Property changes of aluminum oxide thin films deposited by atomic layer deposition under photon radiation

Article abstract of DOI:10.1149/1.2186179

Aluminum oxide thin films were deposited by atomic layer deposition (ALD) using trimethyl aluminum and water under UV radiation. Films were prepared at two different deposition temperatures of 260 and 370°C. The films deposited at 370°C had a higher OAl atomic ratio than those deposited at 260°C. The upper valence band (VB) spectrum, observed by X-ray photoelectron spectroscopy (XPS), of the aluminum oxide thin film deposited under UV radiation during water purge was different from that of a nonradiated thin film, indicating a structural change by the UV radiation. For the film deposited at 370°C, the effective mass (m*) was estimated to be larger than 0.35 m0, and an effective electron barrier height of 3.8 eV was obtained for m* =0.4 m0 by fitting the I-V curve with the Fowler-Nordheim (FN) tunneling model. The films, of which leakage current densities were reduced by UV radiation, showed an increase in electron barrier height for FN tunneling conduction. In addition, the conduction band offset with Si, calculated from the VB offsets and the bandgap values obtained by XPS, was also increased for the UV-radiated sample.

Full text of DOI:10.1149/1.2186179

Journal of The Electrochemical Society, 153 ͑6͒ F87-F93 ͑2006͒
F87
0013-4651/2006/153͑6͒/F87/7/$20.00 © The Electrochemical Society
Property Changes of Aluminum Oxide Thin Films Deposited
by Atomic Layer Deposition under Photon Radiation
,z
*
*
Sang Yong No, Dail Eom, Cheol Seong Hwang, and Hyeong Joon Kim
School of Materials Science and Engineering and Inter-University Semiconductor Research Center, Seoul
National University, Seoul 151-744, Korea
Aluminum oxide thin films were deposited by atomic layer deposition ͑ALD͒ using trimethyl aluminum and water under UV
radiation. Films were prepared at two different deposition temperatures of 260 and 370°C. The films deposited at 370°C had a
higher O/Al atomic ratio than those deposited at 260°C. The upper valence band ͑VB͒ spectrum, observed by X-ray photoelectron
spectroscopy ͑XPS͒, of the aluminum oxide thin film deposited under UV radiation during water purge was different from that of
a nonradiated thin film, indicating a structural change by the UV radiation. For the film deposited at 370°C, the effective mass
͑m*͒ was estimated to be larger than 0.35m0, and an effective electron barrier height of 3.8 eV was obtained for m* = 0.4m0 by
fitting the I-V curve with the Fowler–Nordheim ͑FN͒ tunneling model. The films, of which leakage current densities were reduced
by UV radiation, showed an increase in electron barrier height for FN tunneling conduction. In addition, the conduction band
offset with Si, calculated from the VB offsets and the bandgap values obtained by XPS, was also increased for the UV-radiated
sample.
© 2006 The Electrochemical Society. ͓DOI: 10.1149/1.2186179͔ All rights reserved.
Manuscript submitted August 1, 2005; revised manuscript received January 26, 2006. Available electronically April 6, 2006.
The continuing drive to shrink silicon devices requires a higher
capacitance for gate dielectric layers of complementary metal oxide
semiconductor field effect transistors ͑CMOSFETs͒ and storage ca-
pacitors of dynamic random access memory ͑DRAM͒ devices.1-3
The thickness shrinkage of the dielectric layer for a higher capaci-
tance produces problems of a high leakage current by electron tun-
neling and device instability.4,5 For circumventing these problems,
metal oxides with dielectric constants ͑␬͒ larger than those of SiO2
or Si3N4 have been investigated intensively because the same ca-
pacitance is obtained by thicker films of high-␬ materials with re-
duced tunneling current.6
Al2O3 has dielectric constant ␬ ͑8 ϳ 9͒ at least twice higher
than SiO2 ͑3.9͒ and shows thermodynamic stability on Si. The thick-
ness of the Al2O3 film can be effectively reduced without causing a
large tunneling current due to the larger conduction band offset than
other high-␬ materials. Thus, Al2O3 is a candidate material to re-
place SiO2 layers in microelectronic devices, such as the dielectric
layer of CMOSFETs, and DRAM capacitors.7-9 Recently, a mixture
of Al2O3 and other higher ␬ materials has been studied to obtain a
more stable insulation with a higher effective ␬.10-13
were radiated during the ALD process. The changes under far UV
are reported for TMA or H2O molecules adsorbed on the surface.
Methyl groups are removed from the TMA molecules by photo-
dissociation.22-24 Photon-stimulated desorption or photodissociation
of water may occur by UV radiation.25 In ALD using TMA and
H2O, the surface is covered by methyl groups after TMA injection.
After the methyl groups react with the H2O molecules subsequently
injected, hydroxyl groups are left on the surface.19 When UV is
radiated on the surface after TMA injection ͑UV1͒, part of the me-
thyl groups are removed before a reaction with H2O. At these sites,
aluminum oxide grows during the H2O injection step without typical
ALD reaction, which produces CH4 molecules. For a UV exposure
after H2O injection ͑UV2͒, parts of the hydroxyl groups or adsorbed
H2O weakly bonded to the surface are removed before a reaction
with TMA. Thus, film formation by the reaction of TMA with
weakly bonded species may be reduced. Because detailed film struc-
ture can be altered by variation of film formation reaction, the
change of film properties was investigated according to the UV ra-
diation sequence. In addition, the deposited films underwent various
postdeposition treatments and the effects were also investigated.
