Nanometer-thick conformal pore sealing of self-assembled mesoporous silica by plasma-assisted atomic layer deposition

Article abstract of DOI:10.1021/ja061097d

On a porous substrate, regular atomic layer deposition (ALD) not only takes place on top of the substrate but also penetrates into the internal porosity. Here we report a plasma-assisted process in which the ALD precursors are chosen to be nonreactive unless triggered by plasma, so that ALD can be spatially defined by the supply of plasma irradiation. Since plasma cannot penetrate within the internal porosity, ALD has been successfully confined to the immediate surface. This not only gives a possible solution for sealing of porous low dielectric constant films with a conformal layer of nm-scale thickness but also enables us to progressively reduce the pore size of mesoporous materials in a sub-A/cycle fashion for membrane formation. Copyright

Full text of DOI:10.1021/ja061097d

Published on Web 08/09/2006
Nanometer-Thick Conformal Pore Sealing of Self-Assembled Mesoporous
Silica by Plasma-Assisted Atomic Layer Deposition
Ying-Bing Jiang,† Nanguo Liu,† Henry Gerung,† Joseph L. Cecchi,*,† and C. Jeffrey Brinker*,†,‡
Department of Chemical and Nuclear Engineering, UniVersity of New Mexico, Albuquerque, New Mexico 87131,
and Sandia National Laboratories, Albuquerque, New Mexico 87185
Received February 15, 2006; Revised Manuscript Received July 27, 2006; E-mail: cjbrink@sandia.gov; cecchi@unm.edu
As device dimensions in semiconductor integrated circuits (ICs)
continue to shrink, low dielectric constant (low k) materials are
needed as interlevel dielectrics (ILD) to mitigate issues caused by
reduced line widths and line-to-line spacings, such as increasing
RC delay. To satisfy the technical requirements established by the
microelectronics road map (where k values <2 are ultimately
specified), future generation ILD materials must incorporate poros-
ity. However, the pores, typically on the order of angstroms to a
few nanometers and interconnected at elevated porosities, can trap
moisture, gas precursors, and other contaminants in subsequent
processing steps, making practical pore-sealing techniques essential
to low k implementation.1-3
To be useful for microelectronics applications, a pore-sealing
coating must be conformal to the 3D topology of patterned ILD
films. In addition, at the 65 nm or smaller technology node, it must
be only several nanometers thick, so its impact on the overall ILD
k value is negligible.1 These requirements exclude most thin film
deposition techniques with the exception of atomic layer deposition
(ALD), for which the coatings are inherently conformal and
precisely controlled at subnanometer thicknesses. However, on a
porous substrate, regular ALD not only takes place on top of the
substrate but also penetrates into the internal porosity (see Sup-
Figure 1. Cross-sectional TEM images showing (a) conformal 5 nm thick
pore-sealing coating of SiO2 prepared on a patterned mesoporous low k
silica film by plasma-assisted ALD, and (b) enlarged image at the interface
between PA-ALD layer and the mesoporous film.
porting Information Figure 1), filling pores and drastically increasing
the effective ILD k value.4-6 Therefore, an approach capable of
localizing ALD to the ILD surface and allowing spanning of the
pores is needed.
pump to a base vacuum of 5 × 10-7 Torr. An RF coil surrounded
Generally, this is hard to achieve because ALD is a surface
the Pyrex tube for plasma generation. Samples were mounted in a
adsorption-based deposition process that takes place wherever gas
remote plasma zone for reduced ion bombardment and plasma-
precursor adsorption occurs, including throughout the complete
heating effects (see Supporting Information). Oxygen and TEOS
network of connected internal porosity. Short precursor exposure
(tetraethyl orthosilicate Si(OCH2CH3)4) were used as the precursors
times may reduce the ALD penetration depth, but with small ILD
for SiO2. In the absence of plasma, they remain unreactive at room
feature dimensions, even a 10 nm penetration depth means that a
temperature. These precursors were admitted into the reactor
substantial portion of the ILD porosity would be filled.
alternately via pneumatic timing valves. A constant Ar flow of 15
Here we report a plasma-assisted process in which ALD is
sccm was used as the carrier gas as well as the purging gas.
confined to the immediate surface, allowing pore sealing at minimal
The mesoporous silica thin film samples were prepared on silicon
ILD thickness. To date, several groups have demonstrated the use
substrates by evaporation-induced self-assembly.10 The nonionic
of plasmas to enhance the extent of ALD reactions and achieve
surfactant Brij-56 was used to direct the formation of a cubic
better film quality.7-9 In our work, the purpose of the plasma is to
mesostructure characterized by a continuous 3D network of
define the location of ALD. If ALD precursors are chosen to be
connected pores with diameters ∼2 nm (Figure 1b). These films
nonreactive unless ‘triggered’ by plasma, then ALD can be spatially
exhibit excellent mechanical strength and thermal stability, along
defined by the supply of plasma irradiation. In this regard, it is
with an isotropic k and low surface roughness, important for etching
important to recognize that the Debye length (µm) and the molecule
or chemical mechanical polishing. With ∼50 vol % porosity, the k
mean free path in a typical plasma (µm-mm) greatly exceed the
value was measured to be 2.42 (Supporting Information), consistent
pore dimension of a porous low k material (nm), thus plasma cannot
with effective medium predictions. Prior to PA-ALD, the samples
penetrate (and ALD cannot occur) within the internal porosity.
were patterned by interferometric lithography and etched with a
The experiments were carried out in a self-designed, home-built
plasma-assisted ALD (PA-ALD) system. The deposition chamber
was a 25 mm diameter Pyrex tube, evacuated by a turbomolecular
CHF3/Ar plasma to create 400 × 400 nm trenches (Figure 1a). Then
the photoresist and any residual organics were removed by oxygen-
plasma treatment.
Plasma-assisted ALD was performed by first introducing TEOS
vapor into the reactor, followed by Ar purging to obtain a monolayer
† University of New Mexico.
‡ Sandia National Laboratories.
9
11018
J. AM. CHEM. SOC. 2006, 128, 11018-11019
10.1021/ja061097d CCC: $33.50 © 2006 American Chemical Society

