Structural and compositional dependence of gadolinium-aluminum oxide for the application of charge-trap-type nonvolatile memory devices

Article abstract of DOI:10.1063/1.3309693

The structural and compositional dependence of gadolinium-aluminum oxide (GdAlO) for application to nonvolatile memory is investigated. An addition of Gd into AlO reduces the leakage current, which improves the erase window. The GdAlO film crystallizes into many different phases after annealing depending on the Gd percentage when the amount of Gd exceeds 49%. The crystallization of the GdAlO film causes a change in the band gap of the GdAlO film, resulting in a change of the retention properties. It is also found that crystallized GdAlO is more vulnerable to the generation of traps by electrical stress. The results indicate that careful optimization of the Gd percentage in GdAlO is necessary to utilize the benefit of GdAlO with minimum deterioration in the charge retention property.

Full text of DOI:10.1063/1.3309693

Structural and compositional dependence of gadolinium-aluminum oxide for the
application of charge-trap-type nonvolatile memory devices
Youngmin Park, Jong Kyung Park, Myeong Ho Song, Sung Kyu Lim, Jae Sub Oh, Moon Sig Joo, Kwon Hong
, and Byung Jin Cho
Citation: Applied Physics Letters 96, 052907 (2010); doi: 10.1063/1.3309693
View online: http://dx.doi.org/10.1063/1.3309693
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/96/5?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Impact of composition and crystallization behavior of atomic layer deposited strontium titanate films on the
resistive switching of Pt/STO/TiN devices
J. Appl. Phys. 116, 064503 (2014); 10.1063/1.4891831
Performance improvement of metal-Al2O3-HfO2-oxide-silicon memory devices with band-engineered Hf-
aluminate/SiO2 tunnel barriers
J. Vac. Sci. Technol. B 31, 041201 (2013); 10.1116/1.4807842
Experimental evidence and modeling of two types of electron traps in Al2O3 for nonvolatile memory
applications
J. Appl. Phys. 113, 074501 (2013); 10.1063/1.4792038
Effects of rapid thermal annealing on structure and electrical properties of Gd-doped HfO 2 high k film
Appl. Phys. Lett. 98, 082906 (2011); 10.1063/1.3556652
Low temperature crystallization of high permittivity Ta oxide using an Nb oxide thin film for
metal/insulator/metal capacitors in dynamic random access memory applications
J. Vac. Sci. Technol. B 23, 80 (2005); 10.1116/1.1829060
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:
193.0.65.67 On: Wed, 10 Dec 2014 10:31:53

APPLIED PHYSICS LETTERS 96, 052907 ͑2010͒
Structural and compositional dependence of gadolinium-aluminum oxide
for the application of charge-trap-type nonvolatile memory devices
Youngmin Park,1 Jong Kyung Park,1 Myeong Ho Song,2 Sung Kyu Lim,2 Jae Sub Oh,2
Moon Sig Joo,3 Kwon Hong,3 and Byung Jin Cho1,a͒
1Department of Electrical Engineering, Korea Advanced Institute of Science and Technology,
373-1 Guseong-dong, Yuseong-gu, Daejeon 305-701, Republic of Korea
2National Nanofab Center, 373-1 Guseong-dong, Yuseong-gu, Daejeon 305-701, Republic of Korea
3Hynix Semiconductor Inc., San 136-1, Ami-ri, Bubal-eub, Icheon-si, Gyeonggi-do 467-701,
Republic of Korea
͑Received 11 December 2009; accepted 15 January 2010; published online 5 February 2010͒
The structural and compositional dependence of gadolinium-aluminum oxide ͑GdAlO͒ for
application to nonvolatile memory is investigated. An addition of Gd into AlO reduces the leakage
current, which improves the erase window. The GdAlO film crystallizes into many different phases
after annealing depending on the Gd percentage when the amount of Gd exceeds 49%. The
crystallization of the GdAlO film causes a change in the band gap of the GdAlO film, resulting in
a change of the retention properties. It is also found that crystallized GdAlO is more vulnerable to
the generation of traps by electrical stress. The results indicate that careful optimization of the Gd
percentage in GdAlO is necessary to utilize the benefit of GdAlO with minimum deterioration in the
charge retention property. © 2010 American Institute of Physics. ͓doi:10.1063/1.3309693͔
