In situ reaction mechanism studies of plasma-assisted atomic layer deposition of Al2O3

Article abstract of DOI:10.1063/1.2357886

Reaction mechanisms during plasma-assisted atomic layer deposition (ALD) of Al₂O₃ from Al(CH₃)₃ and O ₂ plasma were studied by time-resolved quartz crystal microbalance measurements, mass spectrometry, and optical emission spectroscopy. Al(CH ₃)₃ chemisorption on the oxide surface after the plasma pulse releases CH₄ products while from the detection of CO, CO ₂, and H₂O in the O₂ plasma it is established that surface -CH₃ groups are predominantly removed by O radical-driven combustionlike reactions. Also a second pathway exists for -CH₃ removal driven by H₂O generated in this plasma step. These reaction pathways are expected to be generic for plasma-assisted ALD of oxides from metal organic precursors.

Full text of DOI:10.1063/1.2357886

In situ reaction mechanism studies of plasma-assisted atomic layer deposition of Al 2
O 3
S. B. S. Heil, P. Kudlacek, E. Langereis, R. Engeln, M. C. M. van de Sanden, and W. M. M. Kessels
Citation: Applied Physics Letters 89, 131505 (2006); doi: 10.1063/1.2357886
View online: http://dx.doi.org/10.1063/1.2357886
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/89/13?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Surface chemistry of plasma-assisted atomic layer deposition of Al 2 O 3 studied by infrared spectroscopy
Appl. Phys. Lett. 92, 231904 (2008); 10.1063/1.2940598
Reaction mechanisms during plasma-assisted atomic layer deposition of metal oxides: A case study for Al 2 O 3
J. Appl. Phys. 103, 103302 (2008); 10.1063/1.2924406
N, NH, and N H 2 radical densities in a remote Ar – N H 3 – Si H 4 plasma and their role in silicon nitride
deposition
J. Appl. Phys. 100, 093303 (2006); 10.1063/1.2358330
Analysis of the molecular structure of fluorocarbon deposits produced by C 4 F 8 and C 4 F 8 – H 2 plasmas
J. Vac. Sci. Technol. A 19, 689 (2001); 10.1116/1.1349198
Modeling of SiO 2 deposition in high density plasma reactors and comparisons of model predictions with
experimental measurements
J. Vac. Sci. Technol. A 16, 544 (1998); 10.1116/1.581096
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: 131.170.6.51
On: Sat, 06 Sep 2014 18:28:15

APPLIED PHYSICS LETTERS 89, 131505 ͑2006͒
In situ reaction mechanism studies of plasma-assisted atomic layer
deposition of Al O
2
3
a͒
S. B. S. Heil
Department of Applied Physics, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven,
The Netherlands
P. Kudlacek
Department of Physics, West Bohemia University, P.O. Box 314, 306 14 Plzen, Czech Republic
b͒
E. Langereis, R. Engeln, M. C. M. van de Sanden, and W. M. M. Kessels
Department of Applied Physics, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven,
The Netherlands
͑
Received 12 July 2006; accepted 4 September 2006; published online 28 September 2006͒
Reaction mechanisms during plasma-assisted atomic layer deposition ͑ALD͒ of Al O from
2
3
Al͑CH ͒ and O plasma were studied by time-resolved quartz crystal microbalance measurements,
3
3
2
mass spectrometry, and optical emission spectroscopy. Al͑CH ͒ chemisorption on the oxide surface
3
3
after the plasma pulse releases CH products while from the detection of CO, CO , and H O in the
4
2
2
O2 plasma it is established that surface –CH3 groups are predominantly removed by O
radical-driven combustionlike reactions. Also a second pathway exists for –CH removal driven by
3
H O generated in this plasma step. These reaction pathways are expected to be generic for
2
plasma-assisted ALD of oxides from metal organic precursors. © 2006 American Institute of
Physics. ͓DOI: 10.1063/1.2357886͔
The search for appropriate high-k dielectrics for semi-
out experimental support. It is also shown that reaction prod-
conductor applications has made atomic layer deposition
ucts created by the plasma ͑particularly H O͒ can lead to
2
͑
ALD͒ of oxides a very active research field. Recently, the
concurrent reaction pathways at the surface and that reaction
products can significantly affect the plasma properties.
