Hydrogen Barrier Layer Against Silicon Oxidation during Atomic Layer Deposition of Al2 O3 and Hf O2

Article abstract of DOI:10.1149/1.2405839

Growth at moderate temperatures, below ～100°C, is shown to prevent interfacial silicon oxidation during atomic layer deposition of high-permittivity (high- κ) gate dielectrics on hydrogen-terminated Si(100). Trimethylaluminum, employed for aluminum oxide

Full text of DOI:10.1149/1.2405839

G44
Journal of The Electrochemical Society, 154 ͑2͒ G44-G48 ͑2007͒
0013-4651/2007/154͑2͒/G44/5/$20.00 © The Electrochemical Society
Hydrogen Barrier Layer Against Silicon Oxidation during
Atomic Layer Deposition of Al2O3 and HfO2
Martin M. Frank,a,z Yu Wang,b Ming-Tsung Ho,b Rhett T. Brewer,b
Naim Moumen,c,d and Yves J. Chabala
aIBM T. J. Watson Research Center, Yorktown Heights, New York 10598, USA
bDepartment of Physics and Astronomy, Laboratory for Surface Modification, Rutgers, The State University
of New Jersey, Piscataway, New Jersey 08854, USA
cIBM Systems and Technology Group, Hopewell Junction, New York 12533, USA
dSEMATECH, Austin, Texas 78741, USA
Growth at moderate temperatures, below ϳ100°C, is shown to prevent interfacial silicon oxidation during atomic layer deposition
of high-permittivity ͑high-␬͒ gate dielectrics on hydrogen-terminated Si͑100͒. Trimethylaluminum, employed for aluminum oxide
growth, leaves the hydrogen layer completely intact. Tetrakis͑ethylmethylamido͒hafnium partially scavenges the hydrogen layer
during hafnium oxide nucleation, resulting in an interface composed of H-terminated silicon atoms and Si–O–Hf bridges. In both
cases, high-␬ dielectrics are grown without formation of interfacial SiO2. Once grown at low temperatures, subnanometer Al2O3
layers effectively prevent interfacial SiO2 formation during subsequent growth at higher temperatures required for optimum high-
␬ quality. This two-step deposition scheme may thereby be useful for gate-stack scaling.
© 2007 The Electrochemical Society. ͓DOI: 10.1149/1.2405839͔ All rights reserved.
Manuscript submitted August 1, 2006; revised manuscript received October 2, 2006. Available electronically January 2, 2007.
Metal oxides such as hafnium oxide, aluminum oxide, or
hafnium aluminate are being seriously considered for Si-based mi-
croelectronics. Such materials, their alloys with silicon and/or nitro-
gen, and laminated stacks are expected to replace silicon oxynitride
͑SiON͒ soon as gate insulators in metal-oxide-semiconductor field
effect transistor ͑MOSFET͒ technology.1-3 Their distinctive features
are higher permittivity ͑high-␬͒ than for SiON and sufficient band
offsets and thermodynamic stability in contact with silicon.4 By in-
creasing gate-stack capacitance, the high-␬ materials increase device
speed and/or reduce gate leakage.1 High-␬ dielectrics should thereby
help sustain the scaling of commercial silicon microprocessor tech-
nology, because the thickness of the traditional SiON gate insulator
has been reduced close to its fundamental physical limit of ϳ1 nm
where insulating properties disappear.5,6
For high-␬ dielectrics on Si to be competitive, however, good
chemical and electrical quality of films and interfaces is essential.
Minimum high-␬ layer thickness ͑ϳ1 to 3 nm͒ and low gate leak-
age need to be achieved. Forming sufficiently continuous films with
low surface roughness on essentially oxide-free Si has turned out to
be a nontrivial task from a nucleation and growth chemistry perspec-
tive. In principle, conformal compound layers can be deposited with
monolayer-level thickness control using a variant of chemical vapor
deposition ͑CVD͒ known as atomic layer deposition ͑ALD͒.7,8 This
technique is based on alternating exposures of precursors with self-
saturating surface adsorption properties, e.g., a metal oxide can be
grown from a metal containing molecule ͑M͒ and water vapor
͑H2O͒ according to the sequence M–H2O–M–H2O–... . Popular
precursor combinations include trimethylaluminum ͓TMA,
Al͑CH3͒3͔ and water for Al2O3 growth8,9 and hafnium tetrachloride
͑HfCl4͒ and water for HfO2 growth.10 The metal precursors are
generally designed to react with OH groups, and ultrathin layers can
ity with respect to O2, H2O, NH3, and high-␬ precursors has been
reviewed comprehensively.9 While low surface reactivity may be
advantageous in terms of oxidation resistance, it also causes poor
nucleation characteristics of many ALD-grown high-␬ films, result-
ing in nonlinear growth kinetics and the formation of discontinuous
and electrically leaky gate stacks. Indeed, this is the case for the
HfCl4/water10,14,15 and Al͑CH3͒3/water9,14,16 processes at 300°C.
