Non-aqueous routes to metal oxide thin films by atomic layer deposition

Article abstract of DOI:10.1002/anie.200705550

(Figure Presented) Controlled growth: A non-aqueous approach inspired from sol-gel chemistry and adapted to the formation of metal oxide thin films by means of atomic layer deposition is presented. The process is based on the reaction of a carboxylic acid with an alkoxide. Growth of metal oxides is achieved at temperatures as low as 50°C on various supports including carbon nanotubes (see TEM picture). The as-grown films show excellent uniformity and possess good dielectric properties.

Full text of DOI:10.1002/anie.200705550

Communications
DOI: 10.1002/anie.200705550
Thin Films
Non-Aqueous Routes to Metal Oxide Thin Films by Atomic Layer
Deposition**
Erwan Rauwel, Guylhaine Clavel, Marc-Georg Willinger, Protima Rauwel, and Nicola Pinna*
The controlled deposition and growth of high-purity metal
oxide thin films on various supports has tremendous impli-
cations not only in fundamental fields, such as surface science,
but also in applied fields. In microelectronics, for example,
new processes, methods, and approaches towards the produc-
tion of ultrathin metal oxides are sought after. The main aim
is to replace, in the near future, the ubiquitous gate oxides in
field-effect transistors. Reduction of the equivalent gate-
oxide thickness (EOT) has emerged as the most difficult
challenge associated with future device scaling.[1,2] To meet
the strict leakage-current requirements in low-power appli-
cations, the introduction of higher-dielectric-constant materi-
als, in which the tunneling current can be suppressed while
maintaining the drain current, will be necessary. Therefore,
improving the deposition processes to control the silicon/
metal interface is indispensable.[3]
surface in a self-limiting process,[4,5] that is, without the
presence of a gas-phase reaction. As a consequence, ALD
offers excellent surface conformity. Inherent to the process is
the possibility to accurately control the thickness of the
deposited film at an almost atomic level by simply counting
the number of deposition cycles.
Usually, the deposition of oxide thin films by means of
ALD is achieved via hydrolytic reactions in which a
transition-metal precursor (for example, a halide, an amide
or an alkoxide) is reacted with water. Upon hydrolysis and
condensation, a metal oxide network is built. General
limitations of the traditional approach mainly reside in the
presence of undesired impurities, which range from unreacted
carbon species to halides. Therefore, a follow-up postsyn-
thetic heat treatment is required to improve the quality of the
film.
A noble and economic process for the formation of oxide
thin films is atomic layer deposition (ALD). ALD is a form of
chemical vapor deposition (CVD) in which the reaction
between precursor materials is separated into successive
surface reactions. In this manner, the precursor materials
remain separated until the adsorbed species react at the
Even though recent studies have demonstrated that H-
terminated Si(001) surfaces are stable against oxidation by
water at temperatures up to 3008C,[6,7] metal oxide thin films
grown by means of ALD on Si(001) substrates usually present
an oxidized interfacial layer (silica or silicates), except in
particular cases.[8] Similar problems occur in other common
ALD processes that employ stronger oxidizing agents, such as
ozone.
