ALD and characterization of aluminum oxide deposited on Si(100) using tris(diethylamino) aluminum and water vapor

Article abstract of DOI:10.1149/1.2239258

A nonpyrophoric, oxygen-free, halogen-free tris(diethylamino) aluminum (TDEAA) precursor was used for atomic layer deposition (ALD) of aluminum oxide on Si(100). ALD of aluminum oxide using TDEAA and water was found to be self-limiting with respect to both reactants. The temperature window for ALD in the hotwall reactor used was found to be between 200 and 400°C. The ALD rate was 1.4 A/cycle at optimum conditions. Fourier transform infrared (FTIR) analyses indicated negligible interfacial SiO₂ growth during deposition. Both FTIR spectra and transmission electron micrographs showed the ALD aluminum oxide to be amorphous. Also, FTIR and X-ray photoelectron spectral (XPS) analyses showed negligible carbon and nitrogen (< 1% atomic) contamination in the film. Z-contrast images and electron energy loss spectra showed uniform aluminum oxide film with an abrupt interface with Si. XPS analysis revealed aluminum oxide film to be stoichiometric with no detectable Al-Al cluster formation. Also, XP spectra showed no silicate formation at the interface of as-deposited alumina films.

Full text of DOI:10.1149/1.2239258

Journal of The Electrochemical Society, 153 ͑10͒ C701-C706 ͑2006͒
C701
0013-4651/2006/153͑10͒/C701/6/$20.00 © The Electrochemical Society
ALD and Characterization of Aluminum Oxide Deposited on
Si„100… using Tris(diethylamino) Aluminum and Water
Vapor
a
a
Rajesh Katamreddy,a,b Ronald Inman,a, Gregory Jursich, Axel Soulet, and
*
Christos Takoudisb,
*
,z
aAmerican Air Liquide, Chicago Research Center, Countryside, Illinois 60525, USA
bDepartments of Chemical Engineering and Bioengineering, University of Illinois at Chicago, Chicago,
Illinois 60607, USA
A nonpyrophoric, oxygen-free, halogen-free tris͑diethylamino͒ aluminum ͑TDEAA͒ precursor was used for atomic layer deposi-
tion ͑ALD͒ of aluminum oxide on Si͑100͒. ALD of aluminum oxide using TDEAA and water was found to be self-limiting with
respect to both reactants. The temperature window for ALD in the hotwall reactor used was found to be between 200 and 400°C.
The ALD rate was 1.4 Å/cycle at optimum conditions. Fourier transform infrared ͑FTIR͒ analyses indicated negligible interfacial
SiO2 growth during deposition. Both FTIR spectra and transmission electron micrographs showed the ALD aluminum oxide to be
amorphous. Also, FTIR and X-ray photoelectron spectral ͑XPS͒ analyses showed negligible carbon and nitrogen
͑Ͻ1% atomic͒ contamination in the film. Z-contrast images and electron energy loss spectra showed uniform aluminum oxide film
with an abrupt interface with Si. XPS analysis revealed aluminum oxide film to be stoichiometric with no detectable Al–Al cluster
formation. Also, XP spectra showed no silicate formation at the interface of as-deposited alumina films.
© 2006 The Electrochemical Society. ͓DOI: 10.1149/1.2239258͔ All rights reserved.
Manuscript submitted March 14, 2006; revised manuscript received May 8, 2006. Available electronically August 4, 2006.
In future minimization of transistor devices, alternative high-
dielectric-constant materials are needed to replace SiO2 and its first-
generation replacement material, silicon oxynitride.1 Among the po-
tential replacement materials, aluminum oxide has gained
considerable interest. Aluminum oxide has many favorable proper-
ties like high bandgap2,3 and thermal stability on Si4 and it remains
amorphous after high-temperature annealing for future silica
replacement. However, it has a relatively low dielectric constant
ever, this compound polymerizes easily and thus exists as a mixture
of polymers or oligomers. Each polymer or oligomer has a different
vapor pressure. As a result, the vapor pressure of this precursor is
unpredictable and difficult to control. Aluminum 2-ethyhexanoate
has also been demonstrated as a precursor for Al2O3 deposition, but
its low vapor pressure results in low deposition rates, which limits
its usefulness.17 In addition, many of the nonpyrophoric candidates
contain aluminum–oxygen ͑Al–O͒ bonds. These bonds are quite
strong ͑122 kcal/mol͒ compared to aluminum–nitrogen ͑Al–N͒
bonds ͑71 kcal/mol͒, which gives alkyl amino aluminum precursors
a clear advantage in the energetics of reaction.