More importantly, interfacial SiO2 is formed during growth, par-
tially defeating the purpose of starting with an O-free substrate.9,14,15
Nucleation can be enhanced, and more linear growth achieved,
when employing in situ activation of the H/Si surface by a very
reactive oxygen precursor such as O3. However, this results in even
more interfacial SiO2.17,18
In the present work, we demonstrate an approach to minimize
interfacial SiO2 formation on H/Si͑100͒. This approach is based on
ALD growth at reduced temperatures, e.g., 50–100°C, exploiting
the remarkable kinetic stability of H/Si against oxidation.9 We use in
situ transmission Fourier transform infrared ͑FTIR͒ absorption spec-
troscopy to study the early stages of the ALD process and the nature
of the Al2O3/Si and HfO2/Si interfaces. For Al2O3 growth, we ex-
amine the TMA/water system. Because the incorporation of impuri-
ties can be substantial at low growth temperatures,19 we also discuss
a two-step low/high temperature growth scheme that allows for
minimum impurity concentration while ensuring minimum interfa-
cial layer thickness. For HfO2 growth, we use tetrakis͑ethylmethy-
lamido͒hafnium ͓TEMAH, Hf͑NC2H5CH3͒4͔ and water. TEMAH
has been shown to be particularly suitable for the formation of
impurity-free, conformal, and smooth films at deposition tempera-
tures as low as 50°C.20
Experimental
indeed be grown in
a linear fashion on hydrophilic SiO2
The substrates are double-side polished, float-zone grown, and
lightly doped ͑␳ ϳ 10 ⍀ cm͒ Si͑100͒ wafers. Hydrogen termination
is achieved by HF etching ͑ϳ20%, 1 min͒ after RCA cleaning.11
The H-terminated silicon samples are immediately loaded into
home-built, nitrogen-purged ALD reactor chambers9,21,22 connected
to Fourier transform infrared ͑FTIR͒ spectrometers for in situ sur-
face analysis. The infrared transparent KBr windows of the ALD
cells are protected from contamination by shutters or gate valves
during precursor exposures. High-␬ dielectric growth is performed
by alternating metal precursor and water pulses delivered in N2 car-
rier gas, with N2 purge between precursor pulses. Metal precursors
are TMA ͑exposure per pulse: ϳ5 ϫ 107 langmuir͒ for Al2O3
growth and TEMAH ͑exposure per pulse: ϳ3 ϫ 104 langmuir͒ for
HfO2 growth. Isotopically marked water ͑D2O͒ is employed as oxy-
substrates.8,10 However, such SiO2 nucleation layers partially com-
promise the benefit of high-␬ dielectrics by reducing the effective
gate stack capacitance.
In an attempt to obtain atomically sharp Si/high-␬ interfaces,
oxide-free, H-terminated Si ͑H/Si͒ substrates have been utilized.
Such H/Si͑100͒ can be prepared quickly and reproducibly by a hy-
drofluoric acid ͑HF͒ wet etch ͑“HF-last”͒ followed by rinsing in
ultrapure deionized water.11 H/Si is remarkably resistant to oxida-
tion in laboratory air and even in O2- or H2O-containing environ-
ments at temperatures as high as 300°C.12,13 Recently, H/Si reactiv-
z E-mail: mmfrank@us.ibm.com
Downloaded on 2015-06-22 to IP 128.192.114.19 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, 154 ͑2͒ G44-G48 ͑2007͒
G45
Figure 2. Si–H stretch mode intensity as a function of the number of Al2O3
and HfO2 ALD cycles performed at the specified temperatures ͑H-terminated
Si = 1͒. Data for Al2O3 grown at 300°C was extracted from spectra in
Ref. 12.
