Sticking coefficient and growth rate during Al chemical vapor deposition

Article abstract of DOI:10.1063/1.116982

Studies of aluminum chemical vapor deposition (CVD) from dimethylethylamine alane on GaAs(100) 2×4 surfaces were used to identify the high-temperature, flux-limited growth regime and determine the effective sticking coefficient α. Following a short induction period, α was found to achieve a largely temperature-independent steady-state value (α=0.13±0.04), substantially lower than expected based on current CVD models. We propose that steric effects - previously ignored in CVD - play a role in determining the upper limit of a and therefore the maximum achievable growth rate.

Full text of DOI:10.1063/1.116982

Sticking coefficient and growth rate during Al chemical vapor deposition
I. Karpov, W. Gladfelter, and A. Franciosi
Citation: Applied Physics Letters 69, 4191 (1996); doi: 10.1063/1.116982
View online: http://dx.doi.org/10.1063/1.116982
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/69/27?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Ion–surface interactions in low temperature silicon epitaxy by remote plasma enhanced chemical–vapor
deposition
J. Vac. Sci. Technol. A 14, 3024 (1996); 10.1116/1.580166
UV irradiation effects in Al chemical vapor deposition on titanium nitride
Appl. Phys. Lett. 69, 340 (1996); 10.1063/1.118053
Carbon thin films deposited by Capillary Assisted Chemical Vapor Deposition
AIP Conf. Proc. 378, 279 (1996); 10.1063/1.51136
Microstructure and deposition rate of aluminum thin films from chemical vapor deposition with
dimethylethylamine alane
Appl. Phys. Lett. 68, 3567 (1996); 10.1063/1.116639
Postgrowth annealing effects of TiO2 thin films grown on InP substrate at lowtemperature by metalorganic
chemicalvapor deposition
J. Appl. Phys. 79, 4459 (1996); 10.1063/1.361757
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.252.67.66 On: Sun, 21 Dec 2014 23:55:24

Sticking coefficient and growth rate during Al chemical vapor deposition
I. Karpov
Department of Chemical Engineering and Materials Science and Center for Interfacial Engineering,
University of Minnesota, Minneapolis, Minnesota 55455
W. Gladfelter
Department of Chemistry and Center for Interfacial Engineering, University of Minnesota, Minneapolis,
Minnesota 55455
A. Franciosia)
Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis,
Minnesota 55455 and Laboratorio Tecnologie Avanzate Superfici e Catalisi dell’Istituto Nazionale per la
Fisica della Materia, Area di Ricerca, I-34012 Trieste, Italy
͑
Received 29 July 1996; accepted for publication 25 October 1996͒
Studies of aluminum chemical vapor deposition ͑CVD͒ from dimethylethylamine alane on
GaAs͑100͒ 2ϫ4 surfaces were used to identify the high-temperature, flux-limited growth regime
and determine the effective sticking coefficient ␣. Following a short induction period, ␣ was found
to achieve a largely temperature-independent steady-state value ͑␣ϭ0.13Ϯ0.04͒, substantially lower
than expected based on current CVD models. We propose that steric effects—previously ignored in
CVD—play a role in determining the upper limit of ␣ and therefore the maximum achievable
growth rate. © 1996 American Institute of Physics. ͓S0003-6951͑96͒00353-1͔
Chemical vapor deposition ͑CVD͒ of metals is a fast-
growing area of high scientific and technological interest.1
A major advantage of CVD relative to other deposition
and therefore also determine the maximum growth rate
achievable during CVD.
–6
All films were grown in a multichamber ultrahigh
vacuum ͑UHV͒ system with Auger electron spectroscopy
͑AES͒ and reflection high energy electron diffraction
͑RHEED͒ instrumentation and that includes an UHV-
3
methods is the high degree of conformal coverage. The use
of multilevel metallizations in integrated circuits requires
deposition methods that can fill trenches and holes with high
z-aspect ratio, a problem also referred to as step coverage,3
CVD techniques may be used to achieve high step coverage
provided that suitable conditions govern the interaction of
,4
7
compatible CVD reactor. GaAs epitaxial substrates ͑Si
16
18
Ϫ3
doped, nϭ10 –10 cm ) were prepared by molecular
beam epitaxy ͑MBE͒, followed by adsorption of an As cap
4
8
the precursor molecule with the surface. A parameter that
layer. Immediately before Al deposition the As cap layer
has an important effect on conformality is the effective stick-
ing coefficient ␣, defined here simply as the ratio between
the number of metal atoms that are incorporated into the film
per second per unit area and the flux of precursor molecules.