Charge-trap-type nonvolatile flash memory is considered
to be suitable for further scaling of flash memory devices.1–3
Most of charge-trap-type memory flash memory devices
use the TANOS ͑TaN gate electrode—Al2O3 blocking
oxide—Si3N4 charge trapping layer—SiO2 tunnel oxide—
silicon substrate͒ structure.4 However, the relatively low di-
electric constant ͑ϳ9͒ of the Al2O3 blocking oxide sets a
limitation for further scaling of these devices. In order to
meet the requirements of both a fast operating speed and
good charge retention in further scaled devices, a blocking
layer with a higher k value without much sacrifice of the
conduction band offset is required.5 Among high-k oxide
candidates, gadolinium oxide possesses desirable properties
for this type of blocking layer, such as a relatively high di-
electric constant ͑17͒,6 a large band gap ͑6.4 eV͒,7 and a
large conduction band offset ͑3.1 eV͒.7 It was shown in a
previous report that sputtered Gd2O3 had the advantages of a
faster erase speed and comparable charge retention compared
to a Al2O3 blocking layer in charge-trap-type flash memory.8
Further improvement of the charge retention characteristics
was also reported after an addition of aluminum into gado-
linium oxide through a cosputtering process.9 However,
there is no detailed study for the physical properties of
gadolinium-aluminum oxide. This paper presents the
structural and compositional dependence of gadolinium-
aluminum oxide ͑GdAlO͒ on the physical and electrical
properties of the memory devices with a GdAlO blocking
layer.
͑ALD͒ method using the Eureka 3000 by Jusung Engineer-
ing. For the ALD process, trimethyl aluminum and
Gd͑iPrCp͒ precursors provided by UP Chemical Co. were
3
used to form Al2O3 and Gd2O3 films, respectively, with
ozone used as an oxidant. The gadolinium percentage in the
GdAlO layer was controlled by adjusting the ALD cycles of
the Al2O3 and Gd2O3. For comparison, the thickness of each
high-k blocking oxide layer was selected so that the equiva-
lent oxide thickness ͑EOT͒ of the entire gate stack was simi-
lar ͑ϳ13.5 nm͒. A 200 nm thick TaN layer was deposited for
the metal gate by reactive sputtering. After the patterning of
the gate electrode and source/drain implantation, a rapid
thermal annealing process was done at 900 °C for 30 s in a
N2 ambient for dopant activation. X-ray photoelectron spec-
troscopy ͑XPS͒ measurements were carried out to measure
the atomic concentration, band gap, and band offset. The
thickness of each high-k blocking oxide was measured by a
spectroscopic ellipsometer.
A low leakage current of the blocking oxide is important
to improve the erase speed and enhance the erase window by
minimizing the electron back-tunneling action from the gate
during erase operations. Figure 1 shows the leakage current
density of GdAlO films with a various percentages of gado-
linium. The 0% Gd and the 100% Gd concentrations in the
figure refer to pure aluminum oxide and pure gadolinium
oxide, respectively. The leakage current of the GdAlO film
was evaluated using whole gate stack ͑GdAlO/Si3N4/SiO2
tunnel oxide͒ devices that simulated actual charge-trap de-
vices, as illustrated in the inset of Fig. 1. Because high-k
dielectric properties such as nucleation and interface reaction
depend on the substrate, it would be more useful to compare
the properties using the actual device structure which the
high-k is to be applied for. The Si3N4 and SiO2 thicknesses
were fixed at 6 nm and 4.5 nm, respectively. The result
shows that the GdAlO devices have much lower leakage cur-
rent and higher dielectric breakdown voltages compared to
the Al2O3 device. For easy recognition of the Gd percentage
After a standard gate precleaning process, a 4.5 nm thick
tunnel oxide ͑SiO2͒ was thermally grown on a p-type Si
substrate and a 6.0 nm thick Si3N4 layer was deposited by
low-pressure chemical vapor deposition to form the charge-
trapping layer. Blocking oxides of Al2O3, GdAlO, and
Gd2O3 were deposited using an atomic layer deposition
a͒
Author to whom correspondence should be addressed. Electronic mail:
bjcho@ee.kaist.ac.kr.