The advantage in using higher alkyl amino aluminum com-
pounds, such as diethyl or higher carbon chain amino aluminum
compounds, rather than methyl amino aluminum compounds, is that
the higher alkyl amino aluminum compounds are less susceptible to
self degradation by oligomerization. Another advantage of using di-
ethyl amino compounds is that, unlike methyl amino aluminum
compounds, the diethyl amino aluminum compound exists as a liq-
uid. Thus, it is easier to control and provides more consistent deliv-
ery to the deposition tool. In this work, we present a nonpyrophoric,
halogen-free, oxygen-free precursor, tris͑diethylamino͒ aluminum
͑TDEAA; U.S. Pat. Appl. 20050003662͒, which along with water is
used for ALD of aluminum oxide. Self-limiting kinetics for both
precursor and water half-cycles are shown. Effects of temperature
on deposition rate and on the vapor pressure of the novel precursor
are also presented. Further, aluminum oxide film and its interface
with Si are analyzed using Fourier transform infrared spectroscopy
͑FTIR͒, X-ray photoelectron spectroscopy ͑XPS͒, and scanning
transmission electron microscopy ͑STEM͒/electron energy loss
spectroscopy ͑EELS͒ techniques.
5
͑ϳ9͒ and this makes it a short-term replacement material. In order
to take advantage of the material benefits of aluminum oxide with-
out the electrical disadvantage of low dielectric constant, a combi-
nation film of aluminum oxide with higher dielectric metal oxides is
potentially a better, longer term replacement material. Of possible
combinations with aluminum oxide, hafnium oxide and zirconium
oxide are currently being investigated.2,6-8 Some recent studies have
reported improvement in crystallization temperature and interfacial
stability by mixing aluminum oxide with hafnium oxide.9,10 It has
been widely assumed that a thin silicon oxide interfacial layer is
required for excellent electrical properties, but the presence of a
silicon oxide interfacial layer places another constraint on the physi-
cal thickness. Aluminum oxide films also have a significant market
in capacitor and some nonintegrated circuit applications like mag-
netic heads.
Atomic layer deposition ͑ALD͒ is being increasingly used to
grow thin films in the microelectronics industry. The ALD process
involves a sequential alternating series of gas-surface reaction under
surface saturation conditions. Precursor and oxidizer are pulsed, al-
ternately separated by inert gas purging periods to prevent gas phase
reactions and obtain self-limiting film deposition. Each cycle results
in a submonolayer growth of the required film. ALD has attractive
features like excellent uniformity, conformality, and submonolayer
thickness control.11
ALD of Al2O3 films is commonly performed today using alkyl
aluminum or alkyl aluminum hydride as Al source.12-15 One of the
disadvantages of alkyl aluminum precursors such as tri-methyl alu-
minum ͑TMA͒ is that they are pyrophoric, which requires special
handling and storage precautions. In addition to pyrophoricity, TMA
contains aluminum–carbon ͑Al–C͒ bonds, which may result in un-
desirable carbon incorporation in the film. Aluminum isopropoxide
has also been used as precursor to deposit aluminum oxide.16 How-
Experimental
Aluminum oxide thin films were deposited using tris͑diethy-
lamino͒ aluminum precursor and water as oxidizing agent on
p-doped 1 ϫ 1 in. Si͑100͒ substrates in a hot wall tube reactor with
a resistive heater coil. Prior to deposition, Si wafers were cleaned by
RCA standard cleaning ͑SC-1͒ to remove organic contaminants and
particulates, followed by 1% HF dip for 40 s. Each of the two steps
was followed by thorough deionized ͑DI͒ water rinse and drying in
N2. This cleaning procedure left approximately 10 Å of native oxide
on the surface.