The Al2O3 thin film can be prepared by various methods such
as molecular beam epitaxy,14 chemical vapor deposition,15 plasma-
enhanced metallorganic chemical vapor deposition,16 reactive sput-
tering,17 and atomic layer deposition ͑ALD͒.18,19 For the applica-
tions mentioned above, ALD has some advantages compared to the
other deposition methods due to precise thickness control, enhanced
thickness uniformity, and excellent conformal deposition on three-
dimensional structures.
ALD of Al2O3 thin films using trimethylaluminum ͑TMA͒ and
H2O is a well-established process. The films deposited by ALD are
amorphous, having only a short-range order. Because crystalline Al
oxide and their hydroxide have diverse structures,20 the films may
have different forms of short-range order according to the deposition
processes or postdeposition treatments. The different structures are
expected to affect the electrical properties of films.
In this study, the films were deposited by ALD at two different
deposition temperatures ͑Td͒ of 260 and 370°C, and the temperature
influence on the structure and electrical properties was investigated.
From earlier reports on Al deposition by TMA in the range of
350–550°C,21 it was indicated that a slight thermal decomposition
of TMA occurs at 370°C and the thermal decomposition may af-
fect the film properties. In addition, photons of the far-UV range
Experimental
Figure 1 shows a schematic diagram of the ALD system
equipped with a UV source. UV light is radiated into the substrate
through the quartz window and controlled by an air-operating shut-
ter. TMA and H2O are used as Al source and oxidant, respectively.
Al2O3 thin films are deposited at a Td of 260 and 370°C on p-type
Si͑100͒ wafers which are prepared by SPM cleaning ͑sulfuric-
peroxide mixture, H2SO4:H2O = 4:1,120°C͒ with the native oxide
removed by dilute HF solution ͑HF:H2O = 1:10͒. The chamber wall
temperature is maintained at 120°C to facilitate the desorption of
source and oxidant from the wall surface during the purge sequence.
The temperature of the TMA and H2O canister was maintained at
5°C by circulating water. The flow rate of the Ar purge gas was
fixed at 600 sccm, and the chamber pressure was maintained below
0.3 Torr. To completely remove the residual source or oxidant, rela-
tively long intervals of 8 and 30 s were used for the purge of TMA
and H2O, respectively. A commercially available low-pressure mer-
cury lamp was used as the UV source. It mainly emits UV light of
253.7 and 183.9 nm with intensity ratio of ϳ5/2. The intensity be-
neath the quartz view port was measured to be 3 mW/cm2, which
was detected in the range of 200–280 nm, and the distance between
the sample and view port is about 5 cm. As shown in Fig. 2, UV was
switched on for 3 s during the purge step of TMA or H2O in each
ALD cycle. Three kinds of samples were prepared according to the
existence and exposure step of UV radiation: sample No UV was
*
Electrochemical Society Active Member.
z E-mail: cheolsh@plaza.snu.ac.kr
Downloaded on 2015-05-07 to IP 128.122.253.212 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

F88
Journal of The Electrochemical Society, 153 ͑6͒ F87-F93 ͑2006͒
Figure 1. Schematic of the ALD chamber with UV lamp.
deposited without UV exposure, and samples UV1 and UV2 were
exposed to UV radiation for 3 s in the purge step of TMA and H2O,
respectively.
Thickness and crystallinity were observed by high-resolution
transmission electron microscopy. The surface roughness of the
films was checked by atomic force microscopy ͑PSI, Autoprobe
CP͒. Composition and chemical properties of the films were ana-
lyzed by X-ray photoelectron spectroscopy ͑XPS, Kratos, AXIS-
His͒ and Auger electron spectroscopy ͑AES, Perkin-Elmer 660͒.
The Pt electrode was deposited using electron-beam evaporation
through a shadow mask, and forming gas annealing ͑FGA͒ was
carried out at 400 and 450°C in dilute H2 ͑N2 + 5%H2͒ ambient for
30 min. The electrical properties of the fabricated capacitors were
measured by an HP4140 pA meter and HP4194 impedance analyzer.
Figure 3. AES concentration profiles of ͑a͒ sample No UV deposited at
260°C, ͑b͒ sample No UV deposited at 370°C, ͑c͒ sample UV1 deposited at
370°C, and ͑d͒ sample UV2 deposited at 370°C.
decomposition of adsorbed Al precursor. It was reported that TMA is
thermally decomposed in the range of 350–550°C, resulting in an Al
deposition containing a large amount of carbon.21 Thus, thermal
decomposition of adsorbed Al precursor is expected at a Td of
370°C during the TMA purge period in this experiment, although
the decomposition is not significant enough to lead to a film growth
without oxidant injection. It has been reported that the decomposi-
tion of TMA molecules that were adsorbed on Si͑100͒ generated
Al-CHx ͑x = 1,2͒ bonds by thermal reactions between the methyl
groups bonded to the Al atoms with formation of CH4.28 The CH3
attached to Al can be removed in the form of CH4 by obtaining only
one H from the reaction with H2O. However, CHx ͑x = 1,2͒ species
should get more H from the reaction with H2O to form the volatile
CH4. Thus, it seems that these CHx groups generated by thermal
decomposition are removed by H2O less effectively than CH3 at-
tached to Al, resulting in an increased carbon contamination. UV
radiation reduced the carbon incorporation at a Td of 370°C below
1 atom % for both radiation conditions, as shown in Fig. 3c and d.