[*] Dr. E. Rauwel, G. Clavel, Dr. M.-G. Willinger, Dr. N. Pinna
Department of Chemistry, CICECO, University of Aveiro
3810-193 Aveiro (Portugal)
To date, only a few alternative routes based on nonoxid-
izing agents have been proposed to ALD. Amongst these, the
most successful depositions were reported for the reaction of
metal alkoxides with metal halides,[9] in which the metal oxide
is formed via an alkyl halide elimination. The formation of
silica and silicates based on metal alkylamide and alkoxysi-
lanol species has also been reported.[10,11]
Taking a closer look at the chemistry, it is noticed that the
evolution of ALD follows a development which is similar to
that of sol–gel chemistry. In both cases, approaches were
pursued to improve the metal oxide formation, either by
modifying the precursor[12,13] or by using alternative non-
hydrolytic reactions.[9,14–17]
Inspired by non-aqueous sol–gel routes for the formation
of metal oxides in solution, we introduce herein a new
approach for the ALD of metal oxide films. This approach is
based on the reaction between a metal alkoxide and a
carboxylic acid on a substrate at temperatures between 50 and
3508C. The main advantages of this procedure are: 1) the low
oxidative character of carboxylic acids, 2) a large deposition-
temperature range, and 3) the high-quality and purity of the
films. Moreover, the low-temperature process avoids decom-
position of the precursor and enables a self-limiting growth. It
should be noted that this has not been achieved so far for the
ALD of hafnia from alkoxides and water.[18,19] In fact, the
depositions of hafnia and titania according to the present
Fax: (+351)234-370-004
E-mail: pinna@ua.pt
Dr. M.-G. Willinger
Fritz Haber Institute of the Max Planck Society
Department of Inorganic Chemistry
Faradayweg4–6, 14195 Berlin (Germany)
Dr. P. Rauwel
Department of Ceramics, CICECO, University of Aveiro
3810-193 Aveiro (Portugal)
[**] The authors thank Prof. Dr. Wolf Widdra and Ralf Kulla, from the
Institute of Physics at the Martin Luther University, Halle (Ger-
many), for support on the conception and realization of the ALD
tool; Dr. Mato Knetz, from the Max Planck Institute of Micro-
structure Physics, Halle (Germany), and Prof. Dr. Markku Leskelä,
from the Department of Chemistry, University of Helsinki, Finland,
for many fruitful discussions; Dr. Dangsheng Su, Achim Klein, and
Rosa Arrigo, from the Fritz Haber Institute of the Max Planck
Society, Berlin (Germany), for the preparation of the thin-film cross-
sections and for supplyingthe carbon nanotubes; as well as Dr.
Frederique Ducroquet, from IMEP-MINATEC, Grenoble, France, for
fruitful discussions. Financial support from FCT project N8 (PTDC/
CTM/65667/2006), FAME network of excellence, and Marie Curie
(MEIF-CT2006-041632) is acknowledged.
Supportinginformation (the Experimental Section) for this article is
available on the WWW under http://www.angewandte.org or from
the author.
3592
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2008, 47, 3592 –3595

Angewandte
Chemie
approach with acetic acid as the oxygen source show typical
ALD windows between 175–250 and 150–2008C, respectively
(see Figure 1). In these temperature ranges, the growth-per-
cycle (GPC) values are constant (namely, 0.5 and
Figure 2. GPC of titania as a function of the openingtime of the
alkoxide valve (at 2008C).
Figure 1. Growth per cycle (GPC) of a) titania deposited on a silicon
substrate (usingacetic acid) as a function of the temperature. At
3008C, the GPC is modulated by the openingtime of the alkoxide
~
&
valve ( : 1 s, : 1.5 s, : 2 s). b) Analogous results for hafnia.
*
0.6 Šcycleꢀ1, respectively) and are in agreement with the
values usually observed for these oxides.[20–23] However, at
temperatures above 2508C (for hafnia) and 2008C (for
titania), the decomposition of the metal alkoxide species
begins, thus leading to an increase in the GPC. At such high
temperatures, the ALD regime is no longer sustained and the
GPC depends on the amount of alkoxide introduced (see
Figure 1a, experimental points at 3008C). In the ALD
window, the trend of the GPC as a function of the opening
time further proves the self-limiting behavior. In fact, at
2008C, the GPC achieves a steady regime at an opening time
of 0.4 s (see Figure 2) and does not increase further. Similar
results were obtained by replacing acetic acid with formic acid
(see Table 1 in the Supporting Information). However,
because of the higher reactivity and lower temperature
stability of formic acid, the ALD window is narrower and
shifted towards lower temperatures.