The reactor consists of a high-purity quartz tube ͑38 mm diam
ϫ 480 mm long͒ with an aluminum block serving as a substrate
*
Electrochemical Society Active Member.
z E-mail: takoudis@uic.edu
Downloaded on 2015-06-23 to IP 131.215.225.9 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

C702
Journal of The Electrochemical Society, 153 ͑10͒ C701-C706 ͑2006͒
Figure 1. Schematic of the ALD reactor
used for aluminum oxide deposition on
Si͑100͒.
holder, as shown in Fig. 1. Silicon substrate sits in the groove on the
aluminum block with the deposition surface being perpendicular to
reactant flow. The reactor can be heated to 650°C, although typi-
cally depositions are done at 150–450°C. The temperature of the
heated zone during deposition ͑and called “reactor temperature”
from now on͒ was measured using a type-K thermocouple glued to
the outside surface of the quartz tube and controlled using a propor-
tional integral differential ͑PID͒ controller ͑Omega CN 9000͒. PID
settings were optimized so that the temperature control was obtained
within 1°C of the set point. The temperature uniformity of the
heated zone was measured by inserting a thin type-K thermocouple
probe along the cylindrical axis of the reactor and measuring tem-
perature with this probe as it was positioned at different distances
along the zone. Within the length of heater coil, the thermocouple
measurements showed less than 2% deviation over a 10 cm dis-
tance. The temperature measured inside the reactor was about 25°C
lower than the temperature measured on the outside surface of the
quartz tube in the 200–450°C range. The inside temperature likely
represents the deposition surface temperature.
The ALD precursor was introduced into the reactor by pulsing
vaporized precursor using nitrogen carrier gas. The precursor vessel
was heated in a copper block and the temperature was controlled
using a PID controller. The precursor was introduced using a simple
pulsing sequence. First, the precursor container and manifold leg
were filled with carrier gas typically for 1 s and then the carrier gas
line was closed using the pneumatic valve VN ͑Fig. 1͒ to empty the
vaporized precursor diluted in carrier gas into 2the reactor. This cycle
of filling carrier gas in the precursor vessel and subsequent evacua-
tion into the reactor is called a precursor “plug.” We repeated this
plug to inject the total amount of precursor needed for an ALD cycle
͑typically in an ALD reactor, the amount of precursor injected is
controlled by changing the pulse time͒. The use of plugs enhances
mass transport by increasing the pressure gradient between the pre-
cursor vessel and the reactor. A similar kind of precursor injection
has been reported earlier by Becker et al.18 This step is followed by
a nitrogen purge to remove physisorbed precursor. The temperature
of the precursor manifold was independently controlled typically
10–15°C above the reservoir temperature in order to prevent precur-
sor condensation. The moisture reagent was introduced into the re-
actor as a diluted mixture with nitrogen carrier gas. The mixture
originated from a separate manifold as indicated in Fig. 1. The
source for moisture was ultrapure deionized water which was deliv-
ered from a bubbler maintained near 0°C using an ice bath and
nitrogen carrier gas. The moisture pulse was then followed by a
nitrogen purge to remove physisorbed water. This cycle was re-
peated as many times as desired to grow thicker films.
controlled. The precursor vessel can be cooled to as low as 77 K
using a liquid nitrogen bath or heated with a resistive band heater in
conjunction with a PID temperature controller. The combination of
both heating and cooling was used to induce freeze–pump–thaw
cycles, thereby discriminating between true precursor vapor pressure
and artificially high vapor pressure from possible liquid degassing
and degradation by-products.
In this study film thicknesses were measured using spectral el-
lipsometry ͑J. A. Woollam Co., Inc., model M-44͒ and reflectometry
͑Mission Peak Optics, Inc., MP100-S͒. FTIR spectroscopy ͑Nicolet,
Magna-IR 560͒ was used in the normal transmission mode over the
wave number range of 4000–400 cm−1. Spectra of cleaned wafers
prior to deposition were used for background subtraction of the
spectra of postdeposition wafers. Such subtraction spectra provide
information about the deposited film and changes at the interface
during deposition without the spectral features of the substrate itself.