The experiments were carried out in a remote-plasma
ALD system with an inductively coupled plasma source. A
continuous flow of O2 serves both as plasma gas and as
purge gas. Al͑CH3͒3 dosing and plasma exposure are alter-
nated by releasing a trapped volume of Al͑CH ͒ at room
investigations have been extended to ALD processes in
which the commonly used H O dosing step is replaced by O
2
3
dosing or O plasma exposure, because of potential advan-
2
tages such as improved film quality, increased flexibility in
process conditions, and feasibility of lower deposition tem-
peratures. For the pure thermal process, the reaction mecha-
nisms underlying film growth have been investigated for sev-
3
3
temperature and by pulsing 100 W plasma power for 2 s,
respectively. During the cycle the background pressure was
kept constant at ϳ10 mTorr and the deposition temperature
was set at 70 °C by the wall heating of the reactor. Film
growth was monitored by a quartz crystal microbalance
͑QCM͒ with a bakeable crystal holder positioned close to the
substrate holder. The gold covered quartz crystals were first
coated with Al2O3 before taking any data. The reaction prod-
ucts were sampled by a differentially pumped quadrupole
mass spectrometer ͑QMS͒ fitted at the side of the reactor
while plasma emission just above the substrate holder was
detected by optical emission spectroscopy ͑OES͒. All diag-
nostics were operated simultaneously and their data acquisi-
tion rate was set to 10 Hz. In these experiments the intermit-
ted purge times were set as such to allow for a detailed study
by the diagnostics. The Al O material properties obtained
eral materials but for the O -based or plasma-assisted ALD
3
process virtually no dedicated studies have appeared yet. Es-
pecially the thermal process for Al O ALD from trimethy-
2
3
laluminum ͓Al͑CH ͒ ͔ and H O, which can be considered as
3
3
2
1
a model system for ALD, has been examined quite thor-
oughly both experimentally and theoretically.
2
–9
Details
about the reaction mechanism have been established for this
process, which is ruled by ͑i͒ chemisorption of Al͑CH ͒ on
3
3
–
OH surface groups forming adsorbed ͑–O–͒ Al͑CH ͒ by
n 3 3−n
splitting off one of more ligands in the form of CH and ͑ii͒
4
the oxidation of the metal by H O regenerating –OH surface
2
groups while forming CH4.
In this letter, we report on the reaction mechanism of
plasma-assisted ALD of Al O by Al͑CH ͒ and O plasma.
2
3
3 3
2
In situ diagnostics such as quartz crystal microbalance and
quadrupole mass spectrometry were employed and also op-
tical emission spectroscopy was shown to provide unique
time-resolved information. Details of the chemisorption step
of Al͑CH ͒ are presented and key insight about the removal
2
3
1
0,11
with this experimental setup have been reported.
A typical example of the mass change during the
plasma-assisted ALD cycles as measured by the QCM is
shown in Fig. 1͑a͒. When pulsing the Al͑CH ͒ in the reac-
3 3
3
3
tor, the chemisorption of Al͑CH ͒ on the surface can be
of –CH ligands by the plasma is provided by revealing com-
3 3
3
observed as a sharp increase in mass ͑m ͒ which takes place
bustionlike reactions by O radicals. Such reactions have been
hypothesized for metal organic precursors earlier, but with-
1
within the time resolution of the QCM ͑100 ms͒. When the
plasma is turned on, first a large negative spike is observed in
the QCM signal followed by a recovery of the signal to a
slightly higher level in mass ͑m ͒ than after Al͑CH ͒ chemi-
a͒
Electronic mail: s.b.s.heil@tue.nl
Author to whom correspondence should be addressed; electronic mail:
b͒
2
3 3
w.m.m.kessels@tue.nl
sorption. The spike during plasma exposure can be attributed
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: 131.170.6.51
003-6951/2006/89͑13͒/131505/3/$23.00 89, 131505-1 © 2006 American Institute of Physics
On: Sat, 06 Sep 2014 18:28:15
0

131505-2
Heil et al.
Appl. Phys. Lett. 89, 131505 ͑2006͒
FIG. 2. ͑Color online͒ Time-resolved mass spectrometry signals for several
ALD cycles. The most likely parent molecules contributing to the signals at
the selected mass-to-charge ratios m/z are indicated.
FIG. 1. ͑Color online͒ Apparent mass change as determined from the QCM
signal for ͑a͒ an ALD cycle including Al͑CH3͒ dosing and O2 plasma
3
sorption at the surface releases –CH groups by a reaction
3
exposure and ͑b͒ only O plasma exposure.
