Supercritical-carbon dioxide-assisted cyclic deposition of metal oxide and metal thin films

Article abstract of DOI:10.1063/1.2181651

Thin films of aluminum oxide and palladium were deposited on silicon at low temperatures (70-120 °C) by a cyclic adsorption/reaction processes using supercritical CO2 solvent. Precursors included Al(hfac) 3, Al(acac) 3, and Pd(hfac) 2, and aqueous H2 O2, tert-butyl peracetate, and H2 were used as the oxidants or reductants. For the precursors studied, growth proceeds through a multilayer precursor adsorption in each deposition cycle, and film thickness increased linearly with the number of growth cycles.

Full text of DOI:10.1063/1.2181651

Supercritical-carbon dioxide-assisted cyclic deposition of metal oxide and metal thin
films
Dipak Barua, Theodosia Gougousi, Erin D. Young, and Gregory N. Parsons
Citation: Applied Physics Letters 88, 092904 (2006); doi: 10.1063/1.2181651
View online: http://dx.doi.org/10.1063/1.2181651
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/88/9?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
{001} Oriented piezoelectric films prepared by chemical solution deposition on Ni foils
J. Appl. Phys. 116, 014105 (2014); 10.1063/1.4886597
Characteristics of high-purity Cu thin films deposited on polyimide by radio-frequency Ar/H2 atmospheric-
pressure plasma jet
J. Appl. Phys. 113, 123301 (2013); 10.1063/1.4795808
Fabrication for multilayered composite thin films by dual-channel vacuum arc deposition
Rev. Sci. Instrum. 79, 065104 (2008); 10.1063/1.2937195
Reduction of thin oxidized copper films using a hot-filament hydrogen radical source
J. Vac. Sci. Technol. A 25, 415 (2007); 10.1116/1.2712197
Damascene Cu electrodeposition on metal organic chemical vapor deposition-grown Ru thin film barrier
J. Vac. Sci. Technol. B 22, 2649 (2004); 10.1116/1.1819911
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:
130.64.175.185 On: Mon, 01 Dec 2014 14:54:41