A JEOL 2010F TEM/STEM was used which has a Schottky field
emission gun source and operated at 200 kV with a JEOL annular
dark-field detector and postcolumn Gatan imaging filter. The Gatan
imaging filter was used in spectroscopy mode to collect EEL
spectra.19 The lens conditions in the microscope were defined for a
probe size of 0.2 nm, with a convergence angle of 13 mrad and a
collection angle of 52 mrad. As the two techniques do not interfere,
Z-contrast images could be used to position the electron probe at the
desired spot in the sample to acquire EEL spectra.20,21 The accumu-
lation time of each spectrum acquisition was 0.1 s for low-loss spec-
trum and 5.0 s for core-loss spectrum. Energy loss spectra at 0–
100 eV ͑low loss͒ were measured to give information about the
valence electrons of the specimen. Along with these low-loss spec-
tra, O K-edge spectra in the 550–600 eV range were also measured
to examine other core electron states within the specimen.
The Kratos AXIS-165 surface analysis system equipped with a
monochromatic Al K␣ and a concentric hemispherical analyzer
coupled with a charge neutralizer was used for XPS analysis of
4.5 nm aluminum oxide on Si substrate. The analyzer was operated
at pass energy of 40 eV to record Al 2p and Si 2p core-level spectra
and the takeoff angle was 90°.
Results and Discussion
TDEAA vapor pressure dependence on temperature.— One of
the most critical thermal parameters of precursors is its vapor pres-
sure as a function of temperature. Sufficient precursor vapor density
is needed to allow adequate deposition rates and yet care is needed
in not overheating precursor beyond the temperature of self-
decomposition. Thus, from knowledge of thermal temperature limit
of the precursor prior to decomposition and its vapor pressure, one
can assess the maximum range of permissible precursor reservoir
temperature for the process with a given precursor.
The vapor pressure of the TDEAA was measured directly using a
heated vacuum manifold equipped with a heated capacitance ma-
nometer. The pressure gauge and manifold were both temperature
Downloaded on 2015-06-23 to IP 131.215.225.9 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 ͑10͒ C701-C706 ͑2006͒
C703
Figure 4. Deposition rate ͑Å/cycle͒ of aluminum oxide on Si͑100͒ for five
plugs vs reactor temperature. Error bars indicate film uniformity across the
wafer sample.
Figure 2. TDEAA vapor pressure vs temperature ͑semilog plot͒ showing
exponential increase in vapor pressure with temperature.
Temperature.— The deposition rate dependence on reactor tem-
perature is shown in Figure 4 with error bars indicating film unifor-
mity. The temperature of the reactor is measured at the outside sur-
face of the quartz tube within the middle of the heated zone. This
data shows the wide temperature window for ALD of aluminum
oxide with TDEAA from about 200 to 400°C with no significant
change in deposition rate. The film uniformity seems to improve
with temperature in this ALD temperature regime. There is a sharp
increase in deposition rate at 450°C, which most likely is due to
precursor decomposition. In all other data presented here, alumina
was deposited at 400°C.
In the vapor pressure measurements, the precursor vessel is
heated to the required temperature and the vapor pressure is mea-
sured from an evacuated heated manifold. The measurements are
considered reliable only after the pressure in the manifold is stable
over time. For each temperature, the measurements are repeated to
make sure that results are reproducible and to verify that decompo-
sition by-products are not influencing the vapor pressure measure-
ments. In Fig. 2, the TDEAA precursor vapor pressure is given as a
function of temperature ͑semi-log plot͒. TDEAA vapor pressure ap-
pears to have exponential dependence on temperature with highest
rate of increase between 55 and 105°C. For our ALD of aluminum
oxide with TDEAA precursor, the temperature of the precursor ves-
sel was maintained at 105°C.
Inert gas purge after the moisture pulse.— In this work, the
moisture pulse length was set to 0.05 s, which for ALD represents
the shortest pulse time of the computer-controlled manifold system.
However, due to the response time of the pneumatic valves, the
pulse time is longer than this as there is no difference in deposition
growth as moisture pulse is adjusted from 0.05 to 0.8 s. Therefore,
in this study excess physisorbed moisture was removed by increas-
ing the nitrogen purge length following the moisture pulse in order
to obtain ALD. Figure 5 shows the effect of inert gas purge length
following the moisture pulse on the growth rate of ALD alumina.