C O M M U N I C A T I O N S
On the basis of studies of plasma-enhanced CVD of SiO2 from
TEOS and O2 performed by Aydil et al.,13 where silanol groups
were identified after exposure of TEOS to an oxygen plasma, we
expect silanols to form similarly during PA-ALD. However, due
to the monolayer (or submonolayer) tSi-OH coverage, we cannot
quantify the extent of surface hydrolysis. Additionally, we expect
that the plasma could serve a catalytic role by generating nucleo-
philic oxo radicals, tSi-O•, that promote siloxane bond formation.
Apparently, at room temperature, the extent of these plasma-assisted
hydrolysis and condensation reactions is less than that for ammonia-
catalyzed hydrolysis and condensation reactions during conventional
room temperature ALD,11 explaining the low deposition rates.
Consistent with a low rate of siloxane bond formation is the highly
conformal and dense PA-ALD layer indicative of a reaction-limited
monomer-cluster growth processsin our case confined exclusively
to the plasma-activated surface.12
Here we emphasize PA-ALD as a means of sealing pores.
However, with the very high degree of thickness control that we
demonstrate (0.03 nm/cycle), which remains linear for at least 150
cycles, we envision that, prior to complete pore sealing, we will
progressively reduce the pore size of the mesoporous silica in a
sub-Å/cycle fashion. This combined with the thin PA-ALD layer
thickness could have very important implications for membrane
formation, where extremely thin inorganic films with precisely
controlled pore size could enable the synthesis of robust mimics
of natural ion or water channels of interest for sensors and water
purification.
Figure 2. TEM images demonstrating the pore-sealing effectiveness by
PA-ALD: (a) regular cross-sectional TEM image showing the mesoporous
sample treated by PA-ALD pore-sealing process and then exposed to TiO2
ALD conditions; (b) Ti-mapping image in the same area acquired with
electron-energy-loss image filtering mode.
(or submonolayer) of adsorbed precursor. RF power was then
delivered to the coil, creating an O2 and Ar plasma. The associated
radicals convert surface-adsorbed TEOS into reactive silanols and
may promote further conversion to siloxane. Following this, the
deposition chamber was purged again to remove the residual
gaseous products, completing one cycle; 150 cycles were performed,
each cycle requiring 5 s.
Figure 1a,b shows cross-sectional TEM images of the sample.
A 5 nm thick SiO2 coating is observed as the smooth dark rim
bordering the patterned mesoporous silica feature. Clearly, the
coating is conformal to the patterned morphology and uniform in
thickness. No penetration of the SiO2 into the porous matrix can
be observed. The measured k of the corresponding planar sample
equaled 2.49 (Supporting Information), consistent with minimal
penetration of the PA-ALD layer.
To verify the pore-sealing effectiveness, the PA-ALD-coated
sample was introduced into a conventional ALD reactor, where
we performed thermal TiO2 ALD, shown previously to infiltrate
surfactant-templated mesoporous silica. At 180 °C, the PA-ALD-
coated sample was treated with 100 thermal ALD cycles using TiCl4
and H2O as the precursors. Figure 2a is a regular cross-sectional
TEM image, where we observe two ALD layers. The inner, lighter
layer is the PA-ALD SiO2 coating, and the outer, darker layer is
the TiO2 thermal ALD coating. The mesoporous low k silica appears
completely unaffected, suggesting that TiCl4 and H2O cannot
