Mat. Res. Soc. Symp. Proc. Vol.

716 2002 Materials Research Society

Low Temperature Metal Organic Chemical Vapor Deposition of Aluminum Oxide Thin
Films for Advanced CMOS Gate Dielectric Applications
Spyridon Skordas, Filippos Papadatos, Zubin Patel, Guillermo Nuesca, Eric Eisenbraun, Evgeni
Gusev1, and Alain E. Kaloyeros.
University at Albany Institute for Materials, School of Nanoscience and Nanoengineering,
Albany, NY 12203, U.S.A.
1
IBM, T. J. Watson Research Center, P. O. Box 218, Yorktown Heights, NY 10598, U.S.A.
ABSTRACT
A low-temperature metal organic chemical vapor deposition (MOCVD) process for the
growth of aluminum oxide for gate dielectric applications has been developed. Amorphous films
were deposited on 200-mm Si(100) wafers, employing an Al -diketonate precursor [Al(III) 2,4pentanedionate] and H2O. Chemical and microstructural properties of films grown in a
temperature range of 250-450oC were studied using x-ray photo-electron spectroscopy (XPS), xray diffraction (XRD), Rutherford back-scattering spectrometry (RBS), nuclear reaction analysis
(NRA), crosssectional (X) scanning and transmission electron microscopy (XSEM and XTEM).
A design of experiment (DOE) method was used for process mapping and optimization. An
optimized process window was defined for the growth of dense, amorphous films with carbon
and hydrogen inclusion as low as 1 at. % and 3 at. % respectively. Post-deposition annealing
studies indicated that chemical and structural film properties are generally stable up to 650C.
The electrical performance of the films was evaluated by capacitance-voltage (C-V) and currentvoltage (I-V) measurements on metal-oxide-semiconductor (MOS) structures. The dielectric
constant (k) obtained was 6.7-9.6 (depending on annealing conditions), with equivalent oxide
thickness (EOT) as low as 1.3 nm. Leakage current densities lower than that of equivalent SiO2
films were also achieved.
INTRODUCTION
As allowable SiO2 thickness shrinks, the semiconductor industry is facing the replacement of
SiO2-based gate dielectrics owing to leakage current-related reliability issues. High dielectric
constant(k) insulators, which allow thicker layers while maintaining equivalent capacitance, are a
solution1. Many materials are under consideration, especially those possessing amorphous
microstructures, good thermal and chemical stability, high band-gap and breakdown field, and
compatibility with potential integration scenarios1. Al2O3 offers these characteristics, as well as
low permeability of alkali and other impurities2,3,4. In previous studies, Al2O3 films were grown
via various CVD and physical vapor deposition (PVD) methods5,6.
Low temperature CVD is attractive since potential issues such as ion-induced damage to
substrates and undesirable thermal strain are avoided. MOCVD methods for Al2O3 growth have
been reported, with various precursors including Al -diketonates, such as Al(III) 2,4pentanedionate [Al(acac)3], due to their desirable properties7,8. These include stability in air,
decomposition pathways in oxidizing ambient leading to contaminant-free oxide phases at low
temperature (<450oC), and precursor volatility, the last allowing delivery by commercially
available systems. We report herein on the development of a manufacturable 200-mm wafer LTMOCVD Al2O3 via an Al(acac)3 and H2O route. The chemical and structural properties of the
B4.11.1