2
between the –CH and hydrogen at the surface, e.g., by
3
chemisorption of Al͑CH ͒ on –OH surface groups.
3
3
During the plasma step, the signals at m/z=18, 28, and
to an artifact caused by the plasma, most probably due to
charging of the crystal. The spike is also observed when
exposing the QCM just to a plasma pulse as shown in Fig.
4
4 increase indicating the detection of H O, CO, and CO
2
2
reaction products. The signal at m/z=28 can partly be attrib-
uted to dissociative ionization of CO in the QMS, but from
2
1
͑b͒; however, in this case the signal returns exactly to its
the cracking pattern and calibration measurements it has
been established that CO is also present in the plasma. From
starting level before the plasma pulse. The mass increase
after the plasma step, m −m , can therefore be attributed to
2
1
these observations it can be concluded that the –CH groups
3
the replacement of –CH ligands by oxide surface groups.
3
existing at the surface after Al͑CH ͒ chemisorption are re-
3
3
From the ratio of the mass increase in the two half reac-
tions, the fraction of ligands released during the Al͑CH3͒3
pulse can be estimated following the approach proposed by
moved from the surface by combustionlike reactions. These
reactions as well as the oxidation of the Al are predominantly
1
6
1
2
driven by the abundant O radicals in the plasma. This con-
Rahtu and Ritala. In the present case, at 70 °C, ϳ1.8 me-
sumption of O radicals affects the O mass balance in the
2
thyl ͑–CH ͒ groups per chemisorbed Al͑CH ͒ molecule are
3
3 3
plasma and leads to a reduction of the O signal in the initial
2
released indicating that the Al͑CH3͒3 mainly binds
phase of the plasma step as can be observed at m/z=16 and
bifunctionally on the surface creating predominantly
3
2.
͑
–O–͒ Al͑CH ͒ surface groups. This is similar to what has
2 3
6
During the O plasma pulse also an increase in the signal
2
been reported for thermal ALD ͑150–350 °C͒ by QCM.
Remarkably different is, however, that the growth rate
as determined from the QCM measurements
54 ng cm /cycle͒ as well as from in situ spectroscopic el-
lipsometry ͑1.7 Å/cycle͒ is much higher than for H O
at m/z=15 is observed. Closer observation of the data re-
veals that this signal is observed for a significantly shorter
duration than the signals at m/z=18, 28, and 44 ͑about twice
−
2
͑
as short͒. We attribute the signal to CH released from the
4
2
surface by the interaction of H O produced by the combus-
−
2
2
͑
32 ng cm /cycle or 1.1 Å/cycle͒ ͑Ref. 4͒ or O3
1.1 Å/cycle͒ ͑Ref. 13͒ based processes at ϳ70 °C. A higher
tionlike reactions with –CH surface groups. This forms a
3
͑
concurrent reaction path to the combustionlike process once
H O is produced but only when still a considerable amount
growth rate per cycle has been reported earlier for both re-
1
4,15
2
mote and direct plasmas,
and we attribute this higher
of –CH groups is present at the surface. This indirect “ther-
3
growth rate to the availability of a higher number of reactive
mal ALD” reaction becomes insignificant when the combus-
oxide surface groups for Al͑CH ͒ chemisorption after the
3
3
tionlike reactions have removed most of the –CH groups. It
3
O plasma.
2
should be noted that the detection of CH at m/z=16 is
4
Figure 2 shows the QMS signal recorded for selected
mass-to-charge ͑m/z͒ ratios during several ALD cycles. The
most likely parent molecules leading to the different signals,
which can either be due to direct or dissociative ionization of
the corresponding molecule, are indicated. During Al͑CH3͒3
pulsing, the signal at m/z=15 and 16 increases which can be
obscured by the consumption of O during the plasma pulse.
2
Furthermore, we were not able to observe the formation of
larger hydrocarbons, C H , in the plasma as, for example,
x
y
suggested by Elliot et al. for the interaction between O atoms
17
and –CH groups.