Figure 1. ͑Color online͒ In situ transmission infrared spectra of
H-terminated Si͑100͒ ͑a͒ before and ͑b͒–͑f͒ after the specified number of
TMA and water exposures at 50°C, ͑g͒ after subsequent inert anneal to
300°C, and ͑h͒ after a final water exposure at 300°C. The reference surface
for oxide phonon and C–H stretch regions was H-terminated Si ͑because no
independent reference surface is available for spectrum ͑a͒ from
H-terminated Si, the ideal spectrum is indicated with a dotted line͒ and for
Si–H stretching region was SiO2/Si.
ever, attack of the Si surface is insignificant. The H/Si signal inten-
sity originating from O-free surface sites ͑ϳ2085–2135 cm−1͒ is
undiminished ͑Fig. 1b and 2͒ and essentially full H coverage is
therefore maintained ͑a minor increase in Si–H stretch intensity is
believed to be due to changes in the dielectric environment, but such
effect is small͒. A slight broadening indicates perturbation of the
H–Si vibrational modes by interaction with the partially oxidized
Al͑CH3͒x film physisorbed on top of the H/Si layer. We note that the
small H–Si feature at ϳ2250 cm−1, originating from minority defect
sites with oxygen in Si–H surface backbonds, is suppressed by TMA
exposure, indicating that TMA may chemisorb at such sites.
gen precursor to ensure clear spectral separation of hydroxyl groups
on the sample from those present in the infrared optical setup. This
is expected to leave growth chemistry unaffected. Growth tempera-
tures are 50°C for Al2O3 and 100°C for HfO2, chosen to be as low
as possible while ensuring ͑Al2O3͒ good temperature control in the
experimental setup used and ͑HfO2͒ consistency with previous stud-
ies. Postdeposition annealing via resistive heating is performed in
situ under purified N2 gas purge ͑oxygen impurities Ͻ10−6 ppm͒.
During infrared ͑IR͒ analysis, the sample is kept at a constant tem-
perature of 50°C ͑Al2O3 experiments͒ or 60°C ͑HfO2 experiments͒
to minimize artifacts due to bulk Si phonon absorption. The IR beam
is incident at 70° from the surface normal ͑approximately the Brew-
ster angle͒.
The subsequent water pulse ͑Fig. 1c͒ fully oxidizes the Al and
causes the loss of most ͑though not all͒ Al-bonded methyl groups, as
expected for a good ALD growth. Infrared signals originating from
H/Si ͑Fig. 2͒ and from Si-bonded methyl groups remain undimin-
ished. Also, there is no evidence of H bonded to oxidized Si atoms
24
͑On–Si–H, 2130–2300 cm−1
͒
or of SiO2 phonon signals ͑ca.
1000–1200 cm−1͒.32 Based on phonon intensity calibration using
SiO2 films of known thickness,12,21 we find that the amount of SiO2
is substantially below a total average thickness of 1 Å. This shows
that the water pulse leaves both the hydrogen layer and the Si-
bonded methyl groups fully intact, and Si oxidation is small.
These characteristics of surface reactivity at 50°C are in marked
contrast to what was previously observed during precursor expo-
sures at 300°C, otherwise employing identical exposure
conditions.9,12,14 At 300°C, the first TMA pulse and ͑more so͒ the
subsequent water pulse scavenge H from the surface, causing a H
loss of 60% during the first ALD cycle ͑Fig. 2͒, and a 1.2 Å thick
interfacial SiO2 film is formed, catalyzed by surface Al. In contrast,
at 50°C the Al2O3 is formed on top of a nearly unaffected H layer,
and interfacial SiO2 formation does not occur. This shows that hy-
drogen provides a barrier layer against silicon oxidation when ex-
posed to Al2O3 ALD precursors at moderate temperatures.
The remarkable stability of the H layer at 50°C becomes even
more evident when following Al2O3 growth during subsequent ALD
cycles. A second TMA exposure ͑Fig. 1d͒ again deposits Al–͑CH3͒x
species on the surface. These are subsequently oxidized by the sec-
ond water pulse ͑Fig. 1e͒, leaving the H–Si signal intensity unaf-
fected ͑Fig. 2͒, without the appearance of any distinct SiO2 phonon
mode or incorporation of oxygen in backbonds ͑OxSi–H͒. Even af-
ter 8.5 ALD cycles ͑Fig. 1f͒, the interface remains oxide-free. In-
Results and Discussion
We start by briefly reviewing the structure and infrared spectrum
of the starting surface, i.e., wet-chemically H-terminated Si͑100͒. Its
infrared spectrum ͑Fig. 1a͒ exhibits a structured Si–H stretching
band, reflecting the fact that the H/Si͑100͒ surface is atomically
rough on the double-layer scale, with several types of hydride ter-
mination: mono-͑ϳ2085 cm−1͒, di-͑ϳ2110 cm−1͒, and trihydride
͑ϳ2135 cm−1͒ species.23 A very small concentration of chemical
defects ͑oxygen in Si–H surface backbonds͒ is evidenced by a signal
24
at ϳ2250 cm−1
.