This was mainly attributed to the effects of photoexcitation of car-
bon elimination reactions or photodissociation of CHx groups at-
tached to Al or the oxide surface.
Results and Discussion
The deposited films showed different properties according to Td
and UV radiation. Figure 3 shows AES concentration profiles of
aluminum oxide films deposited in various conditions. From com-
parison of Fig. 3a and b, the O/Al atomic ratio increased roughly
from 2.4 to 3.9 as the substrate temperature increased from 260 to
370°C. Deposition of oxygen–excess aluminum oxide was also re-
ported by Klie et al.26 In this experiment of ALD using TMA and
H2O, it is concluded that deposition temperature was a critical factor
for the formation of nonstoichiometric thin films. The incorporation
of carbon impurity was below 1 atom % at a substrate temperature
of 260°C. A growth rate of 0.9 Å/cycle, which is a typical value for
the ALD using TMA and H2O,19,27 was also observed even at a high
Td of 370°C. Although it is well known that carbon bonded to TMA
molecules is easily removed in the form of CH4 during ALD of
Al2O3 thin films using TMA and H2O,19 the samples deposited at
370°C showed a higher carbon contamination of 5 atom %. The
increase of carbon concentration is attributed to a slight thermal
The effects of UV radiation were investigated by XPS. Figure 4
shows the XPS valence band ͑VB͒ spectra of aluminum oxide in
three samples, which were all deposited at 370°C, with the indica-
Figure 4. VB spectra of sample ͑a͒ No UV, ͑b͒ UV1, and ͑c͒ UV2 prepared
at a Td of 370°C at two different emission angles of 0 and 60° ͑2-nm-thick
aluminum oxide on Si͒. The background spectrum of Si was subtracted using
H-terminated Si VB spectra.
Figure 2. Schematic of a UV radiation sequence during the ALD process.
Downloaded on 2015-05-07 to IP 128.122.253.212 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 153 ͑6͒ F87-F93 ͑2006͒
F89
tion of Si–O–Si bonding was similar and that carbon-related species
containing oxygen was ignorable,30 the amount of hydrated bonding,
H–O–Al, was not significantly changed by UV radiation.
Though at a ␪ of 0° a subtle change of chemical binding states
e
by UV radiation could be detected in the VB spectra, core-level
spectra of Al 2p and O 1s spectra ͑not shown͒ did not show a
notable difference according to UV radiation. The peak shapes of Al
2p and O 1s spectra could not be differentiated from each other. At
a ␪e of 0°, the O/Al atomic ratio, estimated from the O 1s and Al 2p
peak areas and the sensitivity factors, was 4.16, 4.05, and 3.91 for
sample No UV, UV1, and UV2, respectively.
For the differences in the intensity of the low-binding-energy
part in the UVB spectra measured at 60°, a structural change by UV
radiation is considered. Though the addition of hydroxyl species to
the film was reported to increase the intensity of the low-binding-
energy part in the UVB,20 the difference in the degree of hydration
was ruled out due to the different intensity of low-binding-energy
part, because core-level spectra detected a similar amount of hy-
drated species. Thomas and Sherwood showed that an octahedral
unit ͑AlO96−͒ produces a high-binding-energy peak with larger inten-
sity than the low-binding-energy peak in UVB, whereas a tetrahe-
dral unit ͑AlO54−͒ causes one distinct peak in the lower-binding-
energy region of UVB with reduction of intensity in the high-
binding-energy part by X␣ calculation of the VB spectra.20 It is
believed that structures containing tetrahedrally coordinated Al are
dominant over those of octahedrally coordinated Al in the region
close to the surface for sample No UV and UV1 in contrast to the
UV2 sample. It is known that Al3+ ions in bulk aluminum hydrox-
ides have a strong tendency to occupy octahedral sites in an oxygen
lattice.31 Generally, incompletely bonded atoms near the surface re-
gion are rearranged to minimize the energy. Thus, the surface region
could have stable structures different from the bulk region. It seems
that in the surface region a surface reconstruction, due to incom-
pletely bonded atoms and impurities, increased the favor of tetrahe-
dral coordination for sample No UV and UV1. The decrease of
intensity in the low-binding-energy part for sample UV2 is attrib-
uted to the reduction of tetrahedral coordination in the near-surface
region. Because the radiation of UV2 decreased the oxygen content
in the surface significantly, the circumstance near surface region
may be altered, changing the stable structures into octahedral coor-
dination again.
For the differences in the clearness of the distinction between the
low and high-binding-energy part in the UVB spectra measured at
0°, the Al-O bond distance should be considered. The Al-O bond
distance strongly affects the binding energy of Al hybridization
bonding states, contributing to the high-binding-energy part in UVB.