films were systematically correlated with processing parameters (and optimized) using a DOE
approach. Annealing studies were carried out to determine thermal stability, and electrical testing
(C-V, I-V) was performed to complement material performance.
EXPERIMENTAL DETAILS
Al2O3 films were grown on 200-mm p- and n-type Si(100) wafers in a warm-wall, high
vacuum capable reactor, allowing multiple processing without exposure to air. Samples were
dipped in HF to remove native oxide, and rinsed with de-ionized H2O before they were moved
on a resistively heated chuck in the reactor. Some films were deposited on ultrathin thermally
grown oxynitride. For systematic development and optimization of the MOCVD Al2O3 process,
a DOE study was employed. Al(acac)3 flow was controlled by a MKS 1153A solid-source
delivery system, and H2O flow by a MKS DLI-25B direct liquid injector (DLI)9 The Al(acac)3
reservoir was kept at 185C and delivery lines from it to the reactor were kept at 190-220C,
preventing Al(acac)3 condensation. The H2O reservoir was held at 40oC, and the lines between
DLI and reactor at 120-160C, to prevent condensation. All reactants entered the reactor via a
cone injector above the wafer, allowing uniform flow distribution.
Thickness was measured by CS-SEM, and composition (Al, O, and C relative at. %) was
measured by XPS depth-profiling. Incorporation of H was measured by NRA, using a 6.38 MeV
15 ++
N beam, via the 1H(15N,)12C nuclear reaction. XRD was used to measure microstructure.
Film density was obtained by RBS, employing a 2.0 MeV He+ beam. Relatively thick films (120
nm) were used in materials analyses to obtain best precision, while ultra-thin (<8.0 nm) films
were used for electrical characterization. Thermal and chemical stability were evaluated by
measuring film properties after a 30 min anneal at 650C in 620 torr Ar ambient.
For electrical evaluation, films of various thickness (< 8.0 nm) were grown on 0.8 nm Si
oxynitride / n-Si(100). The oxynitride layer was used to avoid SiO2 formation during Al2O3
growth10. Al dots (area ~1.3x10-4 cm2, 100 nm thick) were evaporated on Al2O3 through a
shadow mask, to form the MOS structures. C-V curves at 10 kHz, 100 kHz, and 1 MHz were
obtained, as well as I-V curves. Annealing using various conditions was carried out, and C-V and
I-V measurements were repeated to compare with the as-deposited film behavior. The annealing
conditions used were: a) Forming gas anneal (FGA, 90% Ar, 10% H2) at 550oC for 30 min, b)
rapid thermal annealing (RTA) in N2 at 700oC for 30 s, followed by FGA at 550oC for 30 min,
and, c) O2 anneal at 550oC for 30 min followed by FGA at 550oC for 30 min.
In the DOE optimization study, Minitab statistical software was used to generate a list of
depositions and employ resulting data to model the process11. A response surface model was
used, including nonlinear terms, to best model film growth11. The model yielded 31 runs at 25
points of the process variable space (7 runs at center-point to check reproducibility and statistical
relevance of results). The process ranges employed were: pressure 1.0-5.0 torr, wafer
temperature 250-450oC, Al(acac)3 flow rate 1.0-4.0 sccm, and H2O flow rate 0.2-1.0 cm3/min.
The optimized process suggested involved pressure 3.0 torr, wafer temperature 350oC, Al(acac)3
flow rate 2.5 sccm, and H2O flow rate 0.6 cm3/min.
DISCUSSION
The films produced were analyzed and results were input to the DOE software. Regression
analysis yielded models for the influence of process parameters on film properties. All DOE data
B4.11.2

