Surface reactions during atomic layer deposition of Pt derived from gas phase infrared spectroscopy

Article abstract of DOI:10.1063/1.3176946

Infrared spectroscopy was used to obtain absolute number information on the reaction products during atomic layer deposition of Pt from (methylcyclopentadienyl)trimethylplatinum [(MeCp) PtMe₃] and O ₂. From the detection of CO_2<

Full text of DOI:10.1063/1.3176946

Surface reactions during atomic layer deposition of Pt derived from gas phase
infrared spectroscopy
W. M. M. Kessels, H. C. M. Knoops, S. A. F. Dielissen, A. J. M. Mackus, and M. C. M. van de Sanden
Citation: Applied Physics Letters 95, 013114 (2009); doi: 10.1063/1.3176946
View online: http://dx.doi.org/10.1063/1.3176946
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/95/1?ver=pdfcov
Published by the AIP Publishing
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:
93.180.53.211 On: Tue, 04 Feb 2014 08:12:19

APPLIED PHYSICS LETTERS 95, 013114 ͑2009͒
Surface reactions during atomic layer deposition of Pt derived
from gas phase infrared spectroscopy
a͒
W. M. M. Kessels, H. C. M. Knoops, S. A. F. Dielissen, A. J. M. Mackus, and
M. C. M. van de Sanden
Department of Applied Physics, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven,
The Netherlands
͑
Received 30 January 2009; accepted 23 June 2009; published online 10 July 2009͒
Infrared spectroscopy was used to obtain absolute number information on the reaction products
during atomic layer deposition of Pt from ͑methylcyclopentadienyl͒trimethylplatinum
͓
͑MeCp͒PtMe ͔ and O . From the detection of CO and H O it was established that the precursor
3 2 2 2
ligands are oxidatively decomposed during the O pulse mainly. Oxygen atoms chemisorbed at the
2
Pt lead to likewise ligand oxidation during the ͑MeCp͒PtMe pulse however the detection of a
3
virtually equivalent density of CO and CH also reveals a concurrent ligand liberation reaction. The
2
4
surface coverage of chemisorbed oxygen atoms found is consistent with the saturation coverage
reported in surface science studies. © 2009 American Institute of Physics.
͓
DOI: 10.1063/1.3176946͔
Atomic layer deposition ͑ALD͒ of metals has only been
studied to a limited extent despite the growing importance
of ultrathin and conformal metal films in a wide variety of
applications. One metal ALD process that has become popu-
lar is ALD of Pt from ͑methylcyclopentadienyl͒trim-
ethylplatinum ͓͑MeCp͒PtMe ͔ and O dosing as developed
nucleation due to the initial absence of catalytic activation of
oxygen on Pt-free substrates. Under the experimental condi-
tions used, described in Ref. 4, a growth per cycle of
0.045Ϯ0.002 nm/cycle was obtained in good agreement
1
–3
with reports in the literature.
Rutherford backscattering
1
4
−2
spectrometry revealed that ͑3.0Ϯ0.2͒ϫ10 cm Pt atoms,
3
2
1
4
by Aaltonen et al. This process has been adopted for a va-
corresponding to 0.2 ML Pt, are deposited per cycle. As
2
–4
riety of applications,
while it also represents a class of
every ͑MeCp͒PtMe3 molecule consists of nine C atoms and
16 H atoms ͑cf. Fig. 1͒ the amount of C and H atoms that
need to be removed from the surface by ligand oxidation and
other reactions can be calculated when taking the heated sub-
strate area into account.
ALD processes of noble metals in which the catalytic activ-
ity of the film is used to dissociate reactants for the subse-
quent decomposition of the metal precursor ligands. More
specifically, the O dissociates on the Pt surface during the
2
O2 pulse and oxidatively decomposes the ligands of the
Figure 2 shows infrared absorbance spectra taken during
the ͑MeCp͒PtMe and O pulses. The differential spectra re-
͑
MeCp͒PtMe . Mass spectrometry studies have revealed that
3
2
3
flect the difference in gas phase species such as precursor
molecules and reaction products before and after the pulses.