As the Al-O bond distance decreases, the signals of high-binding-
energy part in UVB move toward higher binding energy. Thus,
peaked shape in the high-binding-energy part is broadened for the
films having various Al-O bond distances. It can be concluded that
sample UV2 had a relatively refined Al-O bond distance below the
surface, in consideration of the peaked shape of the high-binding-
energy part of UVB at the angle of 0°. It can be deduced that in the
UV2 condition the film was grown through refined reactions be-
tween TMA and strongly bonded hydroxyl groups due to the elimi-
nation of weakly bonded hydroxyl groups by UV radiation, but at
Figure 5. ͑a͒ Al 2p and ͑b͒ O 1s core-level spectra of sample No UV, UV1,
and UV2 deposited at 370°C for 2-nm-thick aluminum oxide on Si.
tions of the high- and low-binding-energy parts in the upper valence
band ͑UVB͒. High- and low-binding-energy parts were compared
with each other to investigate a structural change of aluminum ox-
ide. The spectrum of each sample, which had a 2-nm-thick alumi-
num oxide film on Si substrate, was obtained at two different emis-
sion angles ͑␪ ͒ of photoelectron 0 and 60°, which were the angles
e
between the direction normal to the film surface and the detector.
More structural information on the surface regions of films can be
obtained at ␪ = 60° rather than ␪ = 0°. The background spectrum
e
e
of the Si substrate was subtracted by the H-terminated Si spectrum.
The inelastic background was roughly indicated by the dotted lines
in the spectra. The films were deposited successively, kept in the
vacuum desiccator soon after the deposition process, and transferred
to the XPS chamber after half a day. Thus, the air exposure time and
conditions were similar. Typical spectra of crystalline Al2O3, in
which portion of an octahedral unit ͑AlO96−͒ is larger than a tetrahe-
dral unit ͑AlO54−͒, have two peaks in UVB distinctively divided into
low- and high-binding-energy parts. The low-binding-energy part
mainly represents the O 2p nonbonding state and the high-binding-
energy part the oxygen–aluminum hybridization level. The low-
binding-energy part generally has a smaller intensity than the high-
binding-energy part in crystalline Al2O3,20 but the low-binding-
energy part of the samples in this experiment had higher intensities
than the high-binding-energy part at ␪ of both 0 and 60°. At a ␪ of
e
e
60°the increase of intensity in the low-binding-energy part of UVB
for sample No UV and UV1 was larger than that of sample UV2. At
a ␪ of 0° sample UV2 showed clear distinction between the high-
e
and low-binding-energy part, in contrast to sample No UV and UV1.
The differences in VB spectra according to UV radiation are dis-
cussed later.
Figure 5a and b shows the XPS spectra of Al 2p and O 1s core
levels, measured at a ␪ of 60°, according to the UV exposure con-
e
dition for the films deposited at 370°C. The peak positions were
calibrated using the binding energy of the adventitious C 1s signal
͑285 eV͒. As shown in Fig. 5a, Al 2p core levels show a symmetric
shape in all samples and were fitted by a single component. The O
1s peaks in Fig. 5b could be deconvoluted with roughly two com-
ponents. The major component, having a lower binding energy, is
attributed to the Al–O–Al bond ͑O1͒. The other component ͑O2͒
may be due to H–O–Al, Si–O–Si, and carbon contamination con-
taining oxygen from air exposure. The binding energy difference
between the Al 2p and O1 components was 456.7 0.1 eV, which
agrees well with a report of Alexander et al.29 The ratio between O
from the O1 component and Al, estimated from the comparison of
the peak areas and the sensitivity factors, was 3.5, 3.4, and 1.6 for
sample No UV, UV1, and UV2, respectively, but the O/Al atomic
ratios, which include only the other component ͑O2͒ and Al 2p,
showed similar values in the range between 0.7 and 1. Because the
inspection of the Si 2p and C 1s spectra showed that the incorpora-
␪
of 60° the distinction between low- and high-binding-energy
e
parts disappeared because the contribution of modified Al–O bonds
with various bond distances from the rearrangement of atoms was
increased.
Figure 6 shows the estimation of the interface trap densities ͑Dit͒
distribution by the Terman method.32 Sample No UV deposited at
370°C showed an increased Dit on the lower part of the bandgap
between 0.2 and 0.4 eV from the Si VB edge compared to those
deposited at 260°C or the UV-radiated samples. A remarkable
difference between sample No UV deposited at 370°C and the
other samples is the carbon concentration near the interface. At a
Td of 370°C, the AES depth profiles showed that the atomic con-
Downloaded on 2015-05-07 to IP 128.122.253.212 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

F90
Journal of The Electrochemical Society, 153 ͑6͒ F87-F93 ͑2006͒
Figure 6. Estimation of Dit by the Terman method for sample No UV, UV1,
and UV2. The results labeled “Td of 260°C” were those of the nonradiated
sample. The other samples were prepared at a Td of 370°C.
Figure 8. Summary of the current density at −1 V vs CET according to Td,
UV radiation, and RTA. FGA was carried out at 450°C.
deposited films according to the UV exposure condition. A reduction
of the leakage current was observed in the high-field region when
UV was radiated during the ALD process for Td of both 260 and
370°C.
centration of carbon near the interface is 3.2 0.5, 1.6 0.2, and
1.1 0.2 atom % for sample No UV, UV1, and UV2, respectively.
Sample No UV deposited at 260°C showed a smaller carbon con-
tamination of 0.8 0.2 atom % than the samples deposited at
370°C. As mentioned above, the carbon atoms in a film deposited at
370°C are incorporated during the ALD reaction due to a slight
thermal decomposition of adsorbed precursor. It is believed that the
variation of Dit below midgap is due to the carbon contamination of
the interface. The UV radiation reduced Dit by suppressing the car-
bon incorporation near the interface for the films deposited at
370°C.