APPLIED PHYSICS LETTERS 88, 092904 ͑2006͒
Supercritical-carbon dioxide-assisted cyclic deposition of metal oxide
and metal thin films
Dipak Barua,a͒ Theodosia Gougousi,b͒ Erin D. Young,c͒ and Gregory N. Parsons
North Carolina State University, Raleigh, North Carolina 27695
͑Received 31 August 2005; accepted 23 January 2006; published online 3 March 2006͒
Thin films of aluminum oxide and palladium were deposited on silicon at low temperatures
͑70–120 °C͒ by a cyclic adsorption/reaction processes using supercritical CO2 solvent. Precursors
included Al͑hfac͒3, Al͑acac͒3, and Pd͑hfac͒2, and aqueous H2O2, tert-butyl peracetate, and H2 were
used as the oxidants or reductants. For the precursors studied, growth proceeds through a multilayer
precursor adsorption in each deposition cycle, and film thickness increased linearly with the number
of growth cycles. © 2006 American Institute of Physics. ͓DOI: 10.1063/1.2181651͔
Deposition of metal and dielectric films from precursors
dissolved in supercritical solvents has been demonstrated for
a variety of electronic and optical applications.1–4 Most pre-
vious work from supercritical CO2 ͑sc CO2͒ has been per-
formed in a continuous growth mode ͑analogous to chemical
vapor deposition͒ where all precursors are introduced simul-
taneously to the reactor cell. One of the limitations of this
technique is that it is typically done in batch mode so that the
film thickness is controlled by the amount of precursor
loaded and the extent of reaction. In this work, we present
the concept of sequential solvent-phase deposition, where
precursors dissolved in supercritical fluid flow into the reac-
tion chamber in a cyclic A/B exposure scheme, similar to an
atomic layer deposition ͑ALD͒ process.5 For the precursors
and conditions used in this work, the process resulted in
reproducible multilayer linear growth rather than the self-
limiting submonolayer film growth observed during typical
ALD processes.
Because the density of sc CO2 ͑critical temperature
31 °C and critical pressure 7.4ϫ106 Pa͒ is orders of magni-
tude higher than any gas,6 it can dissolve large concentra-
tions of organic species. The diffusivity of the dissolved spe-
cies is between that for liquids ͑ϳ10−5 cm2/s͒ and gases
͑ϳ1 cm2/s͒ under typical process conditions. As a result, the
precursor may be transported to the surface similarly to
vacuum processes but at a significantly larger concentration.
Another anticipated benefit of sc CO2 is its high solvation
energy, which can be used in promoting the deposition reac-
tion, and facilitating the removal of reaction byproducts at
low temperatures. In this work, the deposition of aluminum
oxide ͑Al2O3͒, and palladium ͑Pd͒ using cyclic reactant ex-
posures is described. These materials are of interest in mi-
croelectronics for potential applications in advanced metal-
oxide-semiconductor transistor gate stacks as gate dielectric
or metal gate materials.7,8
teflon O rings. A similar larger cell ͑ϳ40 ml͒ was used for
the dissolution of the precursors. An ISCO 260D syringe
pump delivered the CO2 through 1/16 in. ͑outer diameter͒
tubing and high pressure manifold and valves. The tempera-
ture of the cells and lines were maintained with individually
controlled external resistive heaters. The reaction effluent
passed though an activated charcoal bed. The Si͑100͒ sub-
strates ͑ϳ2ϫ1 cm2͒ were prepared by exposing to JTB-100
Baker clean for 5 min, followed by de-ionized water rinse
͑10 min͒, and nitrogen-blow dry. Precursors for deposition
included aluminum acetylacetonate ͑99%͒ ͓Al͑acac͒ ͔, alu-
3
minum
hexafluoroacetylacetonate
͑minimum
98%͒
͓Al͑hfac͒ ͔, and palladium ͑II͒ hexafluoroacetylacetonate
3
͓Pd͑hfac͒ ͔, and all were used as received. The fluorinated
2
precursors, ͓Al͑hfac͒ ͔ and ͓Pd͑hfac͒ ͔were stored in a well
3
2
sealed desiccator to prevent reaction with the ambient.
Before each experiment, the cleaned premix cell was
flushed with CO2, and approximately 20–30 mg of precursor
was loaded. The cell was then heated to 70–80 °C and pres-
surized with CO2 to ϳ3500 psi, corresponding to a CO2 den-
sity of 0.7 g/cm3, and the cell stabilized for ϳ30 min. The
highly soluble fluorinated precursors dissolved completely
leading to a precursor concentration ϳ0.07–0.1 wt % in the
premix cell. The solubility limit for the nonfluorinated
Al͑acac͒ precursors is estimated to be ϳ10−4–10−5 wt %,
3
based on similar experimental conditions.9 After loading a
substrate, the reaction cell was flushed with CO2 for 1 or
2 min at 40–50 °C, then pressurized with CO2 the desired
value and heated to the set point temperature.
The depositions were carried out in a hot wall, stainless
steel
high-pressure
reaction
cell
͑21 ml,
1.2 in.͑inner diameter͒ϫ1.2 in. depth, 1.9 in. wall thick-
ness͒ shown in Fig. 1. The ends of the reaction cell used
sapphire windows ͑ϳ0.6 in. thick͒ sealed with compressed
a͒
Electronic mail: dbarua@ncsu.edu
Present address: Department of Physics, University of Maryland, Balti-
b͒
more County, 1000 Hilltop Circle, Baltimore, MD 21250.
Present address: Intel, Chandler Arizona.
c͒
FIG. 1. Schematic diagram of the experimental setup.
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:
0003-6951/2006/88͑9͒/092904/3/$23.00 88, 092904-1 © 2006 American Institute of Physics
130.64.175.185 On: Mon, 01 Dec 2014 14:54:41