0003-6951/2010/96͑5͒/052907/3/$30.00
96, 052907-1
© 2010 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:
193.0.65.67 On: Wed, 10 Dec 2014 10:31:53

052907-2
Park et al.
Appl. Phys. Lett. 96, 052907 ͑2010͒
FIG. 1. ͑Color online͒ The leakage current density of GdAlO devices with
0%, 18%, 35%, 49%, 71%, 82%, and 100% Gd. The leakage current was
evaluated on Si3N4͑6.0 nm͒/SiO2͑4.5 nm͒ ͑see inset figure͒. All samples
were annealed at 900 °C for 30 s as the simulating source/drain annealing
step.
FIG. 3. ͑Color online͒ ͑a͒ XRD data of GdAlO thin film with 35%, 49%,
71%, and 100% Gd. The crystallized GdAlO film shows many different
phases depending on the Gd percentage. ͑b͒ The band gap ͑filled symbols͒
and the conduction band offset against silicon ͑open symbols͒ measured by
XPS on GdAlO film directly deposited on silicon. Crystallization of GdAlO
results in an increase in the band gap.
dependence on the dielectric properties, the leakage current
density at a fixed voltage of Ϫ20 V ͑filled symbols͒ is plotted
as a function of the Gd percentage together with the erase
window ͑open symbols͒ in Fig. 2͑a͒. The reason for compar-
ing the erase characteristics rather than both program/erase is
that the program speed does not change much by changing
the blocking oxide layer if the EOT of the gate stack is the
same. However, when the blocking oxide is leaky, there will
be significant electron back-tunneling from the gate electrode
during erase operation, which results in early saturation of
the erase state, leading to smaller erase window. Hence, the
erase operation is more sensitive to the property of the block-
ing oxide layer. From the result in Fig. 2͑a͒, it is interesting
to note that the leakage current density decreases with the Gd
percentage up to 35% and then bounces back for a higher Gd
percentage. The erase window also shows the similar trend,
as expected. On the other hand, the dielectric constant of the
GdAlO film increases steadily with the Gd percentage, as
shown in Fig. 2͑b͒. As the final EOT values of the tested
GdAlO films are similar, the physical thickness of GdAlO
film with a higher Gd percentage is thicker. Despite the in-
crease in the physical thickness, the leakage current increases
beyond 35% of Gd in GdAlO.
centages. Figure 3͑a͒ shows the XRD results of GdAlO thin
films with 35%, 49%, 71%, and 100% Gd percentages after
annealing at 900 °C for 30 s. As expected, a higher Gd per-
centage in GdAlO resulted in crystallization of the GdAlO
film. Surprisingly, however, GdAlO films with 49%, 71%,
and 100% Gd showed different phases of orthorhombic,
monoclinic, and cubic, respectively. It is also interesting to
note that GdAlO with 35% Gd remains amorphous after an-
nealing. This percentage is coincident with the turning point
of the leakage current in Fig. 2͑a͒. Therefore, it can be con-
cluded that the increase in the leakage current beyond 35% at
Gd is due to the crystallization of the GdAlO film. It is
known that the conductivity of the dielectric can be increased
by conduction through the grain boundaries of crystallized
dielectrics.10 The band gap and band offset of the GdAlO
thin film as a function of the Gd percentage was investigated
as well. Using the XPS O 1s energy loss spectra and the
valence band spectra,11–13 the band gap and valence band
offsets were measured and the conduction band offset was
determined. These results are shown in Fig. 3͑b͒. It was ex-
pected that the band gap of the GdAlO would decrease as the
Gd percentage increased considering the reported band gap
of Al2O3 and Gd2O5 films. However, this was true only up to
35% of Gd. Beyond 35% of Gd, no additional decrease in the
band gap was observed. The band gap and band offset of
GdAlO increase beyond 71% of Gd. The minimum point of
the band gap is coincident with the point of the structural
change from amorphous to crystalline. The result in Fig. 3
shows that among the crystallized phases, the cubic phase of
GdAlO has the highest band gap and band offset. Such band
gap widening after crystallization and structural change was
also reported in other dielectrics, such as Al2O3.13,14
To investigate the mechanism behind this unusual behav-
ior of the leakage current, x-ray diffraction ͑XRD͒ and XPS
analyses were performed for GdAlO with different Gd per-
The manner in which the structural change of GdAlO
blocking layer affects the data retention properties in charge-
trap-type nonvolatile memory was investigated. The charge
retention property was evaluated using a programmed device
by monitoring the change in the flat-band voltage ͑⌬Vfb͒.