Figure 7a shows the thermal stability of the insulating properties
according to Td. Because the leakage current density strongly de-
pends on the film thickness in this range of capacitance equivalent
oxide thickness ͑CET͒, the significant difference of the leakage cur-
rent according to Td is simply due to the film thickness.7 After rapid
thermal annealing ͑RTA͒ at 920°C for 30 s in N2 ambient, the leak-
age current density of the sample deposited at 260°C increased sig-
nificantly, whereas the properties of the sample deposited at 370°C
were not largely degraded. Though surface roughening could in-
crease the leakage current density, the change of surface roughness
by RTA was ignorable in these experiments: the root-mean-square of
the surface roughness was hardly changed after RTA ͑ϳ0.12 nm in
13 ␮m2 for 5-nm-thick film after RTA͒. The reduction of the RTA
temperature from 920 to 780°C affected the electrical properties
significantly. The leakage current density in the low-field region was
not largely increased by the RTA at 780°C, even after 5-min anneal-
ing, while the leakage current density in the high-field region was
rather reduced. Figure 7b shows the leakage current density of the
Because the electrical properties in the lower CET region are
more important for capacitor applications, the electrical properties
of films having a CET below 2.5 nm were investigated according to
Td, UV radiation, and RTA treatment. Due to various factors such as
interfacial layer formation and Si diffusion into the film, the behav-
ior of insulating properties could be altered in this low-CET region.
The results were summarized by a current density at −1 V vs CET
plot in Fig. 8. For the comparison, the SiO2 results from Ref. 33 are
shown together. Because the SiO2 results were given by equivalent
oxide thickness ͑EOT͒, the EOT was converted to CET by adding
0.3 nm.34 Better insulating properties were observed than SiO2
films. The leakage current density of the films deposited at 370°C
showed better insulating properties than those deposited at 260°C.
After RTA at 920°C, the insulating property for sample UV1 depos-
ited at 370°C was superior to that of the nonradiated films.
The leakage current in the high-field region is well fitted by
Fowler-Nordheim ͑FN͒ tunneling as in other reports.35,36 Figure 9a
shows a FN tunneling plot for the samples deposited at 370°C ac-
cording to UV radiation. From the slope of the FN fitting, we can
estimate a fitting parameter, the product of the effective mass ratio
of an electron ͑m*/m0͒, and the cube of the effective barrier height
͑⌽3b͒.37 The steeper slope represents larger m*/m0⌽3b. The extracted
m*/m0⌽3b was 21.7, 24, and 23.3 for samples No UV, UV1, and
UV2. To estimate ⌽b from these slopes, the m* value should be
known. Various values of m* between 0.2 and 0.45m0 were reported
Figure 7. Variation of the leakage current
density vs bias voltage according to the
͑a͒ RTA treatment and ͑b͒ UV radiation
for the samples deposited at 260 and
370°C. RTA was performed at 920°C for
30 s in N2 ambient. RTA at 780°C was
done for 5 min.
Downloaded on 2015-05-07 to IP 128.122.253.212 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 153 ͑6͒ F87-F93 ͑2006͒
F91
Figure 9. ͑a͒ FN-tunneling plot showing J ͑current density͒/E2 ͑electric
field͒ vs 1/E according to UV radiation, and ͑b͒ band diagram when FN
tunneling begins to work.
for Al2O3 films.35,36,38-40 However, the obtained ⌽b from the slope is
strongly affected by a slight change of m*. Therefore, only the maxi-
mum probable value of ⌽b is estimated. Figure 9b shows a sche-
matic of the band diagram when FN tunneling dominates current
conduction. FN tunneling conduction begins to be observed in the
J-V plot when the voltage drop in the oxide ͉͑Vox͉͒ is larger than ⌽b.
In the FN tunneling fitting, the voltage for which the measured
current begins to be dominated by FN tunneling can be estimated
from the fitting results. If the direct tunneling current in the low-field
region was dominant over other conduction mechanisms, such as
Poole–Frenkel, Schottky emission, and trap-assisted tunneling, etc.,
͉Vox͉, where current conduction begins to follow FN tunneling, is
close to ⌽b. When another conduction mechanism dominates direct
tunneling in the low-field region, ͉Vox͉ is larger than ⌽b.
In Fig. 9a, the bias voltages at the Pt electrode ͑Vg͒ are indicated
from which the data begin to be fitted by FN tunneling. We esti-
mated Vox, which follows the relation Vox = Vg − ⌽ms − ⌽s. The
estimated voltage drop across the semiconductor ͑⌽s͒ is −0.34
0.04 V using the Berglund integration.41 The work function differ-
ence between metal and semiconductor ͑⌽ms͒ was estimated from
the extrapolation of the flatband voltage ͑Vfb͒ vs EOT plot.42,43 Note
that EOT, rather than the CET, has to be used here because the CET
values contain a quantum mechanical contribution from the Si
substrate.
Figure 10a shows EOT and physical thickness ͑tphys͒ vs cycle
number ͑ncy͒ of the ALD process at a Td of 370°C. Growth retar-
dation during the first three cycles was observed from the extra-
polation of the tphys vs ncy plot, which was similar to a reported
value of four cycles.27 The extrapolation of the EOT vs ncy plot to
the y axis shows that an interfacial layer was formed which has an
EOT of ϳ0.5 nm, although there is an initial growth retardation.
The inset figure of Fig. 10a shows the variation of EOT as a function
of tphys. From the inverse of the slope, the bulk k of the grown film
is estimated to be 8.3, which is a reasonable k value for an ALD
Al2O3 film.