3
Due to the relatively low pressure in the plasma, it can
be assumed that the CO and H O observed are predomi-
4
attributed to the detection of CH released during Al͑CH3͒3
2
2
chemisorption at the oxide surface. In principle, also the dis-
nantly formed by the reaction between O radicals and –CH3
ligands. The CO in the plasma can, however, be a product
from incomplete combustion reactions or be produced in the
sociative ionization of Al͑CH ͒ contributes to the signal at
3
3
m/z=15 but it was established that this contribution was
negligible under the current conditions in which the amount
plasma by electron-impact dissociation of CO . Calibration
2
of Al͑CH ͒ dosed is just sufficient to saturate the surface
3
3
experiments using plasmas in CO/O and CO /O mixtures
2
2
2
reactions at the substrate and reactor’s wall. For high over-
showed that a significant level of CO is also created in a
exposures of Al͑CH ͒ , i.e., with a significant amount of
CO -based plasma. Furthermore, it was determined that the
3
3
2
gaseous Al͑CH ͒ in the reactor, a signal appeared at m/z
amounts of CO and CO in the initial phase of the O plasma
3
3
2
2
+
=
27 ͑Al ͒ with a minor contribution at m/z=15. Further-
more, the relative increase at m/z=15 and 16 in Fig. 2 is in
after Al͑CH ͒ dosing reach levels of ϳ7% of the total gas
3 3
density.
The excitation of the plasma species gives also insight
good agreement with the cracking pattern of CH in the
4
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: 131.170.6.51
QMS. It can therefore be concluded that Al͑CH ͒ chemi-
into the plasma-assisted ALD process. In Fig. 3͑a͒ the emis-
3
3
On: Sat, 06 Sep 2014 18:28:15

131505-3
Heil et al.
Appl. Phys. Lett. 89, 131505 ͑2006͒
–
CH ligands from the surface mainly by combustionlike
3
reactions driven by O radicals although a concurrent
thermal-ALD-like reaction path by H O generated by the
2
plasma is also present. The subsequent chemisorption step of
Al͑CH ͒ on the surface appears to be similar as for thermal
3
3
ALD albeit at a higher saturation level. Apparently, after the
O plasma exposure a higher number of reactive oxide sur-
2
face groups are available for Al͑CH3͒3 chemisorption.
Whether these oxide surface groups after O plasma expo-
2
sure are predominantly –OH groups such as in the thermal
ALD process cannot be concluded from the present data.
However, the release of CH during Al͑CH ͒ chemisorption
4
3 3
indicates the availability of hydrogen at the surface and sur-
face –OH groups are likely to be formed in the reaction by
the H O created in the combustionlike reactions. The reac-
2
tion mechanism discussed is expected to have generic rel-
evance for plasma-assisted ALD processes of other oxides
from metal organic precursors and the insight presented has
important implications for processing, e.g., in terms of
growth rate per cycle for plasma-assisted ALD. Furthermore,
it is shown that reaction products from the plasma can sig-
nificantly affect the plasma properties and hardware settings
FIG. 3. ͑Color online͒ ͑a͒ Emission from a pure O2 plasma and an O2
plasma ignited after a preceding Al͑CH3͒3 dose as recorded in the first
͑
e.g., power coupling͒ while a technique such as optical
2
00 ms after the plasma strike. ͑b͒ Time-resolved signal of selected emis-
emission spectroscopy yields unique opportunities to study
and optimize the properties for plasma-assisted ALD pro-
cesses.
sion lines during the O plasma step after a preceding Al͑CH ͒ dose.
2
3 3
sion spectrum of a pure O plasma is shown. Mainly emis-
2
The authors would like to thank A. Rahtu ͑ASM Micro-
chemistry͒ for his assistance with the QCM measurements
and stimulating discussions. The Dutch Technology Founda-
tion STW is acknowledged for their financial support. The
research of one of the authors ͑W.M.M.K.͒ was made pos-
sible by a fellowship from the Royal Netherlands Academy
of Arts and Sciences ͑KNAW͒.
+
4
−
4
sion by O from the first ͑b ⌺ →a ⌸ ͒ and second
2
g
u
2
2
5
͑
→
A ⌸ →X ⌸ ͒ negative systems and by O radicals ͑ P
u
g
5
3
3
S, 777 nm and P→ S, 845 nm͒ can be observed. When
an Al͑CH ͒ dose has preceded the plasma exposure, the
3
3
emission in the first hundreds of milliseconds of the plasma
pulse is remarkably different from the pure O plasma. As
2
also shown in Fig. 3͑a͒ there is an intense broadband emis-
1
R. L. Puurunen, J. Appl. Phys. 97, 121301 ͑2005͒.
sion in the first 200 ms of the plasma pulse which can
2
1
8
S. M. George, A. W. Ott, and J. W. Klaus, J. Phys. Chem. 100, 13121
sclearly be observed by the naked eye as a “blue flash.” At
the end of the plasma pulse the emission is identical to the
3͑
1996͒.