092904-2
Barua et al.
Appl. Phys. Lett. 88, 092904 ͑2006͒
FIG. 2. High resolution XP scans for the Al 2p peaks for two aluminum
based films deposited via cyclic deposition. Both films were deposited at
120 °C, and 1600 psi. Film labeled H2O2 was deposited form Al͑acac͒3 and
FIG. 3. Thickness data determined from ellipsometry for Al-based films
deposited from Al͑hfac͒3 –H2O2 at 100 °C and ϳ2000 psi on native oxide
of Si͑100͒ for different precursor exposure times and sample orientations.
Samples were oriented inside the reaction cell such that the deposition sur-
face ͓polished surface of Si͑100͔͒ faced directly to the precursor inlet ͑b͒,
or the deposition surface faced opposite to the precursor inlet ͑sample
flipped by 180° from previous orientation͒ ͑᭿, ᭡͒.
hydrogen peroxide, while film TBA was deposited from Al͑acac͒ and tert-
3
butyl peracetate.
The experiments were carried out in A/B cycles similar
to an atomic layer deposition process.5 For each cycle, a set
volume ͑corresponding to ϳ0.2 g͒ of precursor/sc CO2 solu-
tion ͑A͒ was injected from the premix cell to the reactor
leading to a pressure rise of ϳ30–50 psi. The injected vol-
ume corresponded to ϳ0.15–0.2 mg of fluorinated precur-
sors or ϳ10−4 mg of Al͑acac͒3. The samples were exposed to
the precursor for a time ranging from 30 s to 3 min, and then
the reaction cell was vented and flushed for 1–2 min with
liquid/supercritical CO2. The reaction cell was then repres-
surized and an oxidizing or a reducing agent ͑B͒ was in-
jected. For oxide films, ϳ0.3 ml of liquid oxidizer ͓30%
aqueous solution of H2O2 or 50% tert-butyl peracetate
͑TBA͒ in mineral spirits͔ was injected in each cycle. For
palladium deposition, ultra high purity H2 ͑99.999%, ϳ3.5
ϫ10−4 mol/cycle͒ was used. In every case, the molar quan-
tity of the injected oxidant/reductant was much larger than
required for stoichiometric reaction. Therefore, it is expected
that the growth rate would be determined by the amount of
precursor adsorbed and the cycle time. After 1–3 min, the
reaction cell was vented and flushed and the cycle was re-
peated to build up a macroscopic film thickness. A typical
cycle time was ϳ20 min, primarily due to the use of manual
hardware. Reasonably short cycle times ͑ϳ1 min͒ can be
expected for a more automated and optimized reactor
system.
C ͑285 eV͒ for both samples, but for the sample deposited
with peroxide, a second peak at 289.3 eV is observed, in-
dicative of carbonate in the film. The intensity of the Al 2p
peak indicates significantly lower growth rates when TBA
was used as the oxidant. Use of the fluorinated precursor
Al͑hfac͒ and H2O2 generally results in thicker films com-
3
pared to Al͑acac͒3–H2O2, but there is evidence for alumi-
num hydroxide and carbonate species present in the as-
deposited films when the Al͑hfac͒ is used.4 Film thickness
3
by ellipsometry versus cycle time was studied for Al2O3
films from Al͑hfac͒3 and H2O2 ͑native silicon oxide, 100 °C,
2000 psi͒, and the results are shown in Fig. 3. Each data
point in Fig. 3 was obtained by taking an arithmetic average
of 3–7 values measured on one or more samples with iden-
tical number of cycles. Clearly, film thickness increases lin-
early with number of deposition cycles. However, a signifi-
cant effect of Al͑hfac͒ exposure time on the film thickness
3
per cycle is observed. For the same process conditions, an
increase in precursor exposure time from 30 s to 3 min re-
sulted in nearly fourfold increase in film growth rate ͑Fig. 3͒.
Also, growth rate dependence on the sample orientation was
observed, suggesting a convective flow effect in the reactor
during the deposition. When the deposition surface ͓polished
surface of Si͑100͔͒ was placed in a direct exposure to the
precursor flow, a high growth rate of 5.6 Å/cycle was ob-
served for 30 s of precursor exposure time. However, when
the deposition surface was directed away from the precursor
inlet, the growth rate was 2.1 Å/cycle for the same exposure
time. The observed thickness per growth cycle and the con-
vective flow dependence of growth rate are consistent with
multilayer growth per cycle. Water is generally immiscible
with sc CO2, so the use of aqueous H2O2 peroxide in the
process is expected to lead to a two-phase condition in the
growth cell, which could lead to nonuniform growth. How-
ever, the Al2O3 film did not show evidence for discontinuous
growth over the film surface, consistent with the growth be-
ing determined by the amount of precursor adsorbed during
the exposure cycle, rather than the delivery of oxidizer to the
Films were analyzed by ellipsometry and x-ray photo-
electron spectroscopy ͑XPS͒.
A Rudolf ellipsometer
͑632.8 nm, 70° fixed angle͒ was used to measure the oxide
thickness. XPS was conducted using a Riber LAS3000
͑MAC2 analyzer, Mg K␣ h␯=1253.6 eV, nonmonochro-
matic x-ray source͒ at 75° take-off-angle with 0.1 eV step
size. Three different sets of precursors were studied for cy-
clic deposition of aluminum oxide: Al͑hfac͒ with H2O2,
3
Al͑acac͒ with H2O2, and Al͑acac͒ with TBA. Figure 2
3
3
shows the high resolution XPS data for Al 2p core electrons
for films deposited using Al͑acac͒ and either H2O2 or TBA
3
at 120 °C and 1600 psi. For the film deposited with H2O2 a
peak is observed at 74.8 eV indicative of Al–O bonds,
whereas, for the film deposited using tert-butyl peracetate the
peak is shifted to higher binding energy ͑ϳ75.5 eV͒, which
indicates bonding of the Al atoms with more electronegative
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:
species. The C 1s spectrum ͑not shown͒ shows adventitious
surface. Alternatively, rapid diffusion and mixing of the
130.64.175.185 On: Mon, 01 Dec 2014 14:54:41

