. 2. Switching characteristics of (a)
the SBNVM device and (c) the ONVM
device; threshold voltage shift of (b)
SBNVM and (d) OTFTNVM.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
86.139.91.230 On: Tue, 20 May 2014 10:34:50

143502-3
Trinh et al.
Appl. Phys. Lett. 100, 143502 (2012)
FIG. 3. (a) Ohmic and Schottky properties of
I–V characteristics of the MSM structure using
Al and Ag electrodes, respectively and (b) dia-
gram of voltage potential distribution over the
structure.
obtained using bias voltages combined with light assisted.
However, the subthreshold swing (SS) of such light-assisted
erasing process degrades when compared to the programming
one. Besides the hole trapping mechanism causing the thresh-
old voltage shift, the degradation of interface state under light
illumination and negative bias stress was supposed to thin film
transistor using a-IGZO material.15,16 During light illumina-
tion, the electron hole pairs were generated and they created
trap sites in material. The stretch-out phenomena of SS can be
attributed to photo- generated-traps in a-IGZO layer including
interface and bulk traps. In Fig. 2(b), the switching properties
of the SBNVM indicate the linear dependence of the threshold
voltage shift and bias voltage. For the programming process,
only gate bias voltage was applied to the structure for 1 ms,
from 10 to 13 V causing a threshold voltage shift from about
0.5 to 3 V. Using the light-assisted process, negative bias vol-
tages were applied from ꢀ4 to ꢀ7 V causing a threshold volt-
age shift from about ꢀ0.6 to ꢀ2.6V. The memory window of
the SBNVM structure was calculated to be higher than about
5.5 V with light assistance.
than the erasing duration as general due to the higher mobility
of electron.
The erasing difficulty phenomenon has also been
reported in some earlier studies,3–5 which could have differ-
ent explanations. Some reports attributed this phenomenon
to the n-type conductivity behavior of IGZO, due to which
very few holes are available for positive charge injection
into the gate insulator to neutralize the stored negative
charges.4 In this study, we attribute the electrical erasing
problem of the SBNVM to the potential barrier between the
Ag electrodes and a-IGZO layer. The Schottky contact char-
acteristics of IGZO/Ag are shown in Fig. 3(a) by measuring
current-voltage (I–V) characteristics of an Al/a-IGZO/Ag
(metal-semiconductor-metal (MSM)) structure. The I–V plot
proves that current through the structure is blocked when a
negative bias is applied on Ag electrode. Meanwhile, the Ag/
a-IGZO contact seems to open out when a positive bias is
applied on the electrode. This kind of Schottky contact
behavior is the key point for TFT operation using high con-
ductivity a-IGZO as active layer. The leakage current in
reversed bias region is assumed to the trap sites distributed
over metal and semiconductor junction as well as barrier
height of the contact. The improvement in the operation of
TFT using high conductivity a-IGZO is shown in Fig. 1.
Unfortunately, despite the advantages of using SB in the
TFT, it causes problems in performance of the NVM devices.
As illustrated in Fig. 3(b), when the negative bias applied on
the gate electrode, the SB increases with the negative bias volt-
age because of the blocking current phenomenon. The high
resistance at this barrier redistributes the voltage dropping over
the structure. In this case, an additional V4 component appears
While using the conventional Ohmic contact structure,
the problem in the erasing process was solved as seen in
Fig. 2(c). The programming and erasing processes were meas-
ured from 610 to 613 V and the largest memory window
achieved was about 2.83 V. However, as mentioned earlier,
the active layer for Ohmic contact structure needs to have a
low conductivity for achieving low leakage current. This is
the reason for the lower performance of ONVM compared to
SBNVM. All the threshold voltage shifts at different program-
ming and erasing biases are shown in Fig. 2(d). From this fig-
ure, the programming performance seems to be more effective
FIG. 4. Retention characteristics of (a)
SB NVM device for the programming
process and (b) Ohmic conventional
NVM structure.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
86.139.91.230 On: Tue, 20 May 2014 10:34:50

143502-4
Trinh et al.
Appl. Phys. Lett. 100, 143502 (2012)
significantly reducing the voltage drop, V3, over the tunneling
layer. This is the reason why the threshold voltage did not shift
even when the gate electrode was biased at ꢀ15 V for 10 s. The
light-assisted erasing process with recombination and tunneling
mechanism of trapped charges caused by photon energy is the
solution to this problem. This erasing process is not preferred
for modern NVM devices because of complexities in perform-
ance. In this study, we attempted to characterize retention
properties of the SBNVM using the electrical bias only for the
programming process.