Alumina growth rate decreases with increasing nitrogen purge
length until the nitrogen purge duration levels off at about 40 s. This
indicates that the 0.05 s moisture pulse saturates the surface with
moisture ͑probably through hydroxylation͒, and physisorbed mois-
ture along with excess gaseous and surface reaction by-products is
ALD aluminum oxide: Precursor plugs.— Figure 3 shows the
effect of the number of precursor plugs on the growth rate of alu-
minum oxide with error bars representing variation of thickness
measurements across the substrate. The graph is similar to a typical
ALD precursor pulse effect on growth rate which indicates that the
discontinuous ͑plugs͒ method of precursor delivery works as well as
continuous precursor injection for ALD.12,13,18 From Figure 3, the
growth rate seems to level off to ϳ1.4 Å/cycle at about five plugs of
precursor with no variation of uniformity within the ALD regime
where the deposition surface is saturated with precursor molecules.
Further increase in number of plugs does not affect the growth rate
as the excess physisorbed precursor is removed during the subse-
quent purge step. The optimal length of subsequent nitrogen purging
was determined to be 5 s.
Figure 5. Growth rate ͑Å/cycle͒ of aluminum oxide on Si͑100͒ at 400°C
reactor temperature, and five plugs vs inert gas purge length following a
moisture pulse 0.05 s long. The alumina growth rate appears to saturate to
1.4 Å/cycle at about 40 s inert gas purge.
Figure 3. Growth rate ͑Å/cycle͒ of aluminum oxide on Si͑100͒ vs number
of precursor plugs; the reactor temperature is 400°C. Alumina growth rate
appears to saturate to 1.4 Å/cycle at about five plugs.
Downloaded on 2015-06-23 to IP 131.215.225.9 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

C704
Journal of The Electrochemical Society, 153 ͑10͒ C701-C706 ͑2006͒
Figure 6. FTIR spectrum of 16 nm thick as-deposited alumina film on Si
͑with ϳ1 nm thick silicon oxide after the standard RCA cleaning͒. The inset
shows an exploded view of the 400–1200 cm−1 spectral region. Other ALD
conditions: 400°C reactor temperature, five plugs of TDEAA precursor fol-
lowed by 5 s of nitrogen purge, and then 0.05 s of moisture pulse followed
by 40 s of nitrogen purge.
Figure 7. TEM micrograph of as-deposited alumina ͑5.2 nm thick͒ on Si
deposited under the same conditions as in Fig. 6.
removed after about 40 s of nitrogen purging. The ALD rate at these
conditions is about 1.4 Å/cycle. Additional experiments were done
with the water pulse set to zero, while everything else was kept the
same in our ALD system; in these cases, no deposition was found on
the substrate, even after 200–300 pulses of the TDEAA precursor.
Unless stated otherwise, the typical ALD conditions used in this
work are five plugs of TDEAA precursor followed by 5 s of nitro-
gen purging and then 0.05 s of moisture pulse followed by 40 s of
nitrogen purge; the ALD temperature is 400°C and the system pres-
sure is 250 mTorr ͑at the end of the precursor plugs it is lower͒.
tection limit of the XPS and FTIR spectrometers ͑i.e., Ͻ1 atom %͒
in the film. Features due to hydroxyl groups at 3400 cm−1 are also
absent.23 Peaks at 640 and 710 cm−1 correspond to the O–Al–O
bending mode and the Al–O stretching mode, respectively.24-26 The
apparent absence of peak atϳ530 cm−1 due to Al–O stretching in
condensed AlO6 octahedra indicates an amorphous structure of the
film.27 An absorbance peak due to Al = O at 1345 cm−1 is also
absent.24 The small peak observed at ϳ600 cm−1 is due to silicon.