The data are taken with the reaction chamber isolated from
the pump to allow for sufficient measurement time, and the
signal-to-noise ratio of the data is also improved by averag-
this oxidation takes place during the O pulse for the ligands
remaining on the surface after precursor adsorption but also
2
during the precursor adsorption process itself, because oxy-
5
gen atoms reside at the Pt surface after the O pulse. A
2
similar reaction mechanism was observed for ALD of Ru
5
from RuCp and O .
2
2
In this letter additional experimental proof for the reac-
5
tion mechanism proposed by Aaltonen et al. is presented,
while the insight into the mechanism is also extended by
more quantitative data on the reaction products. From gas
phase transmission infrared spectroscopy the production of
CO and H O in the oxidation of the ligands is confirmed by
2
2
absolute density information, while also the production of
CH reaction products is observed during the ͑MeCp͒PtMe
4
3
precursor pulse. On the basis of these data the precursor
adsorption reaction including the role of chemisorbed oxy-
gen atoms is addressed.
Figure 1 shows the thickness of a Pt film deposited at
300 °C ͑reactor wall temperature is 70 °C͒ as a function of
the number of cycles as monitored by in situ spectroscopic
ellipsometry. This Pt film, with a preferential ͑111͒ orienta-
tion, was deposited on a Pt seed layer deposited by plasma-
assisted ALD ͑Ref. 4͒ in order to prevent problems with
FIG. 1. ͑Color online͒ Pt film thickness vs the number of cycles as measured
by in situ spectroscopic ellipsometry. The substrate was a Pt seed layer
deposited by plasma-assisted ALD and the substrate temperature was
3
00 °C. The insets show the ͑MeCp͒PtMe precursor and the thickness for
3
a few half-cycles with the half-integer and full-integer data points represent-
a͒Electronic mail: w.m.m.kessels@tue.nl.
ing the ͑MeCp͒PtMe pulse and O pulse, respectively.
3
2
0
003-6951/2009/95͑1͒/013114/3/$25.00
95, 013114-1
3.180.53.211 On: Tue, 04 Feb 2014 08:12:19
© 2009 American Institute of Physics
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:
9

013114-2
Kessels et al.
Appl. Phys. Lett. 95, 013114 ͑2009͒
FIG. 2. ͑Color online͒ Gas phase absorbance spectra for the ͑MeCp͒PtMe3
pulse and O pulse. The maximum peak positions of ͑MeCp͒PtMe
2
3
−
1
−1
−1
͑
2909 cm ͒, CO ͑2360 cm ͒, and CH ͑3016 cm ͒ are indicated. In the
inset the absorbance due to CH can be distinguished from the absorbance
2
4
4
due to ͑MeCp͒PtMe3 from a comparison with a spectrum measured for
͑
MeCp͒PtMe precursor only ͑dotted line͒.
3
ing the data over several ALD cycles. ͑MeCp͒PtMe is dosed
3
by a single long pulse or by several sequential pulses ͑“mi-
cropulses”͒ with the total pressure remaining below 0.015
Torr while O is dosed in a single pulse at a pressure of 0.7
2
Torr. The formation of CO during both the ͑MeCp͒PtMe
2
3
and O pulses can clearly be observed from the absorbance
2
−
1
peaks around 2360 cm , with most of the CO being pro-
FIG. 3. ͑Color online͒ ͑a͒ ͑MeCp͒PtMe3, CO2, and CH4 densities during
2
͑
MeCp͒PtMe dosing in which the precursor is dosed by sequential micro-
duced during the O pulse. During the O pulse the oxidation
3
2
2
pulses of 0.8 s. Typical error bars are shown for a total dosing time of 4 s.
b͒ The saturation of the growth per cycle with ͑MeCp͒PtMe dosing time as
of ligands is sufficiently large in magnitude such that also
the presence of H O can be observed from the absorption
͑
3
2
measured by spectroscopic ellipsometry for a substrate at 300 °C. The lines
serve as a guide to the eye.