Figure 11. The effective ⌽b obtained from FN tunneling fitting is shown
according to m*/m0 with estimated upper limits of ⌽b from Vox indicated by
a solid circle.
also observed in this experiment by the linearly increasing Vfb with
the increase in EOT, as shown in Fig. 10b. If the fixed charges were
distributed in the film, the plot would show nonlinear behavior. The
fixed charge density between Si and SiO2 was assumed to be negli-
gible. The estimated fixed charge density from the slope was
͑6.8 1.5͒ ϫ 1012 cm−2, which corresponds to other reports.43
For an arbitrary metal oxide semiconductor system with no dis-
tinguishable interfacial layer ͑interfacial EOT͒, the y axis intercept
of the Vfb vs EOT plot corresponds to ⌽ms. However, in this case,
⌽
should be the Vfb value at the EOT of the interfacial layer
ms
͑ϳ0.5 nm͒ because the fixed charges center at the interfacial layer/
Al2O3 interface. Therefore, a ⌽ms of 0.29 0.12 V was obtained in
this experiment, and Vox can be expressed by Vox = Vg + 0.05 V.
Figure 11 shows the variation of ⌽b as a function of m*/mo using the
m*/m0⌽3b values extracted from Fig. 9a for samples No UV, UV1,
and UV2 deposited at 370°C. The solid circles with indication of
errors in Fig. 11 are the Vox values when the current density began to
be fitted by FN tunneling conduction. Becaus Vox is larger than or
the same as ⌽b, depending on the conduction mechanism in the
low-voltage region, these values are considered as the upper bound
of ⌽b for each case. As shown in Fig. 11, UV radiation increased the
effective barrier height and thus reduced the leakage current. Con-
sidering the maximum probable value of ⌽b, m* is larger than
0.35m0. At an m* of 0.4m0, the extracted ⌽b values for samples No
UV, UV1, and UV2 are 3.78, 3.91, and 3.88 eV, respectively.
The insulating properties of the samples after RTA could not be
analyzed by the FN tunneling conduction due to the effects of traps.
Thus, changes of XPS spectra by UV radiation were investigated for
the RTA samples. Because the difference in leakage current density
between sample No UV and UV1 in Fig. 8 was evident after RTA,
It was reported that aluminum oxide has negative fixed charges
at the interface between SiO2 and aluminum oxide,43 and this was
Figure 10. ͑a͒ Variation in EOT and tphys
as a function of ncy and ͑b͒ Vfb vs EOT
plot of the film grown at a Td of 370°C.
The inset figure shows the variation of
EOT as a function of tphys
.
Downloaded on 2015-05-07 to IP 128.122.253.212 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

F92
Journal of The Electrochemical Society, 153 ͑6͒ F87-F93 ͑2006͒
estimated to be larger than 0.35m0. After RTA, the film prepared at
a Td of 370°C under UV radiation during the TMA purge step
showed superior insulating properties. VBO and bandgap of the
samples prepared at 370°C were measured by XPS after RTA and
FGA, and the calculated values of CBO with Si showed an increase
from 2.0 0.3 to 2.6 0.3 eV by the UV radiation. Because CBO
is closely related to the effective barrier height for the contact of
aluminum oxide with the Pt electrode, the smaller leakage current
density was attributed to the larger electron barrier height.
Acknowledgment
The work was supported by the Korean Ministry of Science and
Technology through the National Research Laboratories program
and the system IC 2010 program of the Korean government.
Figure 12. Comparison of ͑a͒ VB spectra and ͑b͒ O 1s loss spectra after
RTA at 920°C for sample No UV and UV1. The binding energy of the VB
spectrum is referenced to Si VBM.
Seoul National University assisted in meeting the publication costs of
this article.
References
1. P. S. Peercy, Nature (London), 406, 1023 ͑2000͒.
two samples were analyzed. Figure 12a shows the VB spectra of
10-nm-thick films deposited at 370°C measured by XPS after RTA
at 920°C and FGA at 400°C. VB spectra of sample No UV and
UV1 show a difference in VB shape. The binding energy was refer-
enced to the Si valence band maximum ͑VBM͒. For Si͑100͒, the
linear extrapolation method has been found to agree very well with
the method of Kraut et al.44 in which the experimentally measured
VB leading edge is fitted to a Gaussian broadened theoretical den-
sity of states. It was reported that the binding energy difference
between the VBM of Si and the Si 2p centroid was measured to be
98.95 0.04 eV45 by the Kraut method and 98.94 0.05 eV,46,47
averaged over n- and p-Si͑100͒, by the linear method. Therefore, the
position of Si VBM was set to 0 eV by a shift of the Si 2p centroid
to 98.95 eV. For the XPS measurement of 2-nm-thick aluminum
oxide on Si, the background VB spectrum from the Si substrate was
subtracted using the spectrum of H-terminated Si where the peak
position of Si 2p was aligned to 98.95 eV. The binding energy dif-
ferences between the Al 2p and Si 2p spectra from 2-nm-thick films
on Si and between the Al 2p spectra and aluminum oxide VBM,
measured by linear extrapolation, from 10-nm-thick films were ob-
tained. By aligning the Al 2p levels, the valence band offset ͑VBO͒
between aluminum oxide and Si was obtained.48 After this proce-
dure, the maximum intensity positions of the O 2s peak coincided
with each other at 24.2 eV. As shown in Fig. 12a, the VBO values
were measured to be 4.5 0.2 and 3.7 0.2 eV for film No UV
and UV1, respectively.