Here ⌬Vfb is defined as Vfb͑t͒−Vfb͑0͒, where Vfb͑t͒ is the
flat-band voltage at a certain elapse time during the retention
test, and Vfb͑0͒ is the flat-band voltage immediately after
FIG. 2. ͑Color online͒ ͑a͒ Comparison of the leakage current at Ϫ20 V
͑filled symbols͒ and erase window ͑open symbols͒ and ͑b͒ dielectric con-
stant of GdAlO film as a function of the Gd percentage.
program. Three different evaluation criteria were used as fol-
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:
193.0.65.67 On: Wed, 10 Dec 2014 10:31:53

052907-3
Park et al.
Appl. Phys. Lett. 96, 052907 ͑2010͒
oxide owing to the physically thicker dielectrics for the same
EOT. The reduced leakage current leads to an improvement
of the erase window. When the Gd percentage is high, the
film is crystallized into many different phases depending on
the Gd percentage. The crystallization of the GdAlO film
causes a change of the band gap of the GdAlO film, resulting
in a change of the retention properties. It was also found that
once GdAlO is crystallized, the film becomes more vulner-
able to trap generation. Therefore, for the design of flash
memory devices, it is necessary to consider the trade-off re-
lationship between the erase window and retention properties
which depend on the Gd concentration in GdAlO film.
FIG. 4. ͑Color online͒ The charge retention property of fresh devices at
room temperature ͑open circle͒ and 150 °C ͑filled square͒ and that of a
device cycled for 1000 program/erase at 150 °C ͑open diamond͒. Due to the
increase of the band gap of the crystallized GdAlO, the charge loss property
is improved. However, the crystallized GdAlO is more vulnerable to trap
generation during program/erase cycling.
This work was financially supported by Hynix Semicon-
ductor Inc. The authors would like to thank Jusung Engineer-
ing Co. and UP Chemical Co. for the ALD equipment and
the precursor supports, respectively.
1K. Kim and J. Choi, Proceedings of the IEEE Nonvolatile Semiconductor
Memory Workshop 2006 ͑IEEE, Monterey, CA, 2006͒ , p. 9.
2Y. Park, J. Choi, C. Kang, C. Lee, Y. Shin, B. Choi, J. Kim, S. Jeon, J. Sel,
J. Park, K. Choi, T. Yoo, J. Sim, and K. Kim, Tech. Dig. - Int. Electron
Devices Meet. 2006, 346900.
lows: ͑1͒ ⌬Vfb after 24 h at 150 °C, ͑2͒ ⌬Vfb after 11 days at
room temperature and then projected to 10 years, and ͑3͒
1000 program and erase cycling followed by measuring ⌬Vfb
after 24 h at 150 °C. The charge loss, represented by ⌬Vfb is
plotted as a function of the gadolinium percentage in Fig. 4.
The result shows that at room temperature, ͉⌬Vfb͉ increases
with the Gd percentage up to 71% and then bounces back
and decreases. This indicates that the maximum charge loss
occurs at 71% Gd at room temperature. However, at the high
temperature of 150 °C, the maximum charge loss occurs at
49% of Gd. The charge loss is then reduced for higher Gd
concentrations. It is well known that at a high temperature
such as 150 °C, the thermionic emission over the conduction
band of the blocking oxide becomes even more important for
the charge loss mechanism.15 The retention data curve at
150 °C in Fig. 4 resembles very much the band gap and
band offset curve in Fig. 3͑b͒. Therefore, the retention data is
further evidence of the increase in band gap and band offset
for the crystallized GdAlO. When the device is repeatedly
programmed and erased, charge traps are formed in the
blocking oxide, resulting in degradation of the retention
property. The charge loss properties of the GdAlO devices
were evaluated after 1000 program/erase ͑P/E͒ cycles at
150 °C. This result is shown in Fig. 4 ͑open diamonds͒.