−
1
bands in the regions around 1600 cm ͑not shown͒ and
700 cm . These observations confirm the reaction mecha-
−
1
3
nism proposed by Aaltonen et al., i.e., the oxidation of pre-
CH is produced during the ͑MeCp͒PtMe pulse, i.e., a vir-
cursor ligands takes place during the O pulse as well as
4
3
2
during the precursor pulse, because oxygen atoms, generated
tually equivalent amount of carbon atoms decompose into
CH4 and CO2 during adsorption of the precursor in the ALD
cycle.
6
–8
by O dissociative chemisorption reactions,
reside at the
2
Pt surface after O dosing. Furthermore, the spectrum taken
2
during the ͑MeCp͒PtMe pulse shows that the precursor
The aforementioned results hold for ALD cycles with a
large ͑MeCp͒PtMe3 overdose. The reaction products and the
depletion of precursor were also investigated when dosing
the precursor by micropulses as shown in Fig. 3͑a͒. The
3
dosing reached saturation as a nonzero absorbance of
−
1
͑
MeCp͒PtMe ͑dominated by C–H stretch at 2909 cm as-
3
sociated with Pt–CH and Cp–CH groups͒ can clearly be
observed. Additionally the Q branch ͑3016 cm ͒, and to a
lesser extent the P and R branch, of the C–H stretching mode
3
3
−
1
spectra in Fig. 2 correspond with a total ͑MeCp͒PtMe dos-
3
ing time of 4 s. The data were also compared with a satura-
tion curve measured by spectroscopic ellipsometry at the
300 °C heated substrate holder as shown in Fig. 3͑b͒. Al-
though care needs to be taken when comparing results from
local ellipsometry measurements with global infrared mea-
surements, it is clear that the growth per cycle saturates at a
much smaller precursor dose than the density of the reaction
of CH can be distinguished in the spectrum during the
4
͑
MeCp͒PtMe dosing. This demonstrates the remarkable fact
3
that CH is also produced during the ͑MeCp͒PtMe adsorp-
4
3
tion step during Pt ALD. This reaction product was not re-
ported so far for the Pt ALD process.
Quantitative information on the amount of CO and CH
2
4
produced in both half-cycles of the Pt ALD process was ob-
tained from a calibration of the infrared absorbance intensi-
ties of CO and CH gas over the relevant pressure range.
products. Moreover, the CO density shows a clear saturation
2
behavior but such a behavior is not evident for the CH den-
4
2
4
sity. The ratio of CH4 and CO2 density increases with
For the spectra in Fig. 2, CO densities of ͑3.7Ϯ1.4͒
͑MeCp͒PtMe dosing time when going to precursor overdos-
ing conditions.
2
3
ϫ101 and ͑2.5Ϯ0.4͒ϫ10 cm were found for the O2
4
13
−3
and ͑MeCp͒PtMe pulses, respectively. This implies that ap-
The discrepancy between the results in Figs. 3͑a͒ and
3͑b͒ can be attributed to several effects, for example, by a
͑relative͒ change in the reaction products produced when
reaching saturation. Possibly somewhat colder parts of the
300 °C heated substrate holder play a role and contribute to
a slower saturation behavior. Another explanation is that the
surplus of ͑MeCp͒PtMe precursor reacts with reaction prod-
3
proximately fifteen times more carbon atoms are oxidized
during the O pulse. This is in good agreement with the
2
qualitative results reported by Aaltonen et al., who also ob-
served that only a very small proportion of the ͑MeCp͒PtMe3
ligands where decomposed oxidatively during precursor ad-
1
3
−3
sorption. In addition, a density of ͑3.1Ϯ0.6͒ϫ10 cm of
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://sci 3t ation.aip.org/termsconditions. Downloaded to IP:
93.180.53.211 On: Tue, 04 Feb 2014 08:12:19

013114-3
Kessels et al.