092904-3
Barua et al.
Appl. Phys. Lett. 88, 092904 ͑2006͒
gest time dependent precursor adsorption which is consistent
with the nonequilibrium adsorption of ␤-diketone chelates on
–OH terminated surfaces.12 The multilayer adsorption could,
for example, be initiated by forming a hydrogen bond be-
tween the precursor oxygen and surface –OH, with subse-
quent adsorption proceeding by intermolecular ␲ bonding
between adsorbed- and fluid-phase precursor molecules.12
This work demonstrates that metal oxide and metal films
can be deposited from precursors dissolved in sc CO2 using a
binary deposition/exposure reaction sequence. For the pre-
cursors studied, the results indicate growth proceeds through
a multilayer adsorption of precursor in each deposition cycle,
and that the thickness is determined by the number of cycles
used. The binary reaction scheme in sc CO2 may allow for
formation of unique material phases, for example, where the
substrate material is physically modified ͑such as swollen or
densified͒ in sc CO2 allowing the deposited material to inter-
act with the substrate to alter its material properties.
FIG. 4. Pd 3d XPS peaks for palladium films deposited on native silicon
oxide using alternating exposures cycles of Pd͑hfac͒2 and H2 dissolved in sc
CO2. The film deposited with 20 cycles and 180 s Pd͑hfac͒ exposure per
3
cycle is significantly thicker than the film deposited with fewer cycles and
60 s exposure per cycle.
The authors acknowledge financial support from NSF
͑Grant No. CTS- 0304296͒ and the NSF Science and Tech-
nology Center for Environmentally Responsible Solvents
and Processes. They would also like to acknowledge Dr. Ke
Wang for assistance with the sc CO2 experimental setup.
aqueous-phase could result in uniform exposure of the oxi-
dant to the sample surface.
In addition to aluminum oxide, palladium has been de-
posited using the cyclic deposition approach. Figure 4 shows
XPS data of palladium films deposited from Pd͑hfac͒ and
2
1J. J. Watkins, J. M. Blackburn, and T. J. McCarthy, Chem. Mater. 11, 213
͑1999͒.
H2 ͑70–80 °C, 2000–3200 psi͒ using 20 cycles ͓with 180 s
of Pd͑hfac͒ exposure per cycle͔, and 15 cycles ͑with 60 s
3
2J. M. Blackburn, D. P. Long, A. Cabanas, and J. J. Watkins, Science 294,
141 ͑2001͒.
precursor exposure per cycle͒. Films generally adhered
poorly to silicon and oxide surfaces, making thickness evalu-
ation difficult. The Pd 3d peak height indicates that film de-
posited for 20 cycles was much thicker than that deposited
for 15 cycles. The trend in film thickness suggests multiple-
layer growth per exposure cycle and possibly nucleation in-
hibition during initial film growth. Some asymmetric broad-
ening of the C 1s peak to its higher binding energy side is
also observed, consistent with some C–O, CvO, CFx, and
CF3 present in the films.10,11
3H. Ohde, S. Kramer, S. Moore, and C. M. Wai, Chem. Mater. 16, 4028
͑2004͒.
4T. Gougousi, D. Barua, E. D. Young, and G. N. Parsons, Chem. Mater. 17,
5093 ͑2005͒.
5M. Ritala and M. Leskela, Nanotechnology 10, 19 ͑1999͒.
6M. McHugh and V. J. Krukonis, Supercritical Fluid Extraction
͑Butterworth-Heinemann, Boston, 1994͒.
7T. M. Klein, D. Niu, W. S. Epling, W. Li, D. M. Maher, C. C. Hobbs, R.
I. Hegde, I. J. R. Baumvol, and G. N. Parsons, Appl. Phys. Lett. 75, 4001
͑1999͒.
8E. P. Gusev, M. Copel, E. Cartier, I. J. R. Baumvol, C. Krug, and M. A.
Gribelyuk, Appl. Phys. Lett. 76, 176 ͑2000͒.
The observed thickness per growth cycle in the process
reported here is consistent with multilayer adsorption of pre-
cursors during each deposition cycle. Liquid-phase processes
with similar precursors have shown similar results where ad-
9N. G. Smart, T. Carleson, T. Kast, A. A. Clifford, M. D. Burford, and C.
M. Wai, Talanta 44, 137 ͑1997͒.
10W. B. Lin, B. C. Wiegand, R. G. Nuzzo, and G. S. Girolami, J. Am. Chem.
Soc. 118, 5977 ͑1996͒.
sorption followed the Freundlich isotherm,12 Csolid=KfCn
,
fluid
where Csolid represents the mass of precursor adsorbed/cm2
on the Si surface, and Cfluid is the precursor concentration in
the sc CO2. Kf and n are Freundlich adsorption coefficient
and Freundlich exponent, respectively. Also, our results sug-
11J. J. Senkevich, F. Tang, D. Rogers, J. T. Drotar, C. Jezewski, W. A.
Lanford, G. C. Wang, and T. M. Lu, Chem. Vap. Deposition 9, 258
͑2003͒.
12M. P. T. Bradley and D. A. Pantony, Talanta 16, 473 ͑1969͒.
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:
130.64.175.185 On: Mon, 01 Dec 2014 14:54:41