Monte Carlo simulations have shown that, in the molecular-
flow ͑Knudsen͒ regime, as ␣ is increased from zero to unity,
the step coverage degrades rapidly.
was desorbed by heating the sample, yielding the character-
istic GaAs͑100͒ 2ϫ4 RHEED pattern. The surface was then
8
cooled to the desired deposition temperature TD with no de-
tectable variation in the RHEED pattern, and Al films were
grown at a constant temperature by exposing the surface to
DMEAA using a suitable doser and a leak valve.7
The precursor partial pressure pD during deposition was
monitored by means of a nude ion gauge. The ion gauge
reading was calibrated to take into account the ionization
probability of DMEAA using a capacitance monometer.9
All Al depositions were performed in the absence of a
4
We focus here on CVD of aluminum from one of the
5
most promising low-pyrolysis temperature precursors, dim-
ethylethylamine alane ͑DMEAA͒, which offers several po-
tential advantages5 relative to other precursors such as
tri-isobutylaluminum ͑TIBA͒, dimethylaluminum hydride,
trimethylamine alane ͑TMAA͒, or triethylamine alane
,6
1,2
carrier gas, for DMEAA partial pressures p in the 2
D
Ϫ5
Ϫ4
ϫ10 –2ϫ10
Torr range and TD in the 100–550 °C
͑
TEAA͒. For Al deposition from DMEAA on atomically
range. The film morphology was studied ex situ by atomic
force microscopy ͑AFM͒ using a commercial instrument that
included a cantilever with spring constant of 0.58 N/m and a
scanner with a scan area of 125 ␮mϫ125 ␮m. The average
film thickness for continuous films was measured by means
of a profilometer using a diluted HF etch to remove the film
from selected areas of the substrate. For discontinuous films,
estimates of the average ͑nominal͒ thickness, i.e., the volume
of material deposited per unit area, were obtained from the
AFM images, as was done in Ref. 7. The average growth rate
G was estimated as the ratio between the final film thickness
clean GaAs͑001͒ 2ϫ4 surfaces we identified the range of
deposition temperatures and pressures corresponding to flux-
limited growth. In the high temperature limit, ␣ was found to
reach a steady-state, temperature-independent maximum
value of only 0.13Ϯ0.04. This is much lower than expected
based on the current understanding of CVD processes1
since the high-temperature limit of ␣ is assumed to be unity
or close to unity in most models. A surprising general impli-
cation is that steric effects, hitherto not included in the analy-
sis of CVD processes, might determine the upper limit of ␣
,2
͑typically 0.3–0.5 ␮m͒ and the deposition time.
a͒Also with Dipartimento di Fisica, Universit a´ di Trieste, I-34127 Trieste,
Representative AES results illustrating the composition
Italy. Electronic mail: franciosi@area.trieste.it
of Al films deposited on GaAs͑100͒ 2ϫ4 surfaces at
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
Appl. Phys. Lett. 69 (27), 30 December 1996 0003-6951/96/69(27)/4191/3/$10.00 © 1996 American Institute of Physics 4191
28.252.67.66 On: Sun, 21 Dec 2014 23:55:24
1

FIG. 2. Growth rate of Al films fabricated by CVD from DMEAA on
Ϫ4
Ϫ5
GaAs͑100͒ 2ϫ4 as a function of pD in the 2ϫ10 –2ϫ10 Torr range
FIG. 1. Logarithmic growth rate for Al films fabricated by chemical vapor
deposition ͑CVD͒ from dimethylethylamine alane ͑DMEAA͒ on
for a fixed value of T ϭ200 °C. The slope of the linear interpolation is
D
a
1.0Ϯ0.1, consistent with a flux-limited growth rate. The resulting growth
GaAs͑100 2ϫ4 surface. We show results as a function of the substrate
rate normalized to the precursor partial pressure exceeds 5 ␮m/Torr s.