M. Ritala, H. Saloniemi, M. Leskelä, T. Prohaska, G. Friedbacher, and M.
Grasserbauer, Thin Solid Films 286, 54 ͑1996͒.
emission of a pure O plasma. The difference in emission at
2
4
the start of the plasma pulse can be attributed to the plasma
excitation of reaction products released from the surface
which evidently disappears when the surface reactions are
saturated. From the identification of the spectral lines, we
found that the broadband emission is coming predominantly
from CO ͑with the Ångström and Herzberg systems most
prominent͒. Furthermore, also emission by atomic hydrogen
M. D. Groner, F. H. Fabreguette, J. W. Elam, and S. M. George, Chem.
Mater. 16, 639 ͑2004͒.
5
A. W. Ott, J. W. Klaus, J. M. Johnson, and S. M. George, Thin Solid Films
2
92, 135 ͑1997͒.
6
7
A. Rahtu, T. Alaranta, and M. Ritala, Langmuir 17, 6506 ͑2001͒.
M. Juppo, A. Rahtu, M. Ritala, and M. Leskelä, Langmuir 16, 4034
͑2000͒.
S. D. Elliot and J. C. Greer, J. Mater. Chem. 14, 3246 ͑2004͒.
Y. Widjaja and C. B. Musgrave, Appl. Phys. Lett. 80, 3304 ͑2002͒.
E. Langereis, M. Creatore, S. B. S. Heil, M. C. M. van de Sanden, and W.
M. M. Kessels, Appl. Phys. Lett. 89, 081915 ͑2006͒.
B. Hoex, S. B. S. Heil, E. Langereis, M. C. M. van de Sanden, and W. M.
M. Kessels, Appl. Phys. Lett. 89, 042112 ͑2006͒.
A. Rahtu and M. Ritala, Chem. Vap. Deposition 8, 21 ͑2002͒.
S. K. Kim, S. W. Lee, C. S. Hwang, Y.-S. Min, J. Y. Won, and J. Jeong, J.
Electrochem. Soc. 153, F69 ͑2006͒.
A. Niskanen, K. Arstilla, M. Ritala, and M. Leskelä, J. Electrochem. Soc.
152, F90 ͑2005͒.
S. J. Yun, J. W. Lim, and J.-H. Lee, Electrochem. Solid-State Lett. 7, C13
8
9
͑
H ͒ can clearly be observed which is due to electron-impact
10
␣
dissociation of the H O and CH reaction products.
2
4
11
Figure 3͑b͒ shows time-resolved emission signals of the
reaction products ͑CO Ångström system ␯−␯Ј=0–2 and H ͒
␣
1
1
2
3
5
5
and of the O radicals ͑3p P→3s S͒ in the plasma. From
the time-resolved signal of H and CO, for which 1/e-decay
times of ϳ300 and ϳ500 ms can be determined, respec-
tively, it can be concluded that the surface reactions during
the plasma step are saturated well within 1 s. This is in
agreement with results obtained by in situ spectroscopic el-
lipsometry showing saturation of the growth rate for plasma
pulses of Ͼ0.5 s. The time-resolved emission from O radi-
cals shows that oxygen is consumed by the surface reactions
generating the CO, CO , and H O reaction products as well
1
1
4
5
16͑
2004͒.
J. T. Gudmundsson, I. G. Kouznetsov, K. K. Patel, and M. A. Lieberman,
J. Phys. D 34, 1100 ͑2001͒.
1
1
7
S. D. Elliott, G. Scarel, C. Wiemer, M. Fanciulli, and G. Pavia, Chem.
Mater. 18, 3764 ͑2006͒.
See EPAPS Document No. E-APPLAB-89-331639 for a movie of the
plasma emission during a plasma-assisted ALD cycle. This document can
be reached through a direct link in the online article’s HTML reference
section or via the EPAPS homepage ͑http://www.aip.org/pubservs/
8
2
2
as by oxidizing the Al on the surface.
In summary, for the plasma-assisted ALD process of
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: 131.170.6.51
Al O it has been established that the O plasma removes
epaps.html͒.
2
3
2
On: Sat, 06 Sep 2014 18:28:15