shown are presented in three-dimensional surface contours, with film properties plotted as a
function of two process parameters, with the other ones fixed at center-point values. Growth rate
was measured combining CS-SEM (not shown) for film thickness and deposition time. Fig. 1
shows trends for the dependence of growth rate on temperature and Al(acac)3 flow rate. It
strongly depends on temperature, rising from 1 to 20 nm/min as temperature increases from 250
to 450C. This is attributed to more efficient Al(acac)3 decomposition at high temperature, thus
more precursor contributing to film formation. Growth rate increases nearly linearly with
increasing Al(acac)3 flow, suggesting mass transport limited growth. Regression analysis proved
process pressure and H2O flow rate insignificant in affecting growth rate in the processing
window studied, indicating saturated reaction with respect to H2O.
XPS depth profiling was used to determine chemical composition in the films. The Al 2p, O
1s, and C 1s photoelectron peaks were monitored to identify respective relative atomic
percentages. No contaminants besides C were detected. A typical XPS depth profile from a 120
nm-thick film using the optimized set of parameters. A consistent stoichiometry and low C
contamination throughout the film is observed. Trends for C inclusion as a function of deposition
temperature and pressure are shown in Fig. 3. Temperature seems to have a minor effect on C
levels. Inclusion of 2.0 at. % at 350C and above is exhibited, increasing to 4.0 at. % at 250oC.
This can be explained by incomplete precursor decomposition at lower temperature, although it
should be noted that the variations in C levels (2 at. %) fall within the precision limits of XPS.
Regression analysis showed process pressure, and reactant flows as insignificant with regard to C
at. % in the range investigated. Al:O atomic ratios were also measured by XPS. The desired Al:O
ratio for stoichiometric Al2O3 is Al/O = 0.67. Small variations (<5 at. %) ~0.66 in this ratio
were observed in the window investigated but no statistically significant trends were observed,
and it is felt that these variations are within the experimental precision of XPS.
H inclusion was measured by NRA using the 1H(15N,)12C nuclear reaction. H may originate
from either the Al(acac)3 or H2O reactants, as well as H2O and hydrocarbons from air exposure.
A typical H depth profile is shown in Fig. 4, showing uniform depth distribution. High surface H
levels are attributed to native H2O and Adventitious hydrocarbon contamination, as analyses
were carried out ex-situ. Fig. 5 plots trends for H inclusion related to process pressure and
temperature. The increase in temperature causes decrease in H from 15 at. % at 250C to 6 at. %
at 450C, consistent with the existence of cleaner precursor decomposition pathways at higher
deposition temperature. Regression analyses indicate that neither aluminum precursor nor water
flow exerts influence on H inclusion over the processing window explored, with H levels varying
only 3-4 at. % over the entire flow ranges encompassed.
Film density was measured by RBS (not shown), combined with CS-SEM images for film
thickness. In RBS spectra, the Si, O, and Al peaks are the only ones visible. The C level (<4 at.
%) is too small to be readily observed. In Fig. 6, the density trends as related to deposition
temperature and pressure are shown. The data indicate that the significant factor is process
pressure, an increase of which >3.0 torr typically leads to density > ~3.0 g/cm3 (~80% that of
bulk Al2O3). Small variations of density with temperature over the window investigated exist.
Neither of the reactant flow rates nor process temperature is statistically important for film
density. The maximum density observed, 3.3 g/cm3, is within 10-15% of bulk Al2O3.
The film thermal and chemical stability were evaluated by measuring film properties after exsitu annealing at 650oC for 30 min in 620 torr r ambient. By NRA, there is significant reduction

B4.11.3

Figure 1. Surface response curve for the

growth rate as a function of process
temperature and Al(acac)3 flow.

Figure 2. Typical XPS depth profile of

CVD Al2O3 film. Consistent stoichiometry
is observed.

Figure 3. Surface response curve for the

effect of wafer temperature and process
pressure on carbon contamination.

Figure 4. Representative NRA depth profile

of MOCVD film showing hydrogen
incorporation.

of H inclusion in annealed films, down to 2-4 at. %, regardless of the initial H at. %. Within XPS
precision, C at. % and Al/O ratio seem unaffected. Thickness obtained by CS-SEM, combined
with RBS, demonstrate film densification by 15-20 %, owing to reduction of film thickness,
consistent with H reduction. XRD measurements verified that films remained amorphous.
From XPS, SEM and TEM data, the formation of a parasitic SiO2 layer between Al2O3 and Si
is suggested, probably due to the effects of the oxidizing ambient on Si during initial growth.
This can lead to significantly lower k and EOT values of the stack and must be avoided. The
solution we used before proceeding with electrical characterization was a stable oxynitride layer
(EOT < 1.0 nm) on Si to prevent SiO2 formation.
C-V measurements for as deposited structures (not shown) show relatively low accumulation
capacitance (indicating lower k), stretchout and other features indicating high interfacial trap
B4.11.4

Figure 5. Surface response curve showing

hydrogen incorporation as a function of
process temperature and pressure.