Ϫ4
temperature TD for a constant partial pressure of DMEAA p ϭ2ϫ10
D
Torr. The growth rate is limited by the temperature-activated surface reac-
tion rate only for T Ͻ150 °C. Inset: Representative Auger spectroscopy
D
Al nucleates on the GaAs͑100͒ surface, and the subsequent
growth stage. We have shown earlier that the initial forma-
͑electron-excited, 3 keV primary beam energy͒ results for a 0.34-␮m-thick
7
Al film fabricated by CVD from DMEAA on GaAs͑100͒ 2ϫ4 at
tion of Al on GaAs͑100͒ 2ϫ4 by CVD from DMEAA fol-
lows the Volmer–Weber growth mode, so that the influence
of this initial type of growth on the determination of G
should be assessed. In Fig. 3 we show results for the equiva-
T ϭ200 °C.
D
TDϭ200 °C are shown in the inset of Fig. 1. The dominant
feature at 68 eV is due to Al LMM Auger transitions. For
TDϭ200 °C the films exhibited no detectable oxygen con-
tamination and minimum C and N incorporation. Using the
AES elemental sensitivity factor, and neglecting the possibil-
ity of a composition gradient within the film, the AES inten-
sity ratios would be consistent with C and N concentrations
well below 1%. Slightly higher C and N incorporation ͑but
always at the same percent level͒ was observed for 400
lent film thickness as a function of the deposition time t for
Ϫ5
T ϭ200 °C and p ϭ2ϫ10
Torr. Discontinuous films
D
D
were observed by AFM after 25 and 100 s exposure. The
inset of Fig. 3 shows an AFM image of one such film ͑ob-
tained after a 100 s exposure͒. The sharp increase in the
growth rate in Fig. 3 after 1–2 min was seen to correspond to
the coalescence of the individual Al islands formed through
Volmer–Weber growth to yield a continuous Al coverage of
the substrate. A linear extrapolation of the high-coverage
ϽT Ͻ550 °C.
D
Results for the Al growth rate G(p ,T ) in the 100–
D
D
data to zero coverage suggests an induction period of about
Ϫ4
Ϫ5
5
1
00 °C range for p ϭ2ϫ10 Torr are summarized in Fig.
D
1.5 min at p ϭ2ϫ10 Torr. Therefore for pressures in the
2
D
Ϫ4
, where we plot the logarithm of the growth rate versus
ϫ10 Torr range and total exposure times of the order of
TD to emphasize any temperature-activated dependence.1
,2
The results show, however, that for 150ϽT Ͻ500 °C the
D
films grew at a similar rate Gϳ8 Å/s. For T Ͻ130 °C the
D
growth rate decreased rapidly, dropping to about 0.3 Å/s for
TDϳ100 °C.
A temperature-independent CVD rate is typically asso-
ciated with growth limited by gas-phase transport, i.e., by the
1
,2
precursor flux. To further investigate this possibility, we
gauged the Al growth rate as a function of p in the 2
D
Ϫ4
Ϫ5
ϫ10 –2ϫ10
Torr range for a fixed value of
TDϭ200 °C. The results are summarized in a log–log plot in
Fig. 2. The linear dependence with slope equal to 1.0Ϯ0.1
indicates that in the range of deposition parameters examined
Al growth is a first order process in the precursor partial
pressure, i.e., that film growth is limited by the impingement
rate of the precursor molecules onto the substrate surface.
The growth rate normalized to the precursor partial pressure
FIG. 3. Equivalent thickness of Al films fabricated by CVD from DMEAA
exceeded 5 ␮m/Torr s, which is quite promising for techno-
logical applications.1
–3
on GaAs͑100͒ 2ϫ4 as a function of deposition time for TDϭ200 °C and
Evidence that for T Ͼ150 °C the
D
Ϫ5
pDϭ2ϫ10 Torr. Discontinuous films were observed by AFM after 25
growth rate is largely temperature independent and propor-
tional to pD is also consistent with the results of a recent
and 100 s exposure. The sharp increase in the growth rate after 1–2 min
corresponds to the coalescence of the individual Al islands formed through
Volmer–Weber growth to yield a continuous Al coverage of the GaAs sub-
strate. Inset: Atomic force microscopy ͑AFM͒ image of a representative Al
film during the induction period ͑100 s exposure͒ within a sampled area of
study of Al CVD from DMEAA on Au substrates in a con-
5
ventional low pressure quartz reactor for p ϳ3 Torr.