STEM/EELS studies.— Figure 7 shows a TEM micrograph of the
as-deposited aluminum oxide film on RCA-cleaned Si substrate. The
aluminum oxide film appears to be amorphous in the TEM image,
which is consistent with the FTIR analysis above, and the Si͑100͒ is
easily identified by the crystalline fringes. There is a potential chal-
lenge in using TEM micrographs to define the interface of the two
amorphous layers, i.e., Al2O3 and SiO2. These two layers appear in
different contrast in the micrograph, with silica having the darker
contrast. One may expect that the difference in contrast can be used
to identify thickness. However, the apparent thickness of the inter-
Film and interface characterization: FTIR results.— Figure
6
shows FTIR spectra of aluminum oxide film in the absorbance
mode. The fundamental vibrations of solids are localized in the low-
frequency region of the IR spectrum ͑Ͻ1200 cm−1͒.22 Absence of
any Si–O growth during deposition is evident from the lack of
Si–O–Si symmetric absorption peak at 1070 cm−1. Lack of any hy-
drocarbon absorption peaks at ϳ3000 cm−1 and 1600 cm−1 indi-
cates negligible organic residue present in as-deposited films. FTIR
and XPS analyses show that carbon contamination is below the de-
Figure 8. ͑a͒ Z-contrast ͑high-angle ADF͒ image of alumina ͑ϳ5.2 nm͒ on Si͑100͒ substrate deposited at the same conditions as in Fig. 6; ͑b͒ low-loss EEL
and ͑c͒ O K-edge spectra of silicon substrate, interfacial layer, and aluminum oxide.
Downloaded on 2015-06-23 to IP 131.215.225.9 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 ͑10͒ C701-C706 ͑2006͒
C705
Figure 9. ͑a͒ Al 2p and ͑b͒ Si 2p core-
level spectra of
a 4.5 nm thick as-
deposited alumina film on Si at 400°C.
facial silica layer changes with change of focus probe and it is
difficult to judge the right contrast for defining the true interface.
A more reliable way of determining the film and interfacial layer
thickness is to use the Z-contrast image in which films with different
atomic number appear in different contrast. For thin samples, the
image contrast in high-angled annular dark field ͑HAADF͒ is
roughly proportional to Z2 ͑atomic number͒.28 This gives the
Z-contrast image good sensitivity even for elements with close
atomic numbers like Al ͑13͒ and Si ͑12͒. In STEM, the atomic-
resolution HAADF image is formed by scanning a focused electron
beam ͑a few angstroms in diameter͒ and collecting the electrons that
are elastically scattered to large angles ͑Ͼ50 mrad͒ on an annular
dark field ͑ADF͒ detector. Figure 8a shows Z-contrast ͑high-angle
dark field͒ image of an Al2O3/interfacial layer/Si stack. From these
measurements, the thickness of the alumina layer is determined to
be 52 2 Å and that of the interfacial layer to be 10 2 Å.
Figure 8b shows the EEL spectra in the low-loss region
͑0–100 eV͒ of the three distinct regions in the STEM image. EEL
spectra in the low-loss region give information about valence elec-
trons of species and it is referred to as valence EELS ͑VEELS͒. The
bottom spectrum of Figure 8b has a peak at 23 eV which is attrib-
uted to aluminum oxide.29,30 The top spectrum has the characteristic
Si signal at 18 eV and a second plasmon at about 35 eV.31,32 The
middle spectrum ͑Figure 8b͒ corresponds to the interfacial layer and
it has a broad feature at 22 eV. This peak has been attributed to
silicon dioxide.31,32 FTIR analysis revealed no silica growth during
aluminum oxide deposition. Therefore, the silica at the interface is
the residual native oxide from the standard RCA cleaning procedure
used.