We have previously shown that H/Si is chemi-
cally stable during water vapor exposure at temperatures as high as
300°C.9,12 In the following, we therefore concentrate on dielectric
growth initiated by metal precursor pulses.
Aluminum oxide growth.— First, we discuss low-temperature
Al2O3 growth. The first TMA exposure at 50°C ͑Fig. 1b͒ causes the
rise of modes due to Al–CH3 bonding ͑–CH3 umbrella mode at
ϳ1220 cm−1 and C–H stretching mode at ϳ2940 cm−1͒.25-27
A
Si–CH3 minority species ͑ϳ1270 cm−1͒ is formed as well,28,29 in-
dicating some transfer of methyl groups to Si substrate defect
sites.12 Low-frequency features ͑800–900 cm−1͒ are likely due to
Al–O bonding,30 indicating partial oxidation of the Al by
O-containing trace impurities in the precursor gas stream.31 How-
Downloaded on 2015-06-22 to IP 128.192.114.19 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G46
Journal of The Electrochemical Society, 154 ͑2͒ G44-G48 ͑2007͒
Figure 4. ͑a͒ I–V and ͑b͒ high-frequency C–V characteristics of
Hg/HfO2/H/p-Si͑100͒ MOS capacitors measured with a Hg probe ͑10 kHz,
area 7.3 ϫ 10−4 cm2͒. The HfO2 was deposited using 20 ALD cycles of
TEMAH and water exposures at 100°C, resulting in an effective HfO2 thick-
ness of ϳ18.7 Å as determined by Rutherford backscattering spectrometry.
Figure 3. ͑Color online͒ In situ transmission infrared spectra of
H-terminated Si͑100͒ ͑a͒ before and ͑b͒–͑e͒ after the specified number of
TEMAH and water exposures at 100°C and ͑f͒ after subsequent inert anneal
to 300°C. The reference surface for oxide phonon and C–H stretch regions
was H-terminated Si and for Si–H stretching region was SiO2/Si.
In addition, the in situ infrared characterization of HfO2 growth
reveals that no interfacial SiO2 ͑Ͻ0.5 Å͒ is formed between the
high-␬ layer and silicon substrate at moderate deposition tempera-
tures. Instead of features attributed to SiO2,
a
feature
at ϳ1000 cm−1 indicates that the interfacial structure is likely
Si–O–Hf ͑Fig. 3͒.37 The amount of interfacial Si–O–Hf saturates
after ϳ10 ALD cycles.22 Because half of the initial Si–H is still
present, we infer that the Si–O–Hf species involve approximately
half of the surface Si atoms.
deed, the integrated intensity of the H–Si signal is not diminished
during Al2O3 growth at 50°C, as shown in Fig. 2. At 50°C, thick
Al2O3 films can thus be grown on H/Si while leaving the protective
H layer intact. In contrast, Al2O3 growth at 300°C results in near-
complete H removal after 2 ALD cycles ͑Fig. 2͒ and formation of
4 Å interfacial SiO2 underneath thick Al2O3 films.12
As interfacial Si–H likely reduces the areal density of interfacial
covalent bonds, poor HfO2–Si adhesion could be suspected. A
simple Scotch tape test, i.e., the application and subsequent removal
of adhesive tape, coupled with Rutherford backscattering spectrom-
etry ͑RBS͒ measurements of surface composition before and after
the test, did not provide evidence for facile HfO2 delamination.38
To assess the electrical quality of these unconventional
HfO2/H/Si stacks, we have recorded current-voltage ͑I–V͒ and
high-frequency capacitance–voltage ͑C–V͒ characteristics of
Hg/HfO2/H/p-Si MOS capacitors using a mercury probe ͑Fig. 4͒.
Gate leakage ͑Fig. 4a͒ at moderate gate voltages ͑ϳ10−5 A/cm2͒ is
more than seven orders of magnitude lower than what is acceptable
for high-performance MOSFET applications ͑100 A/cm2 and
more39͒, proving that continuous HfO2 layers are formed. The C–V
characteristics ͑Fig. 4b͒ show that interface state density ͑Dit͒ and
the extent of charge trapping are low ͑we refrain from a quantitative
evaluation of the C–V data, because accurate gate-stack-capacitance
measurements may be hampered by frequency-dependent artifacts40
when employing the mercury probe technique͒. Low Dit without a
forming gas anneal is possibly due to the large areal density of
residual interfacial H, passivating any dangling bonds. We conclude
that good-quality HfO2 layers can be grown on H/Si using a low-
temperature ALD process.