Appl. Phys. Lett. 95, 013114 ͑2009͒
uct species, such as H O, produced during initial
The
oxidative
decomposition
reaction
when
2
͑
MeCp͒PtMe adsorption. This can occur either directly with
2
͑MeCp͒PtMe adsorbs at the surface during the precursor
3
3
H O molecules or indirectly, for example, through –OH sur-
face species generated by the interaction of H O with
pulse takes place via the chemisorbed oxygen atoms. Be-
cause only one CH3 group is liberated per Pt atom during
this pulse, and because the reaction products are CO2, H2O,
2
͑colder͒ surfaces. For one ͑MeCp͒PtMe micropulse, corre-
3
and CH ; the amount of oxygen atoms required on the sur-
sponding to a dosing time of 0.8 s, the precursor is fully
depleted by the surface reactions, while for two micropulses
some of the precursor remains unreacted. For a dosing time
of 0.8 s, the growth per cycle at the substrate ͓Fig. 3͑b͔͒ is
also saturated and, therefore, it is expected that the densities
reported for 1–2 micropulses reflect the reaction products
produced during ALD of Pt at 300 °C well. More support for
this conclusion is obtained when calculating the number of C
4
face per deposited Pt atom can be calculated. This calcula-
tion shows that for every Pt atom, approximately 1.5 oxygen
atoms need to be available as surface-bound oxygen. As
ϳ0.2 ML Pt is deposited per cycle, this implies that a sur-
face coverage of 0.3 ML of oxygen atoms after the O pulse
2
is sufficient for the precursor adsorption reaction to take
place. This surface coverage of oxygen atoms is in very good
agreement with the 0.25 ML saturation coverage of chemi-
sorbed oxygen atoms found in surface science studies on
atoms liberated into the gas phase as reaction products per
ALD cycle. A number of ͑2.4Ϯ0.9͒ϫ1018 C atoms can be
6
,7
Pt͑111͒ exposed to O2. This illustrates the consistency of
the analysis and it provides more evidence that the Pt ALD
reaction proceeds through chemisorbed oxygen atoms
present at the Pt surface. Contrary to the reaction mechanism
calculated from the number of Pt atoms deposited per cycle
at the heated substrate holder, whereas a calculation on the
basis of the CO and CH densities during both half-cycles
and the reactor volume reveals that this number is obtained
for 70% after one ͑MeCp͒PtMe micropulse, for 111% after
two micropulses, and for 153% after five micropulses. Not-
withstanding a large experimental uncertainty, this compari-
son indicates that sufficient ͑MeCp͒PtMe precursor is dosed
2
4
5
5,11
for Ru, the involvement of subsurface oxygen
is there-
3
fore not required.
In conclusion, quantitative insight into the reaction
mechanism of Pt ALD from ͑MeCp͒PtMe and O has been
3
2
3
obtained and can be summarized by the reactions:
into the reaction chamber between one and two micropulses
to achieve ALD saturation conditions for the heated substrate
holder. It also provides support for the aforementioned addi-
tional reactions possibly taking place during precursor over-
dosing.
2
͑MeCp͒PtMe ͑g͒ + 3 O͑ads͒ → 2͑MeCp͒PtMe ͑ads͒
3
2
+ CH ͑g͒ + CO ͑g͒ + H O͑g͒,
͑1͒
4
2
2
Pt
2
͑MeCp͒PtMe ͑ads͒ + 24 O ͑g͒→2 Pt͑s͒ + 3 O͑ads͒
2
2
From the relative densities of CH and CO obtained
4
2
during the ͑MeCp͒PtMe and O pulses, it can be derived
+ 16 CO2͑g͒ + 13 H2O͑g͒,
͑2͒
3
2
that approximately one C atom per precursor molecule is
liberated from the precursor as volatile reaction product dur-
ing adsorption of the precursor on the Pt surface in the ALD
cycle. The other eight C atoms of the precursor molecule
remain at the surface and are oxidatively decomposed during
for ͑MeCp͒PtMe dosing ͓Eq. ͑1͔͒ and O dosing ͓Eq. ͑2͔͒.