To demonstrate the versatility of the process, a wide range
of materials have been coated; these materials include carbon
nanotubes (CNTs); wool and cellulose fibers; and Si and
monocrystalline oxide substrates (that is, LaAlO3, SrTiO3).
Scanning electron microscopy (SEM) images of the CNTs
after ALD of titania are shown in Figure 3A. At a medium
resolution, the tubes reveal a smooth surface, whereas at a
higher resolution, a texturing can be observed. The inset in
Figure 3A shows a region of a single tube with defects in the
coating, thereby exposing the carbon underneath.[24] CNTs
were also coated with HfO2. Again, a tube with a partly
defective coating was selected to demonstrate that they are
indeed coated (see Figure 3B). The tubes are also coated
Figure 3. SEM image of carbon nanotubes coated with A) titania and
B) hafnia, both deposited at 2008C. The insets show high-resolution
SEM and TEM images, respectively. C) HRTEM image of a 3.5-nm
thick hafnia film on Si(001) (deposited at 2258C).
inside, as is evidenced in the transmission electron microscopy
(TEM) image by the high contrast at the outer and inner walls
of the tube (see inset of Figure 3B and the Supporting
Information). Finally, in Figure 3C, a high-resolution TEM
(HRTEM) image of the cross-section of an unprocessed
Si(001) wafer coated with HfO2 is presented. The thickness of
the amorphous SiO2 film after deposition is only 0.7 nm,
which roughly corresponds to the thickness of the native
surface oxide layer of unprocessed Si wafers. The as-deposited
Angew. Chem. Int. Ed. 2008, 47, 3592 –3595
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
3593

Communications
metal oxide thin films were amorphous, as demonstrated by
the metal alkoxide with the carboxylic acid is responsible for
the growth of the metal oxide film.
the HRTEM results where the 3.5-nm-thick HfO2 layer shows
no evidence of crystallinity (Figure 3C). The amorphous
character of the films was also evidenced by means of X-ray
diffraction measurements performed on various films depos-
ited on silicon and selected-area electron diffraction spectra
recorded on coated carbon nanotubes (not shown).
Indeed, 1) the presence of an ALD window (Figure 1)
rules out the decomposition of the metal alkoxide in that
temperature range.[12,29] 2) Reactions of carboxylic acid com-
pounds with metal oxide surfaces were extensively studied in
ultrahigh vacuum. It is generally accepted that the carboxylic
acids adsorb dissociatively.[30,31] Here, no clean oxide surfaces
are directly exposed, and thus, reactions can only take place
with the adsorbed species.[32] 3) The reaction of titanium
isopropoxide and acetic acid in an autoclave at temperatures
up to 1808C leads to the quantitative formation of isopropyl
acetate and—at low yield—to the precipitation of nano-
crystalline titania (see the Supporting Information).
The chemical composition of the metal oxide films
deposited on silicon substrates was analyzed by using X-ray
photoelectron spectroscopy (XPS) whereas that of the films
formed on carbon nanotubes was studied by means of energy
dispersive X-ray spectroscopy (EDX) and EDX mapping (see
Figures 2–4 in the Supporting Information). The experiments
show that metal oxides are formed and XPS, in particular,
proves that they possess a low content of carbon impurities.
Electrical measurements were performed to demonstrate
the quality of the films. All measurements were carried out on
approximately 10-nm-thick films deposited at temperatures
between 100 and 3508C (without postgrowth treatment).
With increasing deposition temperature, the global relative
dielectric constant (er) rises from 12.5 to 21.3 for hafnia and
from 11.5 to 43 for titania, thus confirming the good quality of
the films (see Table 1 in the Supporting Information).
We have demonstrated that the applied process leads to
ALD growth of high-quality films. In the following, we will
discuss the chemical mechanisms taking place during depo-
sition.