D
The average growth rate G in Figs. 1 and 2 makes no
0.7 ␮mϫ0.7 ␮m. The full gray scale in the AFM image corresponds to a
distinction between the initial induction period,1 in which
,2
z range of 60 Å.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
192 Appl. Phys. Lett., Vol. 69, No. 27, 30 December 1996 Karpov, Gladfelter, and Franciosi
28.252.67.66 On: Sun, 21 Dec 2014 23:55:24
4
1

2
1
0 min such as those employed in Fig. 1, the influence of the
precursor would have the time to dissociate, a partial satu-
ration of the surface sites by reaction by-products,10 or steric
effects such as those observed in dissociative
induction period on the determination of G becomes negli-
gible.
13
In our experimental conditions, the relationship between
and the deposition parameters can be derived from simple
chemisorption, which would entail that only a fraction of
the impinging molecules, i.e., those with suitable orientation
relative to the surface might react with the substrate. The fact
that the limiting value of ␣ is largely independent of both
temperature and precursor flux in our opinion clearly sup-
ports the third mechanism since steric effects should depend
only on the initial rotational kinetic energy state of the pre-
␣
kinetic theory considerations. For an ideal gas the flux ␴ of
the gas molecules onto the sample surface is directly propor-
tional to the pressure pD of the precursor and inversely pro-
portional to the square root of the temperature and the pre-
cursor molecular mass m . The growth rate G can be
expressed as
D
13
cursor molecules. Future studies in which the kinetic en-
ergy of the precursor molecules is systematically changed in
beam scattering experiments might lend further support to
this identification.
This work was supported in part by the NSF under Grant
No. DMR-9525758. The Center for Interfacial Engineering
at the University of Minnesota is supported in part by the
NSF under Grant No. CDR-8721551. The authors thank L.
Sorba for her invaluable help in the growth of the GaAs
substrates, Tianhoe Lim for film thickness measurements,
and Kai-Ann Yang for the synthesis of DMEAA.
mAl ␣͑t,T ͒p m
D
D
Al
G͑t,p ,T ͒ϭ␣͑t,T ͒␴͑p ͒
ϭ
,
D
D
D
D
␳Al ␳Al
ͱ
2␲m kT
D
͑1͒
where t is the total deposition time, mAl is the atomic weight
of Al, ␳Al is the corresponding density, mD is the molecular
mass of DMEAA, T is the average temperature of the gas
assumed in thermal equilibrium with the system walls͒, and
k is Boltzmann’s constant. In the flux-limited case, and after
͑
the induction period, the growth rate becomes independent of
time and T , so that ␣ becomes a constant that quantifies the
D
probability of Al deposition from DMEAA onto the Al-
covered substrate for each impinging molecule:
G␳Al
ͱ
D
2␲m kT
1
The Chemistry of Metal CVD, edited by T. Kodas and M. H. Smith ͑VHH,
Weinheim, 1994͒; M. J. Hampden-Smith and T. T. Kodas, Chem. Vap.
Deposition 1, 8 ͑1995͒, and references therein.
␣
ϭ
.
͑2͒
p m
D
Al
From the results of Fig. 2 we can derive the ratio
2
J. R. Creighton and J. E. Parmeter, Crit. Rev. Solid State Mater. Sci. 18,
G/p , and using Eq. ͑2͒ we find ␣ϭ0.13Ϯ0.04. This sur-
175 ͑1993͒.
D
3
H. H. Lee, Fundamentals of Microelectronics Processing ͑McGraw-Hill,
New York, 1990͒.
J.-H. Yun and S.-K. Park, Jpn. J. Appl. Phys. 1 34, 3216 ͑1995͒.
M. G. Simmonds, I. Taupin, and W. L. Gladfelter, Chem. Mater. 6, 935
prisingly low value of the effective sticking coefficient may
have important positive implications for conformal coverage.
For example, simulations have shown that lowering ␣ from
4
5
0.5 to 0.1 would result in a 62%–76% increase in step cov-
͑
1994͒.