Si 2p1/2 ͑99.8 eV͒ features because of the high-energy-resolution
capability.35 There is also a smaller higher binding energy peak at
about 102.9 eV. This component at the higher energy results from a
higher oxidation chemical state of Si. The separation between bulk
Si and Si oxidation state is 3.6–4.0 eV, which suggests the Si+4
oxidation state.33 This indicates the interfacial layer to mainly con-
sist of silicon dioxide. No silicate formation is evidenced.33
Conclusions
In this work we introduce a novel, nonpyrophoric, oxygen-free,
halogen-free TDEAA precursor for aluminum oxide deposition. We
observe the vapor pressure of precursor to increase exponentially
with temperature. ALD of aluminum oxide using TDEAA and water
was observed to be self-limiting with respect to both reactants with
an ALD rate of 1.4 Å/cycle. The temperature window for ALD in
the hotwall reactor was found to be between 200 and 400°C. Uni-
formity was observed to be increasing with temperature even within
the ALD regime. At higher temperatures above 400°C there was a
significant increase in the growth rate which was attributed to pre-
cursor decomposition. FTIR analysis indicates negligible interfacial
SiO2 growth during deposition. Both FTIR spectra and a TEM mi-
crograph show the ALD aluminum oxide to be amorphous. Also,
FTIR analyses show no indication of carbon and nitrogen contami-
nation in the film. Z-contrast images and EEL spectra show uniform
aluminum oxide film with an abrupt interface with Si. XPS analysis
reveals stoichiometric aluminum oxide film on stoichiometric sili-
con dioxide/Si substrate with no Al–Al cluster formation or silicate
formation at the interface of the as-deposited films.
Acknowledgments
Figure 8c shows O K-edge EEL spectra of the three distinct
regions in the STEM image. The top spectrum of Fig. 8c corre-
sponds to the silicon substrate and it has no O K-edge features. The
middle spectrum of Fig. 8c from the interfacial layer includes a
single peak at 570 eV, corresponding to silica, while the bottom
spectrum has characteristic O K-edge features of alumina at 570 and
590 eV. These corroborate our low-loss spectra findings that the
interfacial layer mainly consists of silicon oxide and the interface
with alumina is fairly abrupt.
The authors are grateful to John Roth, at the Research Resource
Center of University of Illinois at Chicago, who provided help with
the electron microscope and XPS. The authors are also grateful to
Laurent Duquesne at American Air Liquide, Chicago, for help with
TDEAA characterization and in setting up the ALD reactor. This
research was supported by the NSF-GOALI grant no. 0329195.
University of Illinois at Chicago assisted in meeting the publication costs
of this article.
XPS results
.— Figure 9a includes Al 2p core loss spectra of a
References
4.5 nm thick alumina film deposited on Si at 400°C. This spectrum
could not be deconvoluted into more than one symmetric peak at
74.1 eV corresponding to Al3+; this indicates that the film mainly
consists of Al2O3 and it lacks Al–Al clusters or substoichiometric
alumina in the film.33,34 Figure 9b shows Si 2p core loss spectra of
the sample used above for Al 2p spectra. The Si 2p spectra show
two main peaks at ϳ99.3 eV corresponding to bulk Si. The Si 2p
peak can be deconvoluted into Si 2p3/2 ͑99.3 eV͒ and
1. G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys., 89, 5243 ͑2001͒.
2. J. Robertson, in Band Offsets of Wide-bandgap Oxides and Implications for Future
Electronic Devices, p. 1785–1791, AVS, Raleigh, NC ͑2000͒.
3. S. Miyazaki, J. Vac. Sci. Technol. B, 19, 2212 ͑2001͒.
4. K. J. Hubbard and D. G. Schlom, J. Mater. Res., 11, 2757 ͑1996͒.
5. R. D. Shannon, J. Appl. Phys., 73, 348 ͑1993͒.
6. E. P. Gusev, E. Cartier, D. A. Buchanan, M. Gribelyuk, M. Copel, H. Okorn-
Schmidt, and C. D’Emic, Microelectron. Eng., 59, 341 ͑2001͒.
7. K. L. Ng, N. Zhan, C. W. Kok, M. C. Poon, and H. Wong, Microelectron. Reliab.,
Downloaded on 2015-06-23 to IP 131.215.225.9 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

C706
Journal of The Electrochemical Society, 153 ͑10͒ C701-C706 ͑2006͒
43, 1289 ͑2003͒.
23. A. C. Dillon, A. W. Ott, J. D. Way, and S. M. George, Surf. Sci., 322, 230 ͑1995͒.
24. Z. Katz-Tsameret and A. Raveh, in Characterization of Aluminum-Based Oxide
Layers Formed by Microwave Plasma, pp. 1121–1127, Denver, CO ͑1995͒.