2. A. I. Kingon, J.-P. Maria, and S. K. Streiffer, Nature (London), 406, 1032 ͑2000͒.
3. International Technology Roadmap for Semiconductors, http://public.itrs.net/
4. J. H. Lee, K. Koh, N. I. Lee, M. H. Cho, Y. K. Kim, J. S. Jeon, K. H. Cho, H. S.
Shin, M. H. Kim, K. Fujihara, H. K. Kang, and J. T. Moon, Tech. Dig. - Int.
Electron Devices Meet., 2000, 645.
5. D. A. Buchanan, E. P. Gusev, E. Cartier, H. Okorn-Schmidt, K. Rim, M. A. Gri-
belyuk, A. Mocuta, A. Ajmera, M. Copel, S. Guha, N. Bojarczuk, A. Callegari, C.
D’Emic, P. Kozlowski, K. Chan, R. J. Fleming, P. C. Jamison, J. Brown, and R.
Arndt, Tech. Dig. - Int. Electron Devices Meet., 2000, 223.
6. Y. Fujisaki, K. Iseki, H. Ishiwara, M. Mao, and R. Bubber, Appl. Phys. Lett., 82,
3931 ͑2003͒.
7. I.-S. Park, B. T. Lee, S. J. Choi, J. S. Im, S. H. Lee, K. Y. Park, J. W. Lee, Y. W.
Hyung, Y. K. Kim, H. S. Park, Y. W. Park, S. I. Lee, and M. Y. Lee, VLSI Tech-
nology Digest, p. 42 ͑2000͒.
8. H. Seidl, M. Gutsche, U. Schroeder, A. Birner, T. Hecht, S. Jakshik, J. Luetzen, M.
Kerber, S. Kudelka, T. Popp, A. Orth, H. Reisinger, A. Saenger, K. Schupke, and
B. Sell, Tech. Dig. - Int. Electron Devices Meet., 2002, 839.
9. F. Fishburn, R. Kauffman, R. Lane, T. McDaniel, K. Schofield, S. Southwick, R.
Turi, and H. Wang, VLSI Technology Digest, p. 75 ͑2003͒.
10. M. Cho, H. B. Park, J. Park, C. S. Hwang, J.-C. Lee, S.-J. Oh, J. Jeong, K. S. Hyun,
H.-S. Kang, Y.-W. Kim, and J.-H. Lee, J. Appl. Phys., 94, 2563 ͑2003͒.
11. J. Lee, Y. Ahn, Y. Park, M. Kim, D. Lee, K. Lee, C. Cho, T. Chung, and K. Kim,
VLSI Technology Digest, p. 57 ͑2003͒.
12. X.-B. Lu, Z.-G. Liu, Y.-P. Wang, Y. Yang, X.-P. Wang, H.-W. Zhou, and B.-Y.
Nguyen, J. Appl. Phys., 94, 1229 ͑2003͒.
13. R. A. B. Devine, J. Appl. Phys., 93, 9938 ͑2003͒.
14. M. Ishida, K. Sawada, S. Yamaguchi, T. Nakamura, and T. Suzaki, Appl. Phys.
Lett., 55, 56 ͑1989͒.
15. E. Fredriksson and J. O. Carlson, Chem. Vap. Deposition, 1, 333 ͑1993͒.
16. C. J. Kang, J. S. Chun, and W. J. Lee, Thin Solid Films, 189, 161 ͑1990͒.
17. S. Zhu, F. Wang, H. Lou, and W. Wu, Surf. Coat. Technol., 71, 9 ͑1995͒.
18. S. K. Kim and C. S. Hwang, J. Appl. Phys., 96, 2323 ͑2004͒.
19. A. W. Ott, K. C. McCarly, J. W. Klaus, J. D. Way, and S. M. George, Appl. Surf.
Sci., 107, 128 ͑1996͒.
20. S. Thomas and P. M. A. Sherwood, Anal. Chem., 64, 2488 ͑1992͒.
21. D. R. Biswas, C. Ghosh, and R. L. Layman, J. Electrochem. Soc., 130, 234 ͑1983͒.
22. Y. Zhang and M. Stuke, Jpn. J. Appl. Phys., Part 2, 27, L1349 ͑1988͒.
23. I. P. Herman, Chem. Rev. (Washington, D.C.), 89, 1323 ͑1989͒.
24. G. S. Higashi and L. J. Rothberg, Appl. Phys. Lett., 47, 1288 ͑1985͒.
25. M. A. Henderson, Surf. Sci. Rep., 46, 1 ͑2002͒.
The bandgap estimated from the O 1s loss spectra is shown in
Fig. 12b. Reported values of the Al2O3 bandgap are in the wide
range of 5.4–8.8 eV.49-51 A bandgap of 7.6 0.1 and 7.4 0.1 eV
was obtained for sample No UV and UV1, respectively. From VBO
and bandgap, the conduction band offset ͑CBO͒ between aluminum
oxide and Si can be easily calculated. After RTA the film deposited
under UV1 condition has a higher CBO of 2.6 0.3 eV compared
to sample No UV, showing a CBO of 2.0 0.3 eV. Thus, better
insulation properties of sample UV1 after RTA is attributed to higher
potential barrier height with the Pt electrode caused by larger CBO.