Figure 6. Surface response curve

demonstrating the density as a function of
process temperature and pressure.

density, high hysteresis indicating significant presence of mobile charges, and flat-band voltage
(Vfb) shift due to fixed charges. Upon annealing, significant changes are noted.
Capacitance increases by ~20 %, interfacial traps become passivated, and hysteresis decreases
significantly, regardless of annealing method. In Fig. 7 the EOT as a function of film thickness is
presented from which we obtain the k values for all cases. As shown, upon annealing k improves
from 6.7 to 9.1-9.6. EOT as low as 1.3 nm was achieved. Fig. 8 summarizes results from I-V
measurements (not shown). Leakage current density is calculated at Vfb+1.0 V and its
logarithmic values are plotted vs. EOT. For reference, the leakage current line corresponding to
similar structures with SiO2 as gate dielectric is shown1.
For the case of FGA there is marginal improvement compared to equivalent SiO2 films,
which becomes substantial (~1 order of magnitude) for O2 anneal followed by FGA.

Figure 7. The EOT as derived from C-V measurements at 100 kHz, plotted as a function of film
thickness. The k values are obtained from the slope of the lines, using the equation listed.
B4.11.5

Figure 8. The logarithmic leakage current density measured by I-V measurements at Vfb+1V
vs.EOT. The straight line corresponds to the leakage performance of equivalent SiO2 films.
The as-deposited samples as well as the ones with RTA N2 anneal followed by FGA exhibit poor
leakage behavior, generally inferior to that of equivalent SiO2 films.
CONCLUSIONS
A low temperature MOCVD process for Al2O3 thin-film growth was developed. The effect of
processing variables on the film properties was evaluated, and an optimized process window was
established (pressure 3.0 torr, wafer temperature 350oC, Al(acac)3 flow rate 2.5 sccm, and H2O
flow rate 0.6 cm3/min).. The Al2O3 films thus deposited are stoichiometric (Al:O~0.66), dense
(~3.0 g/cm3), possess ~2 at. % C and 5-10 at. % H, demonstrate good thermal stability up to
650oC and promising electrical performance (k ~ 9.1-9.6, improved leakage performance
compared to SiO2) for use as high-k dielectrics in advanced CMOS processing.
ACKNOWLEDGMENTS
The authors thank IBM for supporting this work under its University Partnership (UPP) Program.
REFERENCES
1

D. A. Buchanan, IBM J. Res. Dev. 43, 245 (1999).

K. J. Hubbard, and D. G. Schlom, J. Mater. Res., Vol. 11, No. 11, (1996).
3
C. C. Wang, S. A. Akbar, W. Chen, and V. D. Patton, J. Mater. Sci. 30, 1627 (1995).
4
G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys. 89, 5243 (2001).
5
Y. Kim, H. Park, J. Chun and W. Lee, Thin Solid Films 237, 57 (1994).
6
E. Ciliberto, I. Fragal, R. Rizza, and G. Spoto, Appl. Phys. Lett. 67, 1624 (1995).
7
T. Klein, D. Niu and G. Parsons, Mat. Res. Soc. Symp. Proc. Vol. 567, 445 (1999).
8
A. C. Dillon, A. W. Ott, J. D. Way, S. M. George, Surface Science 322 (1995), 230-242.
9
MKS Instruments, Inc., Andover, Massachusetts, http://www.mksinst.com.
10
E. P. Gusev, H. C. Lu, E. L. Garfunkel, T. Gustafsson, M. L. Green, IBM J. Res. Dev. 43, 265 (1998).
11
Minitab Inc., State College, Pennsylvania, http://www.minitab.com.
2

B4.11.6