25. S.-L. Lin and C.-S. Hwang, J. Non-Cryst. Solids, 202, 61 ͑1996͒.
26. A. R. Chowdhuri, C. G. Takoudis, R. F. Klie, and N. D. Browning, Appl. Phys.
Lett., 80, 4241 ͑2002͒.
27. P. Tarte, Spectrochim. Acta, Part A, 23, 2127 ͑1967͒.
28. B. Foran, J. Barnett, P. S. Lysaght, M. P. Agustin, and S. Stemmer, J. Electron
Spectrosc. Relat. Phenom., 143, 151 ͑2005͒.
29. H. Müllejans and R. H. French, J. Phys. D, 29, 1751 ͑1996͒.
30. K. Suenaga, D. Bouchet, C. Colliex, A. Thorel, and D. G. Brandon, J. Eur. Ceram.
Soc., 18, 1453 ͑1998͒.
31. I. Berbezier, J. M. Martin, C. Bernardi, and J. Derrien, Appl. Surf. Sci., 102, 417
͑1996͒.
8. N. Zhan, M. C. Poon, C. W. Kok, K. L. Ng, and H. Wong, J. Electrochem. Soc.,
150, F200 ͑2003͒.
9. M.-H. Cho, H. S. Chang, Y. J. Cho, D. W. Moon, K.-H. Min, R. Sinclair, S. K.
Kang, D.-H. Ko, J. H. Lee, J. H. Gu, and N. I. Lee, Surf. Sci., 554, L75 ͑2004͒.
10. M.-H Cho, Y. S. Roh, C. N. Whang, K. Jeong, H. J. Choi, S. W. Nam, D.-H. Ko,
J. H. Lee, N. I. Lee, and K. Fujihara, Appl. Phys. Lett., 81, 1071 ͑2002͒.
11. M. Ritala and M. Leskela, Atomic Layer Deposition, Vol. 1, Academic Press, San
Diego, CA ͑2001͒.
12. R. Matero, A. Rahtu, M. Ritala, M. Leskela, and T. Sajavaara, Thin Solid Films,
368, 1 ͑2000͒.
13. A. W. Ott, K. C. McCarley, J. W. Klaus, J. D. Way, and S. M. George, Appl. Surf.
Sci., 107, 128 ͑1996͒.
14. G. S. Higashi and C. G. Fleming, Appl. Phys. Lett., 55, 1963 ͑1989͒.
15. S. M. George, A. W. Ott, and J. W. Klaus, J. Phys. Chem., 100, 13121 ͑1996͒.
16. J. A. Aboaf, J. Electrochem. Soc., 114, 948 ͑1967͒.
17. T. Maruyama and T. Nakai, Appl. Phys. Lett., 58, 2079 ͑1991͒.
18. J. S. Becker, S. Suh, S. Wang, and R. G. Gordon, Chem. Mater., 15, 2969 ͑2003͒.
19. E. M. James and N. D. Browning, Ultramicroscopy, 78, 125 ͑1999͒.
20. P. D. Nellist and S. J. Pennycook, Ultramicroscopy, 78, 111 ͑1999͒.
21. N. D. Browning, M. F. Chisholm, and S. J. Pennycook, Nature, 366, 143 ͑1993͒.
22. C. Morterra and G. Magnacca, Catal. Today, 27, 497 ͑1996͒.
32. C. Gatts, G. Duscher, H. Mullejans, and M. Ruhle, Ultramicroscopy, 59, 229
͑1995͒.
33. T. M. Klein, D. Niu, W. S. Epling, W. Li, D. M. Maher, C. C. Hobbs, R. I. Hegde,
I. J. R. Baumvol, and G. N. Parsons, Appl. Phys. Lett., 75, 4001 ͑1999͒.
34. Y.-C. Jung, H. Miura, K. Ohtani, and M. Ishida, J. Cryst. Growth, 196, 88 ͑1999͒.
35. W. R. Salaneck, R. Bergman, J.-E. Sundgren, A. Rockett, T. Motooka, and J. E.
Greene, Surf. Sci., 198, 461 ͑1988͒.
Downloaded on 2015-06-23 to IP 131.215.225.9 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).