3
2
For simplicity, it has been assumed that CH and CO are
4
2
produced in equal amounts during precursor adsorption in
Eq. ͑1͒, whereas in Eq. ͑2͒ the catalytic activity of the Pt is
important for the dissociative chemisorption of the O . This
2
the O pulse. This conclusion is virtually independent of the
2
reaction mechanism for Pt, which is ruled by the saturation
surface coverage of chemisorbed oxygen atoms, can serve as
a model system for ALD processes of noble metals.
number of micropulses considered however the CH :CO
4
2
ratio during precursor adsorption is approximately 1:2 at one
to two micropulses and 1:1 at five micropulses. These obser-
vations can be used to discuss the precursor adsorption
mechanism. Considering the fact that the Pt in ͑MeCp͒PtMe3
is bonded to three CH groups and one CpCH group it can
The Dutch Technology Foundation STW is acknowl-
edged for their financial support ͑STW-TFN 10018͒.
3
3
1
T. Aaltonen, M. Ritala, T. Sajavaara, J. Keinonen, and M. Leskela, Chem.
be hypothesized that one of the CH groups is liberated dur-
3
Mater. 15, 1924 ͑2003͒.
ing precursor adsorption. This CH group can either be oxi-
2
3
X. Jiang and S. F. Bent, J. Electrochem. Soc. 154, D648 ͑2007͒.
3
dized by chemisorbed oxygen or it can react to form CH by
Y. Zhu, K. A. Dunn, and A. E. Kaloyeros, J. Mater. Res. 22, 1292 ͑2007͒.
H. C. M. Knoops, A. J. M. Mackus, M. E. Donders, M. C. M. van de
4
9
4
ligand exchange. The other two CH groups as well as the
3
Sanden, P. H. L. Notten, and W. M. M. Kessels, Electrochem. Solid-State
CpCH group will subsequently be oxidized during the O
3
2
Lett. 12, G34 ͑2009͒.
T. Aaltonen, A. Rahtu, M. Ritala, and M. Leskela, Electrochem. Solid-
pulse. From the stability of covalent ͑substituted͒ cyclopen-
5
tadienyl groups, it is also expected that the CpCH group
State Lett. 6, C130 ͑2003͒.
3
6
remains intact during precursor adsorption. In addition, the
case that the precursor adsorbs with most of the ͑large͒
ligands remaining unreacted on the surface is consistent with
the ellipsometry data obtained for the ALD half-cycles as
shown in the inset of Fig. 1. When film and surface groups
are analyzed with one single dielectric function, the ͑“appar-
J. L. Gland, B. A. Sexton, and G. B. Fisher, Surf. Sci. 95, 587 ͑1980͒.
7
C. T. Campbell, G. Ertl, H. Kuipers, and J. Segner, Surf. Sci. 107, 220
8͑
1981͒.
R. B. Getman, Y. Xu, and W. F. Schneider, J. Phys. Chem. C 112, 9559
9͑
2008͒.
The exact nature of the ligand exchange reaction cannot be assessed on the
basis of the current data. Possible involvement of –OH surface groups can
only be established by surface spectroscopic studies.
10
1
0
ent”͒ thickness, shows a large increase after precursor ad-
sorption as well as a large, albeit somewhat smaller, decrease
after ligand oxidation by O2.
E. Langereis, S. B. S. Heil, H. C. M. Knoops, W. Keuning, M. C. M. van
de Sanden, and W. M. M. Kessels, J. Phys. D 42, 073001 ͑2009͒.
J. F. Weaver, J.-J. Chen, and A. L. Gerrard, Surf. Sci. 592, 83 ͑2005͒.
11
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:
3.180.53.211 On: Tue, 04 Feb 2014 08:12:19
9