Hafnium oxide growth.— With the TEMAH and water precur-
sors, HfO2 dielectric constant and insulating properties are essen-
tially independent of temperature between 200 and 300°C.33 Given
that impurity-free films grow even at deposition temperatures as low
as 50°C,20 we expect gate dielectrics formed at 100°C to still ex-
hibit comparable electrical quality. Also, growth is nearly linear on
H/Si, indicating that near-continuous films are formed.22,33 This is in
contrast to films formed by HfCl4 and water, which require growth
temperatures around 300°C for acceptable electrical quality34 and
suffer from islanding on H/Si.10,14,15 Besides high film quality and
good nucleation at low temperatures, the TEMAH-water process
also exhibits relatively low interfacial oxide formation on H/Si;
even at the fairly high temperature of 250°C, no more than 5 Å of
interfacial SiO2 are formed.33 At 100°C, an even sharper high-␬/Si
interface may be expected.
An initial TEMAH exposure of H/Si͑100͒ causes a decrease of
Si–H vibrational intensity by ϳ 10% ͑Fig. 3a and b, and Fig. 2͒.
The TEMAH thus clearly reacts with H/Si at moderate temperatures
and at much lower exposures than is the case for TMA ͑TEMAH
exposure = 0.06% of TMA exposure in the present study, and Al2O3
islanding occurs when using such low TMA exposures͒.16 It is this
reaction that gives rise to the near-linear growth behavior reported
previously.22 Upon subsequent D2O exposure ͑Fig. 3c͒, no addi-
tional loss of surface hydride occurs. Additional precursor cycles
cause gradual H loss during TEMAH exposure ͑Fig. 3d and 2͒.
However, half of the surface hydrogen remains in place after 20–30
TEMAH/D2O exposure cycles ͑Fig. 3e and 2͒. The absorption band
of Si–H stretch mode also red-shifts and becomes broader, suggest-
ing a chemical interaction in the immediate vicinity of Si–H species
induced by Hf-containing structures.35 The observation of C–H sym-
metric and asymmetric stretch modes in the spectra region of
2800–3000 cm−1 ͑Fig. 3͒ is due to organic ligands remaining inside
and on top of the growing film,36 especially after TEMAH exposure.
Thermal stability and two-step deposition process.— We have
seen that nearly interfacial SiO2-free Al2O3 layers can be grown on
H/Si at moderate temperatures. This may be expected to result in
superior capacitance ͑electrical thickness͒ scaling compared to con-
ventional growth at the more typical growth temperature of
ϳ300°C. However, when grown at moderate temperatures, the qual-
ity of the Al2O3 itself is inferior, as indicated by a lower density.41
For example, the infrared spectra in Fig. 1c and e demonstrate a
higher degree of CHx incorporation, consistent with an increased
amount of hydrogen detected in low-temperature grown films.19,41
Such hydrocarbon species can be removed by higher temperature
Downloaded on 2015-06-22 to IP 128.192.114.19 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, 154 ͑2͒ G44-G48 ͑2007͒
G47
low as ϳ100°C, as indicated by a low concentration of hydrocar-
bons ͑Fig. 3e͒ and other impurities,20 makes a similar two-step pro-
cess for HfO2 less important. However, the passivating effect of
interfacial hydrogen is also operative during thermal processing of
HfO2. For instance, much of the interfacial H remains in place dur-
ing inert gas anneal to 300°C ͑Fig. 3f͒, and interfacial SiO2 forma-
tion ͑with hydroxyl groups from the HfO2 film as oxygen source͒,22
accordingly is insignificant, in line with the Al2O3 results.
Finally, we note that anneals up to ϳ1000°C are required for
transistor fabrication. Such anneals can result in significant interfa-
cial rearrangements. For example, for HfO2 layers grown at 100°C,
interfacial H desorbs and ϳ7 Å SiO2 form during inert 700°C an-
neals, indicating that excess oxygen in the HfO2 film ͑for example,
in the form of hydroxyl groups͒ can react with the depassivated Si
substrate.22 However, gate electrodes are present during device an-
neals, many of which tend to getter diffusing oxygen.42 In this case,
sharp Si/high-␬ interfaces can be preserved during high-temperature
processing.