Based on the above observations, the most plausible
mechanism involves an esterification reaction.[33]
There are two possible mechanisms by which the carbox-
ylic acid can react with the metal precursor in an ALD
process. When the carboxylic acid is introduced, it can react
either by A) a one-step esterification reaction between the
alkoxide ligands on the surface and the acid, thereby
producing hydroxy groups [see reaction (A-1)], or B) the
replacement of the alkoxy ligands by carboxylate groups
under the elimination of an alcohol [see reaction (B-1)].
ꢁMꢀOR0 þ RCOOH ! ꢁMꢀOH þ RCOOR0
ꢁMꢀOR0 þ RCOOH ! ꢁMꢀOOCR þ R0OH
ðA ꢀ 1Þ
ðB ꢀ 1Þ
The reaction between metal alkoxide compounds and
carboxylic acids is found in organic chemistry for the
production of esters[25] and in sol–gel chemistry for modifying
the reactivity of alkoxides.[13] According to recent work on
direct esterification, the reaction of equimolar amounts of
carboxylic acid and alcohol compounds is achievable in the
presence of hafnium alkoxides and other metal salts which act
as catalysts.[25] The Lewis acidity of the metal center is
responsible for the catalytic activity in direct esterification
reactions. In traditional sol–gel chemistry, the reaction of an
alkoxide with a carboxylic acid immediately leads to a mixed
alkoxide–carboxylate complex via a rapid and exothermic
ligand exchange, which is then followed by metal-oxo-cluster
formation.[26] Indeed, an esterification reaction takes place
between free carboxylic acid and the released alcohol
molecules. Hence, the water produced in situ induces
hydrolysis and condensation and leads to the formation of
oxo bridges. A possible secondary mechanism was also
proposed, namely, an intramolecular esterification that
takes place in the coordination sphere of the metal and
leads to a hydrolyzed precursor. However, since the products
of both mechanisms are the same, the intramolecular
esterification could not be experimentally verified.[26–28]
To explore the mechanism behind the proposed ALD
ꢁ
Here, stands for any kind of ligand (for example, OR’,
OH, OOR, OM) coordinated to the metal M, R stands for H
or CH3, and R’ is an alkyl group.
In the next step, when the metal precursor is introduced, it
can react with the hydroxylated surface species (mecha-
nism A), thus leading to the formation of metal–oxygen–
metal bonds under the elimination of an alcohol [reaction (A-
2)]. Alternatively, in the case of mechanism B, a reaction with
the carboxylate surface species takes place under the
elimination of an ester [reaction (B-2)]:
ꢁMꢀOH þ MꢀOR0 ! ꢁMꢀOꢀMꢁ þ R0OH
ðA ꢀ 2Þ
ðB ꢀ 2Þ
ꢁMꢀOOCR þ MꢀOR0 ! ꢁMꢀOꢀMꢁ þ RCOOR0
Additional secondary condensation reactions can occur
during a complete ALD cycle, for example, 1) the condensa-
tion of two adjacent surface M OH groups, thereby forming
ꢀ
ꢀ ꢀ
an M O M bond under the elimination of water, 2) the
ꢀ
condensation of adjacent surface M OH and alkoxy groups
ꢀ
(M OR’) under the elimination of alcohol [R’OH, similar to
reaction (A-2)], and 3) the condensation of two adjacent
surface alkoxy groups under the elimination of ether. The
secondary reactions (1) and (2) require the previous forma-
tion of hydroxy species and can therefore only take place in
the case of the first mechanism (A-1). Regarding the side
reaction (3), it was recently shown that ether elimination is
responsible for the formation of hafnia nanoparticles from an
alkoxide precursor in benzyl alcohol.[34] To rule out ether
elimination, we conducted an ALD control experiment using
process, the reaction products were analyzed by means of 13
C
and 1H nuclear magnetic resonance (NMR) and gas-chroma-
tography-coupled mass spectrometry (GC-MS). As expected,
the unreacted precursors, the alcohol, and the corresponding
ester were detected. Since the ester can also be subsequently
formed on trapping of the products, the NMR and GC-MS
experiments do not conclusively elucidate the reaction
mechanism. However, it is evident that a surface reaction of
3594
www.angewandte.org
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2008, 47, 3592 –3595

Angewandte
Chemie
[2] J. Robertson, Rep. Prog. Phys. 2006, 69, 327.