This work was supported by Priority Research Centers
Program through the National Research Foundation of Korea
(NRF) funded by the Ministry of Education, Science and
Technology (2011-0018397).
1K. Nomura, H. Ohta, K. Ueda, T. Kamiya, M. Hirano, and H. Hosono,
Science 300, 1269 (2003).
2H. Yabuta, M. Sano, K. Abe, T. Aiba, T. Den, H. Kumomi, K. Nomura, T.
Kamiya, and H. Hosono, Appl. Phys. Lett. 89, 112123 (2006).
3N. C. Su, S. J. Wang, and A. Chin, IEEE Electron Device Lett. 31, 201
(2010).
4H. X Yin, S. N. Kim, H. Lim, Y. S. Min, C. J. Kim, I. H. Song, J. C Park,
S. W. Kim, A. Tikhonovsky, J. W Hyun, and Y. S Park, IEEE Trans. Elec-
tron Devices 55, 2071 (2008).
Figure 4 demonstrates the retention characteristics of the
SBNVM (Fig. 4(a)) and ONVM (Fig. 4(b)). The retention
time was measured for 104 s and extrapolated to 10 yr. The
results show a residue of 87% trapped charges after 10 yr
corresponding to the memory window of 2.67 V for
SBNVM. Meanwhile, with P/E biased at 613 V for 1 ms, the
memory window of the ONVM remains at 2.06 V corre-
sponding to 73% trapped charges. These results are matched
to OOxOn NVM properties using other active materials10,11
and suitable for device applications.
5Y. S. Park, S. Y. Lee, and J. S. Lee, IEEE Electron Device Lett. 31, 1134
(2010).
6J. H. Yang, C. G. Ahn, I. B. Baek, M. G. Jang, G. Y. Sung, B. C. Park, K.
J. Im, and S. J. Lee, Thin Solid Films 517, 1825–1828 (2009).
7J. W. Shin, W. J. Cho, C. J. Choi, and M. G. Jang, Appl. Phys. Lett. 94,
053502 (2009).
8J. W. Shin, C. J. Choi, M. G. Jang, and W. J. Cho, IEEE Electron Device
Lett. 29, 1336 (2008).
9K. Nomura, H. Ohta, A. Takagi, T. Kamiya, M. Hirano, and H. Hosono,
Nature 432, 488–492 (2004).
10H. H. Nguyen, V. D. Nguyen, T. T. Trinh, K. S. Jang, K. H Baek, J. Raja,
and J. S Yi, J. Electrochem. Soc. 158(10), H1077 (2011).
11N. V. Duy, S. W. Jung, K. Kim, D. N. Son, N. T. Nga, J. Cho, B. Choi,
and J. Yi, J. Phys. D: Appl. Phys. 43, 075101 (2010).
12CRC Handbook of Chemistry and Physics, edited by David R. Lide, 89th
ed. (Internet Version, 2009), pp. 12–114.
In summary, two types of IGZO TFT/NVM with SB and
Ohmic contacts have been fabricated and compared. The
SBTFT shows better performance compared to the conven-
tional OTFT, which is attributed to the high conductivity.
The difficulty in the erasing process of SBNVM was attrib-
uted to the SB height formed at the interface of the active
layer and electrodes. This barrier prevents the electron in the
reverse bias from moving through the electrodes. This disad-
vantage could be solved by using the ONVM structure,
which shows conventional programming/erasing processes
with a memory window of 2.83 V.
13P. S. Yun, and J. Koike, J. Electrochem. Soc. 158(10), H1034–H1040
(2011).
14K. W. Lee, K. Y Heo, and H. J. Kim, Appl. Phys. Lett. 94, 102112 (2009).
15K. H. Ji, J. I. Kim, H. Y. Jung, S. Y. Park, R. N. Choi, U. K. Kim, C. S.
Hwang, D. S. Lee, H. S. Hwang, and J. K. Jeong, Appl. Phys. Lett. 98,
103509 (2011).
16T. Y. Hsieh, T. C. Chang, T. C. Chen, M. Y. Tsai, W. H. Lu, S. C. Chen,
F. Y. Jian, and C. S. Lin, Thin Solid Films 520, 1427–1431 (2011).
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
86.139.91.230 On: Tue, 20 May 2014 10:34:50