Conclusions
Utilizing the TMA/water ALD process at 50°C, Al2O3 films can
be grown on H/Si͑100͒ while leaving the H termination intact; the
H layer acts as a barrier that prevents interfacial Si oxidation. Only
a few covalent bonds are formed at the Al2O3–H/Si interface. In
most locations, the Al2O3 film can therefore likely be regarded as
physisorbed on the H/Si substrate. The exact nature of interfacial
bonding and Al2O3 film adhesion are important objects of future
study. By contrast, Al2O3 growth at 300°C results in near-complete
H removal and formation of 4 Å interfacial SiO2.9,12
HfO2 growth from TEMAH and water at 100°C results in partial
H removal from the H/Si͑100͒ substrate by the metal precursor,
with formation of Si–O–Hf bridges; however, interfacial Si oxida-
tion is prevented as well. Such HfO2/H/Si stacks exhibit good elec-
trical quality, with low gate leakage and interface state density. Even
after inert anneal at 300°C, significant H passivation remains in
place both for Al2O3 ͑ϳ60% of a monolayer͒ and for HfO2
͑ϳ35%͒, and SiO2 formation is minimal. High-quality Al2O3 films
with drastically reduced interfacial SiO2 thickness ͑from 4 to
ഛ 1.5 Å SiO2͒ can be fabricated using a two-step deposition pro-
cess. First, a sharp Si-Al2O3 interface is ensured via initial growth at
moderate temperature, then high-quality Al2O3 growth is performed
at 300°C. The thin Al2O3 film initially formed acts as a diffusion
barrier, impeding O from accessing to the Si substrate during growth
at 300°C. We conclude that H termination of Si can be maintained
during ALD ͑and likely also CVD͒ growth of oxides utilizing water
as the O precursor if low to moderate temperatures ͑e.g., no higher
than 100°C͒ are maintained during formation of an initial closed
metal oxide film.
It may seem surprising that SiO2 formation does not occur during
HfO2 growth, despite partial removal of interfacial hydrogen during
the first TEMAH exposure. We propose that this may be related to
details of interface structure near Si–O–Hf bridging arrangements
thus formed. The bulky metal precursor itself clearly is not able to
scavenge all surface H, due to steric hindrance. Therefore, residual
H-passivated sites may surround each site reacted with TEMAH.
Our results suggest that, given such a local bonding arrangement,
there is no facile pathway for H2O molecules or O atoms/ions to
move directly into the Si subsurface layer. Molecular dynamic cal-
culations will be required to estimate the activation barrier and prob-
ability of O insertion near such depassivated sites.
Figure 5. In situ transmission infrared spectra for Al2O3 films on
H-terminated Si͑100͒ formed during 8–12 ALD cycles, employing ͑bottom͒
conventional constant temperature growth at 300°C and ͑top͒ a two step
50°C/300°C scheme. Reference surface was H-terminated Si.
processing. We therefore propose that good-quality Al2O3 films with
reduced interfacial SiO2 thickness can be fabricated using a two-step
deposition process: ͑i͒ Initial growth at moderate temperatures
͑Ͻ100°C͒ of a very thin Al2O3 film ͑e.g., ϳ10 Å͒ to avoid inter-
facial SiO2 formation while achieving film uniformity, followed by
͑ii͒ higher temperature ͑ϳ300°C͒ growth to reduce carbon impuri-
ties and obtain a good-quality gate dielectric, until the desired total
film thickness is reached. With this process, the thin Al2O3 film
initially formed acts as a diffusion barrier, preventing O from ac-
cessing the Si substrate during growth at 300°C. In the following,
we examine the relevant process steps individually with infrared
spectroscopy.
Starting with the 50°C-grown film discussed in a previous sec-
tion ͑Fig. 1f͒, the sample is first annealed to 300°C in the inert N2
carrier gas ͑Fig. 1g͒. The methyl groups partially decompose, while
the feature due to interfacial H/Si decreases in intensity to about
60% of the H/Si͑100͒ intensity ͑Fig. 2͒. A water pulse at 300°C
͑Fig. 1h͒ decomposes all methyl groups, leading to higher quality
Al2O3 such as that formed at this conventional Al2O3 growth tem-
perature. Still, the Si–H stretch band retains 40% of its original
intensity ͑Fig. 2͒, demonstrating that interfacial H is still present in
appreciable quantities after the oxidizing anneal. In line with this
finding, despite some overlap with the Al2O3 phonon tail, no sig-
nificant longitudinal-optical phonon mode of SiO2 ͑ϳ1200 cm−1
͒
emerges, demonstrating that interfacial SiO2 formation is prevented.
During the next growth cycle at 300°C ͑Fig. 2 and 3͒, the hydrogen
layer still exhibits ϳ40% of the initial coverage and continues to act
as a barrier layer against interfacial Si oxidation.