[3] G. D. Wilk, R. M. Wallace, J. M. Anthony, J. Appl. Phys. 2001,
89, 5243.
[4] M. Leskelä, M. Ritala, Thin Solid Films 2002, 409, 138.
[5] M. Leskelä, M. Ritala, Angew. Chem. 2003, 115, 5706; Angew.
Chem. Int. Ed. 2003, 42, 5548.
[6] Y. B. Kim, M. Tuominen, I. Raaijmakers, R. de Blank, R.
Wilhelm, S. Haukka, Electrochem. Solid-State Lett. 2000, 3, 346.
[7] M. M. Frank, Y. J. Chabal, G. D. Wilk, Appl. Phys. Lett. 2003, 82,
4758.
[8] M. T. Ho, Y. Wang, R. T. Brewer, L. S. Wielunski, Y. J. Chabal,
N. Moumen, M. Boleslawski, Appl. Phys. Lett. 2005, 87, 133103.
[9] M. Ritala, K. Kukli, A. Rahtu, P. I. Raisanen, M. Leskela, T.
Sajavaara, J. Keinonen, Science 2000, 288, 319.
[10] R. G. Gordon, J. Becker, D. Hausmann, S. Suh, Chem. Mater.
2001, 13, 2463.
[11] D. Hausmann, J. Becker, S. L. Wang, R. G. Gordon, Science
2002, 298, 402.
[12] A. C. Jones, H. C. Aspinall, P. R. Chalker, R. J. Potter, T. D.
Manning, Y. F. Loo, R. OꢀKane, J. M. Gaskell, L. M. Smith,
Chem. Vap. Deposition 2006, 12, 83.
only metal alkoxide as the precursor. No film growth was
observed under these conditions. This result confirms that
ether elimination according to secondary reaction (3) is
unlikely and that a carboxylic acid is definitively required for
thin-film formation. Furthermore, no ether was detected by
means of NMR and GC-MS studies.
To determine whether the process followed mechanism A
or B, we gained additional insights into the kinetics of film
formation by monitoring the film GPC as a function of the
valve opening time and the residence time of the precursors
inside the chamber. In the case of acetic acid, it was observed
that the pause between the acid pulse and the purging did not
influence the GPC. On the other hand, for the metal alkoxide
precursor introduced in the second step, a pause of several
seconds was required to complete the reaction and reach the
nominal GPC value (namely, 0.5–0.6 Š per cycle).
This behavior supports a reaction according to mecha-
nism B. The modification of an alkoxide by a carboxylic acid
[such as in reaction (B-1)] is fast and does not require a
thermal activation.[26] In contrast, the ester elimination
process [reaction (B-2)] that takes place between the carbox-
ylate ligands at the surface of the film and the metal alkoxide
is much slower, which is in agreement with the experimental
observations. If the reaction would follow the path proposed
in mechanism A, the opposite behavior would be expected
because the hydroxy groups are very reactive towards
alkoxide species.[35] In that case, the required residence time
for the alkoxide precursor in the second step would be
shorter. Based on the above reasoning, we conclude that a
surface esterification reaction consistent with mechanism B is
responsible for the growth of the metal oxide films.
In conclusion, the only condition required to deposit oxide
thin films by using the present method appears to be the
reaction of a carboxylic acid with a metal alkoxide. Hence, the
process might be extended to a large variety of metal oxides.