26. R. F. Klie, N. D. Browning, A. Roy Chowdhuri, and C. G. Takoudis, Appl. Phys.
Lett., 83, 1187 ͑2003͒.
27. L. G. Gosset, J.-F. Damlencourt, O. Renault, D. Rouchon, Ph. Holliger, A. Ermo-
lieff, I. Trimaille, J.-J. Ganem, F. Martin, and M.-N. Séméria, J. Non-Cryst. Solids,
303, 17 ͑2002͒.
28. T. R. Gow, R. Lin, L. A. Cadwell, F. Lee, A. L. Backman, and R. I. Masel, Chem.
Conclusions
Mater., 1, 406 ͑1989͒.
Aluminum oxide thin films were deposited by ALD using TMA
and water under UV photon radiation. The films were prepared at
two different deposition temperatures ͑Td͒ of 260 and 370°C,
and the films deposited at a Td of 370°C showed a higher O/Al
atomic ratio than those grown at a Td of 260°C. VB spectra showed
a change in the upper valence band structure by UV radiation. At a
Td of 370°C, carbon incorporation caused by the decomposition of
adsorbed precursors was decreased by UV radiation during the ALD
process. The interface containing carbon atoms showed an increased
interface trap density below the midgap. The leakage current con-
duction in the high-field region was well fitted by FN tunneling
before RTA. At a Td of 370°C, the electron effective mass was
29. M. R. Alexander, G. E. Thompson, and G. Beamson, Surf. Interface Anal., 29, 468
͑2000͒.
30. M. R. Alexander, S. Payan, and T. M. Duc, Surf. Interface Anal., 26, 960 ͑1998͒.
31. B. C. Lippens and J. J. Steggerda, Physical and Chemical Aspects of Adsorbents
and Catalysts, Chap. 4, Academic Press, New York ͑1970͒.
32. L. M. Terman, Solid-State Electron., 5, 285 ͑1962͒.
33. M. Koh, W. Mizubayashi, K. Iwamoto, H. Murakami, T. Ono, M. Tsuno, T. Mi-
hara, K. Shibahara, S. Miyazaki, and M. Hirose, IEEE Trans. Electron Devices, 48,
259 ͑2001͒.
34. Y. H. Wu, M. Y. Yang, A. Chin, W. J. Chen, and C. M. Kwei, IEEE Electron
Device Lett., 21, 341 ͑2000͒.
35. M. D. Groner, J. W. Elam, F. H. Fabreguette, and S. M. George, Thin Solid Films,
413, 186 ͑2002͒.
36. S. Meng, C. Basceri, B. W. Busch, G. Derderian, and G. Sandhu, Appl. Phys. Lett.,
83, 4429 ͑2003͒.
Downloaded on 2015-05-07 to IP 128.122.253.212 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 153 ͑6͒ F87-F93 ͑2006͒
F93
37. D. K. Schröder, Semiconductor Material and Device Characterization, John Wiley
& Sons, New York ͑1988͒.
45. E. T. Yu, E. T. Croke, T. C. McGill, and R. H. Miles, Appl. Phys. Lett., 56, 569
͑1990͒.
38. Q. Q. Shu and W. G. Ma, Appl. Phys. Lett., 61, 2542 ͑1992͒.
39. M. Specht, M. Städele, S. Jakschik, and U. Schröder, Appl. Phys. Lett., 84, 3076
͑2004͒.
46. S. A. Chambers, T. Liang, Z. Yu, R. Droopad, J. Ramdani, and K. Eisenbeiser,
Appl. Phys. Lett., 77, 1662 ͑2000͒.
47. S. A. Chambers, T. Liang, Z. Yu, R. Droopad, and J. Ramdani, J. Vac. Sci. Technol.
40. Y. Yourdshahyan, C. Ruberto, L. Bengtsson, and B. I. Lundqvist, Phys. Rev. B, 56,
8553 ͑1997͒.
41. C. N. Berglund, IEEE Trans. Electron Devices, 13, 701 ͑1966͒.
42. J. K. Schaeffer, L. R. C. Fonseca, S. B. Samavedam, Y. Liang, P. J. Tobin, and B.
E. White, Appl. Phys. Lett., 85, 1826 ͑2004͒.
A, 19, 934 ͑2001͒.
48. L. F. Edge, D. G. Schlom, S. A. Chambers, E. Cicerrella, J. L. Freeouf, B.
Holländer, and J. Schubert, Appl. Phys. Lett., 84, 726 ͑2004͒.
49. K. S. Shamala, L. C. S. Murthy, and K. Narasimha Rao, Mater. Sci. Eng., B, 106,
269 ͑2004͒.
43. R. S. Johnson, G. Lucovsky, and I. Baumvol, J. Vac. Sci. Technol. A, 19, 1353
͑2001͒.
44. E. A. Kraut, R. W. Grant, J. R. Waldrop, and S. P. Kowalczyk, Phys. Rev. Lett., 44,
1620 ͑1980͒.
50. R. H. French, J. Am. Ceram. Soc., 73, 477 ͑1990͒.
51. H. Nohira, W. Tsai, W. Besling, E. Young, J. Petry, T. Conard, W. Vandervorst, S.
De Gendt, M. Heyns, J. Maes, and M. Tuominen, J. Non-Cryst. Solids, 303, 83
͑2002͒.
Downloaded on 2015-05-07 to IP 128.122.253.212 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).