6
7
erage over what is expected for ␣ϭ0.5 already for features
K. M. Chen, T. Castro, A. Franciosi, W. L. Gladfelter, and P. I. Cohen,
Appl. Phys. Lett. 60, 2132 ͑1992͒; J. Han, K. F. Jensen, Y. Senzaki, and
W. L. Gladfelter, Appl. Phys. Lett. 64, 425 ͑1994͒.
I. Karpov, G. Bratina, L. Sorba, A. Franciosi, M. G. Simmonds, and W. L.
Gladfelter, J. Appl. Phys. 76, 3471 ͑1994͒; N. Venkateswaran, I. Karpov,
W. Gladfelter, and A. Franciosi, J. Vac. Sci. Technol. A 14, 1949 ͑1996͒.
Y. Fan, I. Karpov, G. Bratina, L. Sorba, W. Gladfelter, and A. Franciosi,
J. Vac. Sci. Technol. B 14, 623 ͑1996͒.
with aspect ratio near unity.4
The few available studies relating to Al CVD from
TIBA, TMAA, and TEAA have used reactive scattering
studies or growth rate measurements to determine the surface
reaction kinetic parameters and model the Al deposition
8
9
1
0
rate. Their results can be interpreted in terms of a
temperature-dependent ␣ that reflects the adsorption ͑revers-
ible and irreversible͒ probability of the precursor at free Al
surface sites, the surface reaction rate of the adsorbed pre-
cursor molecules, and the desorption of the reaction by-
After a series of freeze-pump-thaw cycles, precursor pressures from 0.2 to
1.7 Torr as recorded with a capacitance manometer were set in the UHV-
compatible gas line. The valve between the gas line and the main UHV
chamber was then open completely to allow the precursor to expand in the
main chamber and achieve equilibrium. From the ion gauge reading in the
main chamber, the known volume ratio between the gas line and the main
chamber, and the initial capacitance manometer reading, we obtained an
ion gauge sensitivity factor for DMEAA relative to nitrogen of 0.9Ϯ0.1,
with the quoted error corresponding to the scatter of eleven experimental
results.
products. The high-temperature limit of ␣ during CVD in all
such models is unity or close to unity.10 For DMEAA we
find, however, a much lower limiting value of ␣ ͑0.13
1
1,12
Ϯ0.04͒. This is not unheard of in semiconductor epitaxy;
1
1
10
11
recent studies of Si homoepitaxy from SiH reported, for
B. E. Bent, R. G. Nuzzo, and L. H. Dubois, J. Am. Chem. Soc. 111, 1634
4
͑
1989͒; L. H. Dubois, B. R. Zegarski, C.-T. Kao, and R. G. Nuzzo, Surf.
example, a temperature-independent limiting value of
Sci. 236, 77 ͑1990͒; L. H. Dubois, B. R. Zegarski, M. E. Gross, and R. G.
Nuzzo, Surf. Sci. 244, 89 ͑1991͒.
K. Werner, S. Butzke, S. Radelaar, and P. Balk, J. Cryst. Growth 136, 322
␣
ϳ0.15 for 460ϽT Ͻ580 °C, whereas ␣ϳ0.12 was ob-
D
served with Si H . In previous studies of other Al precursors,
2
6
however, the limiting value of ␣ was typically assumed to be
͑
1994͒; S. M. Mokler, W. K. Liu, N. Ohtani, J. Zhang, and B. A. Joyce,
1
,2,10
unity.
Surf. Sci. 275, 16 ͑1992͒.
C. E. Otis and R. W. Dreyfus, Phys. Rev. Lett. 67, 2102 ͑1991͒; T. Okada,
Y. Nakayama, W. K. A. Kumuduni, and M. Maeda, Appl. Phys. Lett. 61,
12
Mechanisms that may account for ␣Ӷ1 at temperatures
high enough for Al deposition not to be limited by the pre-
cursor dissociation probability at the surface would in prin-
ciple include a temperature-activated desorption of the pre-
cursor molecules so rapid that only a fraction of the adsorbed
2368 ͑1992͒; C. Gabbanini, S. Gozzini, and A. Lucchesi, Appl. Phys. Lett.
67, 715 ͑1995͒.
13
J. Dai and J. Z. H. Zhang, Surf. Sci. 319, 193 ͑1994͒; M. Beutl, M.
Riedler, and K. D. Rendulic, Chem. Phys. Lett. 247, 249 ͑1995͒.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
Appl. Phys. Lett., Vol. 69, No. 27, 30 December 1996 Karpov, Gladfelter, and Franciosi 4193
28.252.67.66 On: Sun, 21 Dec 2014 23:55:24
1