The non-aqueous chemistry involved allows the growth of
high-quality metal oxide films—even at very low temper-
atures—without the need of oxidizing agents. Moreover,
these findings do not only bring new insights into the fields of
thin films, but they also contribute to the understanding of
sol–gel chemistry and ester synthesis (an industrially relevant
area). This ALD process can be regarded as a model system
for the mechanistic study of the Lewis acid catalysis in the
esterification reaction by metal salts. The elucidation of these
reaction mechanisms will probably allow the development of
new catalysts based on a more fundamental understanding.
[13] C. Sanchez, J. Livage, M. Henry, F. Babonneau, J. Non-Cryst.
Solids 1988, 100, 65.
[14] M. Niederberger, G. Garnweitner, N. Pinna, G. Neri, Prog. Solid
State Chem. 2005, 33, 59.
[15] M. Niederberger, G. Garnweitner, Chem. Eur. J. 2006, 12, 7282.
[16] M. Niederberger, Acc. Chem. Res. 2007, 40, 793.
[17] N. Pinna, M. Niederberger, Angew. Chem. 2008, DOI: 10.1002/
ange.200704541; Angew. Chem. Int. Ed. 2008, DOI: 10.1002/
anie.200704541.
[18] K. Kukli, M. Ritala, M. Leskela, T. Sajavaara, J. Keinonen, A. C.
Jones, J. L. Roberts, Chem. Mater. 2003, 15, 1722.
[19] K. Kukli, M. Ritala, M. Leskelä, T. Sajavaara, J. Keinonen, A. C.
Jones, J. L. Roberts, Chem. Vap. Deposition 2003, 9, 315.
[20] A. Rahtu, M. Ritala, Chem. Vap. Deposition 2002, 8, 21.
[21] S. K. Kim, W.-D. Kim, K.-M. Kim, C. S. Hwang, J. Jeong, Appl.
Phys. Lett. 2004, 85, 4112.
[22] K. Kukli, M. Ritala, M. Leskelä, T. Sajavaara, J. Keinonen, A. C.
Jones, N. L. Tobin, Chem. Vap. Deposition 2004, 10, 91.
[23] E. P. Gusev, C. Cabral, M. Copel, C. DꢀEmic, M. Gribelyuk,
Microelectron. Eng. 2003, 69, 145.
[24] Most likely, these regions were in contact with other nanotubes
during the ALD process and were detached afterwards during
the preparation of the samples for the SEM measurements.
[25] K. Ishihara, S. Ohara, H. Yamamoto, Science 2000, 290, 1140.
[26] S. Doeuff, M. Henry, C. Sanchez, Mater. Res. Bull. 1990, 25, 1519.
[27] N. Steunou, F. Robert, K. Boubekeur, F. Ribot, C. Sanchez,
Inorg. Chim. Acta 1998, 279, 144.
[28] L. Rozes, N. Steunou, G. Fornasieri, C. Sanchez, Monatsh. Chem.
2006, 137, 501.
[29] M. A. Cameron, S. M. George, Thin Solid Films 1999, 348, 90.
[30] K. S. Kim, M. A. Barteau, Langmuir 1990, 6, 1485.
[31] M. A. Barteau, Chem. Rev. 1996, 96, 1413.
[32] It should be noted that the pressure during the deposition
process was in the mbar range.
[33] It should be noted that in ALD, the two precursors react
separately on the surface. Hence, the sol–gel esterification
mechanism that involves the in situ formation of water can be
excluded.
[34] N. Pinna, G. Garnweitner, M. Antonietti, M. Niederberger, Adv.
Mater. 2004, 16, 2196.
[35] J. Livage, M. Henry, C. Sanchez, Prog. Solid State Chem. 1988,
18, 259.
Received: December 4, 2007
Revised: February 4, 2008
Published online: March 31, 2008
Keywords: atomic layer deposition · metal oxides ·
.
nanostructures · sol–gel processes · thin films
[1] The International Technology Roadmap for Semiconductors,
2005.
Angew. Chem. Int. Ed. 2008, 47, 3592 –3595
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
3595