Figure 5 shows Al2O3 phonon spectra for 8–12 cycle Al2O3
films grown conventionally at 300°C ͑Fig. 5, bottom͒ and via the
two-step process proposed herein ͑Fig. 5, top͒. Estimated thick-
nesses, based on published data for Al2O3 thickness increase per
ALD cycle ͑ϳ1.2 Å/cycle at 50°C; ϳ0.95 Å/cycle at 300°C͒,9,41
are 8 and 11 Å Al2O3 for the 300°C films and 12 Å Al2O3 for the
50/300°C film. The infrared spectra show that amorphous Al2O3 is
grown in both cases, identified by a characteristic longitudinal-
optical phonon mode at ϳ950 cm−1. Furthermore, they prove that
interfacial SiO2 formation is greatly reduced by initial low-
temperature growth ͑ഛ1.5 vs 4 Å SiO2͒.9,12
Acknowledgments
This work was supported in part by the National Science Foun-
dation ͑grant no. CHE-0415652͒ and by SEMATECH. The authors
thank Eric L. Garfunkel, Torgny Gustafsson, and Glen D. Wilk for
insightful discussions.
IBM T. J. Watson Research Center assisted in meeting the publication
cost of this article.
The fact that high-quality HfO2 can be formed at temperatures as
Downloaded on 2015-06-22 to IP 128.192.114.19 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G48
Journal of The Electrochemical Society, 154 ͑2͒ G44-G48 ͑2007͒
͑2002͒.
References
21. M. M. Frank, Y. J. Chabal, and G. D. Wilk, Mater. Res. Soc. Symp. Proc., 745, 41
͑2003͒.
22. M.-T. Ho, Y. Wang, R. T. Brewer, L. S. Wielunski, Y. J. Chabal, N. Moumen, and
M. Boleslawski, Appl. Phys. Lett., 87, 133103 ͑2005͒.
1. G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys., 89, 5243 ͑2001͒.
2. J. Robertson, Eur. Phys. J.: Appl. Phys., 28, 265 ͑2004͒.
3. A. A. Demkov and A. Navrotsky, Materials Fundamentals of Gate Dielectrics,
Springer, Dordrecht, ͑2005͒.
4. K. J. Hubbard and D. G. Schlom, J. Mater. Res., 11, 2757 ͑1996͒.
5. D. A. Muller, T. Sorsch, S. Moccio, F. H. Baumann, K. Evans-Lutterodt, and G.
Timp, Nature (London), 399, 758 ͑1999͒.
23. Y. J. Chabal, G. S. Higashi, K. Raghavachari, and V. A. Burrows, J. Vac. Sci.
Technol. A, 7, 2104 ͑1989͒.
24. T. Hattori, T. Aiba, E. Iijima, Y. Okube, H. Nohira, N. Tate, and M. Katayama,
Appl. Surf. Sci., 104/105, 323 ͑1996͒.
25. Y. Imaizumi, Y. Zhang, Y. Tsusaka, T. Urisu, and S. Sato, J. Mol. Struct., 352/353,
447 ͑1995͒.
6. M. L. Green, E. P. Gusev, R. Degraeve, and E. L. Garfunkel, J. Appl. Phys., 90,
2057 ͑2001͒.
7. M. Ritala and M. Leskelä, in Handbook of Thin Film Materials, H. S. Nalwa,
Editor, Vol. 234, p. 183, Academic Press, New York ͑2002͒.
8. R. L. Puurunen, J. Appl. Phys., 97, 121301 ͑2005͒.
26. C. J. Pouchert, The Aldrich Library of FT-IR Spectra, I ed., Aldrich Chemical
Company, Milwaukee, WI ͑1985͒.
27. R. L. Puurunen, M. Lindblad, A. Root, and A. O. I. Krause, Phys. Chem. Chem.
9. M. M. Frank and Y. J. Chabal, in Materials Fundamentals of Gate Dielectrics, A.
A. Demkov and A. Navrotsky, Editors, p. 367, Springer, Dordrecht ͑2005͒.
10. M. L. Green, M.-Y. Ho, B. Busch, G. D. Wilk, T. Sorsch, T. Conard, B. Brijs, W.
Vandervorst, P. I. Räisänen, D. Muller, M. Bude, and J. Grazul, J. Appl. Phys., 92,
7168 ͑2002͒.
11. W. Kern, in Handbook of Semiconductor Wafer Cleaning Technology: Science,
Technology, and Applications, W. Kern, Editor, Noyes Publications, Park Ridge, NJ
͑1993͒.
Phys., 3, 1093 ͑2001͒.