Compared to P/E cycled and fresh devices, there is no deg-
radation after P/E cycling when the Gd percentage is below
35%. However, noticeable degradation occurs in the devices
with Gd percentages of 49% and above, which is the percent-
age range for the crystallization of GdAlO. This result indi-
cates that crystallized GdAlO is more vulnerable to charge
trap generation by electrical stress.
3C. H. Lee, K. I. Choi, M. K. Cho, Y. H. Song, K. C. Park, and K. Kim,
Tech. Dig. - Int. Electron Devices Meet. 2003, 613 ͑2003͒.
4C. H. Lee, C. Kang, J. Sim, J. S. Lee, J. Kim, Y. Shin, K. T. Park, S. Jeon,
J. Sel, Y. Jeong, B. Choi, V. Kim, W. Jung, C. I. Hyun, J. Choi, and K.
Kim, Proceedings of the IEEE Nonvolatile Semiconductor Memory Work-
shop 2006 ͑IEEE, Monterey, CA, 2006͒, p. 54.
5J. A. Kittl, K. Opsomer, M. Popovici, N. Menou, B. Kaczer, X. P. Wanga,
C. Adelmann, M. A. Pawlak, K. Tomida, A. Rothschild, B. Govoreanu, R.
Degraeve, M. Schaekers, M. Zahid, A. Delabie, J. Meersschaut, W. Pols-
poel, S. Clima, G. Pourtois, W. Knaepen, C. Detavernier, V. V. Afanas’ev,
T. Blomberg, D. Pierreux, J. Swerts, P. Fischer, J. W. Maes, D. Manger, W.
Vandervorst, T. Conard, A. Franquet, P. Favia, H. Bender, B. Brijs, S. Van
Elshocht, M. Jurczak, J. Van Houdt, and D. J. Wouters, Microelectron.
Eng. 86, 1789 ͑2009͒.
6W. H. Chang, C. H. Lee, P. Chang, Y. C. Chang, Y. J. Lee, J. Kwo, C. C.
Tsai, J. M. Hong, C.-H. Hsu, and M. Hong, J. Cryst. Growth 311, 2183
͑2009͒.
7T. Hattori, T. Yoshida, T. Shiraishi, K. Takahashi, H. Nohira, S. Joumori,
K. Nakajima, M. Suzuki, K. Kimura, I. Kashiwagi, C. Ohshima, S. Ohmi,
and H. Iwai, Microelectron. Eng. 72, 283 ͑2004͒.
8J. Pu, S. J. Kim, Y. S. Kim, and B. J. Cho, Electrochem. Solid-State Lett.
11, H252 ͑2008͒.
9J. Pu, D. S. H. Chan, S. J. Kim, and B. J. Cho, IEEE Trans. Electron
Devices 56, 2739 ͑2009͒.
10O. Bierwagen, L. Geelhaar, X. Gay, M. Piešiņš, H. Riechert, B. Jobst, and
A. Rucki, Appl. Phys. Lett. 90, 232901 ͑2007͒.
11S. Miyazaki, H. Nishimura, M. Fukuda, L. Ley, and J. Ristein, Appl. Surf.
Sci. 113-114, 585 ͑1997͒.
12H. Y. Yu, M. F. Li, B. J. Cho, C. C. Yeo, M. S. Joo, D.-L. Kwong, J. S.
Pan, C. H. Ang, J. Z. Zheng, and S. Ramanathan, Appl. Phys. Lett. 81,
376 ͑2002͒.
13S. Miyazaki, J. Vac. Sci. Technol. B 19, 2212 ͑2001͒.
14V. V. Afanas’ev and A. Stesmans, J. Appl. Phys. 102, 081301 ͑2007͒.
15A. Arreghini, N. Akil, F. Driussi, D. Esseni, L. Selmi, and M. J. van
Duuren, Solid-State Electron. 52, 1460 ͑2008͒.
In conclusion, this study showed that the use of GdAlO
can significantly reduce the leakage current of the blocking
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:
193.0.65.67 On: Wed, 10 Dec 2014 10:31:53