penetrate through the PA-ALD SiO2 coating to form TiO2 in the
underlying porous silica matrix. This is further supported by the
Ti-mapping image in Figure 2b. The bright border in this image
represents the location of Ti and corresponds to the TiO2 overlayer
shown in Figure 2a. Comparing the Ti-mapping image (Figure 2b)
to the original regular TEM image (Figure 2a), no detectable TiO2
can be found beyond the PA-ALD SiO2 coating. Therefore, the
PA-ALD SiO2 coating, although only 5 nm thick, is sufficiently
dense and defect-free to seal the pores and protect the porous low
k silica from exposure to gaseous chemicals.
Concerning the mechanism of room temperature PA-ALD of
SiO2, we first note that the deposition rate is quite low, 0.03 nm/
cycle, compared to 0.07-0.08 nm/cycle measured by George et
al. for conventional NH3-catalyzed SiO2 ALD.11 Conventional ALD
uses multiple water/TEOS cycles, where water exposures hydrolyze
ethoxysilane bonds to form silanols and alkoxide exposure results
in condensation reactions to form siloxane bonds. As for the related
solution-based “sol-gel” reactions, hydrolysis and condensation
are bimolecular nucleophilic substitution reactions catalyzed by acid
or base.12 In PA-ALD, plasma exposure takes the place of
hydrolysis, activating the alkoxide surface toward TEOS adsorption.
Acknowledgment. This work was supported by the Army
Research Office Grant DAAD19-03-1-0227, DOE Basic Energy
Sciences, Air Force Office of Scientific Research Grant FA9550-
04-1-0087, the NIH Nanomedicine Center program, and the SNL
LDRD program. Sandia is a multiprogram laboratory operated by
Sandia Corporation, a Lockheed Martin Company, for the United
States Department of Energy’s National Nuclear Security Admin-
istration under Contract DE-AC04-94AL85000. We would also like
to thank Prof. S. Brueck at UNM-Center for HTM for access to
interferometric lithography.
Supporting Information Available: Additional supporting figures
and table. This material is available free of charge via the Internet at
http://pubs.acs.org.
References
(1) 2004 International Technology Roadmap for Semiconductors, Interconnect,
SIA.
(2) Peters, L. Semiconductor International, October 2005, 49-53.
(3) de Rouffignac, P.; Li, Z. W.; Gordon, R. G. Electrochem. Solid State
Lett. 2004, 7, G306-G308.
(4) Cameron, M. A.; Gartland, I. P.; Smith, J. A.; Diaz, S. F.; George S. M.
Langmuir 2000, 16, 7435-7444.
(5) Ek, S.; Iiskola, E. I.; Niinisto¨, L. Langmuir 2003, 19, 3461-3471.
(6) Travaly, Y.; Schuhmacher, J.; Baklanov, M. R.; Giangrandi, S.; Richard,
O.; Brijs, B.; Van Hove, M.; Maex, K.; Abell, T.; Somers, K. R. F. J.
Appl. Phys. 2005, 98, 083515.
(7) Elers, K. E.; Winkler, J.; Weeks, K.; Marcus, S. J. Electrochem. Soc.
2005, 152, G589-G593.
(8) Kim, D. H.; Kim, Y. J.; Song, Y. S.; Lee, B. T.; Kim, J. H.; Suh, S.;
Gordon, R. J. Electrochem. Soc. 2003, 150, C740-C744.
(9) Kim, H.; Detavenier, C.; van der Straten, O.; Rossnagel, S. M.; Kellock,
A. J.; Park, D. G. J. Appl. Phys. 2005, 98, 014308.
(10) Brinker, C. J.; Lu, Y. F.; Sellinger, A.; Fan, H. Y. AdV. Mater. 1999, 11,
579-585.
(11) Ferguson, J. D.; Smith, E. R.; Weimer, A. W.; George, S. M. J.
Electrochem. Soc. 2004, 151, G528-G535.
(12) Brinker, C. J.; Scherer, G. W. Sol-Gel Science: The Physics and
Chemistry of Sol-Gel Processing; Academic Press: San Diago, 1990.
(13) Deshmukh, S.; Aydil, E. J. Vac. Sci. Technol. A 1995, 13, 2355-2367.
JA061097D
9
J. AM. CHEM. SOC. VOL. 128, NO. 34, 2006 11019