28. J. A. Glass, E. A. Wovchko, and J. T. Yates, Surf. Sci., 338, 125 ͑1995͒.
29. A. Fidélis, F. Ozanam, and J. N. Chazalviel, Surf. Sci., 444, L7 ͑2000͒.
30. W. Mozgawa, M. Sitarz, and M. Rokita, J. Mol. Struct., 512, 251 ͑1999͒.
31. M. D. Halls, K. Raghavachari, M. M. Frank, and Y. J. Chabal, Phys. Rev. B, 68,
161302͑R͒ ͑2003͒.
32. K. T. Queeney, M. K. Weldon, J. P. Chang, Y. J. Chabal, A. B. Gurevich, J. Sapjeta,
and R. L. Opila, J. Appl. Phys., 87, 1322 ͑2000͒.
12. M. M. Frank, Y. J. Chabal, and G. D. Wilk, Appl. Phys. Lett., 82, 4758 ͑2003͒.
13. X. Zhang, E. Garfunkel, Y. J. Chabal, S. B. Christman, and E. E. Chaban, Appl.
Phys. Lett., 79, 4051 ͑2001͒.
14. M. M. Frank, Y. J. Chabal, M. L. Green, A. Delabie, B. Brijs, G. D. Wilk, M.-Y.
Ho, E. B. O. da Rosa, I. J. R. Baumvol, and F. C. Stedile, Appl. Phys. Lett., 83, 740
͑2003͒.
33. K. Kukli, M. Ritala, J. Lu, A. Hårsta, and M. Leskelä, J. Electrochem. Soc., 151,
F189 ͑2004͒.
34. D. Triyoso, R. Liu, D. Roan, M. Ramon, N. V. Edwards, R. Gregory, D. Werho, J.
Kulik, G. Tam, E. Irwin, X. D. Wang, L. B. La, C. Hobbs, R. Garcia, J. Baker, B.
E. White, and P. Tobin, J. Electrochem. Soc., 151, F220 ͑2004͒.
35. M. J. Kelly, J. H. Han, C. B. Musgrave, and G. N. Parsons, Chem. Mater., 17, 5305
͑2005͒.
36. M. M. Frank, S. Sayan, S. Dörmann, T. J. Emge, L. S. Wielunski, E. Garfunkel,
and Y. J. Chabal, Mater. Sci. Eng., B, 109, 6 ͑2004͒.
37. D. A. Neumayer and E. Cartier, J. Appl. Phys., 90, 1801 ͑2001͒.
38. M.-T. Ho, C.-L. Hsueh, M. Dai, and Y. J. Chabal, Unpublished.
39. International Technology Roadmap for Semiconductors ͑ITRS͒: 2006 Update,
http://public.itrs.net/ ͑2006͒, last accessed Dec 13, 2006.
40. X. Garros, C. Leroux, and J.-L. Autran, Electrochem. Solid-State Lett., 5, F4
͑2002͒.
15. E. P. Gusev, Jr., C. Cabral, M. Copel, C. D’Emic, and M. Gribelyuk, Microelectron.
Eng., 69, 145 ͑2003͒.
16. R. L. Puurunen, W. Vandervorst, W. F. A. Besling, O. Richard, H. Bender, T.
Conard, C. Zhao, A. Delabie, M. Caymax, S. De Gendt, M. Heyns, M. M. Viitanen,
M. de Ridder, H. H. Brongersma, Y. Tamminga, T. Dao, T. de Win, M. Verheijen,
M. Kaiser, and M. Tuominen, J. Appl. Phys., 96, 4878 ͑2004͒.
17. H. B. Park, M. Cho, J. Park, S. W. Lee, C. S. Hwang, J.-P. Kim, J.-H. Lee, N.-I.
Lee, H.-K. Kang, J.-C. Lee, and S.-J. Oh, J. Appl. Phys., 94, 3641 ͑2003͒.
18. S. K. Kim and C. S. Hwang, J. Appl. Phys., 96, 2323 ͑2004͒.
19. J. Buckley, B. De Salvo, D. Deleruyelle, M. Gely, G. Nicotra, S. Lombardo, J. F.
Damlencourt, P. Hollinger, F. Martin, and S. Deleonibus, Microelectron. Eng., 80,
210 ͑2005͒.
41. M. D. Groner, F. H. Fabreguette, J. W. Elam, and S. M. George, Chem. Mater., 16,
639 ͑2004͒.
42. H. Kim, P. C. McIntyre, C. O. Chui, K. C. Saraswat, and S. Stemmer, J. Appl.
Phys., 96, 3467 ͑2004͒.
20. D. M. Hausmann, E. Kim, J. Becker, and R. G. Gordon, Chem. Mater., 14, 4350
Downloaded on 2015-06-22 to IP 128.192.114.19 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).