A R T I C L E S
Palgrave and Parkin
that agglomeration has taken place. However, it is clear that it
is not the entire contents of an aerosol droplet that agglomerates.
spectra of the Au@SiO2 core-shell films showed peaks in
reflectivity in the red and near-IR regions, which were dependent
on particle separation. Mulvaney et al. successfully explained
this using the MG model of dipole interaction between the
particles.4 That this peak also appears in the reflectance spectrum
of the titania/gold nanocomposite films is strong evidence that
the gold particles therein are close enough to interact. As
mentioned above, the gold particles might acquire a thin titania
coating during the deposition process. If so, then this titania
layer would prevent gold particles fusing while allowing them
to come close enough to interact as described by the MG model.
The behavior of nanoparticles in the CVD reactor differs from
that of a conventional molecular precursor due to the relatively
large size of the nanoparticle. Gas-phase particles are subject
to a thermophoretic force when exposed to a temperature
gradient.35 This force is directed away from the hot surface, so
particles act as if repelled from a hot surface and attracted toward
a cold surface. This is usually of benefit in CVD as it prevents
particles (formed by gas-phase reaction of precursors or
otherwise present) from being incorporated into the growing
film. Since the flow of gas in the reactor is laminar rather than
turbulent, thermophoresis is usually the dominant force deter-
mining the deposition location of particles. It is therefore
unsurprising that when deposited alone, without an additional
precursor, the majority of gold nanoparticles are found on the
top plate. However, when deposited with an additional precursor,
gold particles are present throughout the substrate coating,
indicating that the presence of an additional precursor somehow
overcomes the thermophoretic effect. One possibility is that gas-
phase reaction of the precursor on the surface of the gold
nanoparticles produces coated gold particles, which are less
affected by the asymmetric bombardment of gas molecules than
smaller, uncoated particles. They can therefore find their way
to the substrate more easily. Nonetheless, the reduction in Au:
Ti atomic ratio from around 2:1 to around 1:25 in film 6 (as
quantified by EDX analysis) indicates that a large proportion
of the gold is lost during deposition. This indicates that the
majority of gold particles are deposited on the cold walls of
the reactor and the top plate, or pass from the reactor in the
exhaust, rather than deposit on the substrate.
In addition, the MG model predicts red shifting of the
plasmon peak with reduced particle separation, which might
explain the change in plasmon peak on deposition (Figure 3).
However, Mie theory provides an alternative explanation, that
the particles may fuse, creating larger particles, which also show
a red-shifted plasmon peak. Of course it may be that case that
both mechanisms are at work, and both contribute to the plasmon
red shift. Nonetheless, the further red shift on annealing is
probably better explained by the MG model. In the absence of
any gold ions in the sample (as shown by XPS), this must be
caused by either an agglomeration of particles or a reduction
in separation of dipole interacting particles. Annealing is known
to increase film density by eliminating pores and voids, and
thus would be expected to reduce particle separation. The MG
model can thus explain the red shifting of the plasmon peak on
annealing of the films. The absence of any further red shift after
80 min of annealing represents the elimination of all the voids
and the limit of the mobility of the particles within the film.
Potential Applications. A wide range of applications has
been suggested for semiconductor/metal nanocomposites in
general, as outlined in the Introduction. In addition, several other
possibilities present themselves. The reflection properties of the
titania/gold films may lend themselves to heat mirror applica-
tions, since they reflect infrared light and transmit visible light.
Heat mirrors are used in solar control applications. In this case
the higher reflectivity in the near-IR would mean that a window
incorporating these particles would reflect away much of the
heat portion of solar radiation (which is most intense between
ca. 800 and 1500 nm). This would enable a reduction in solar
gain and a reduction in air conditioning costs. The optical clarity
of the films makes them suitable for use as window coatings
where tinted glass is required. This would be a valuable
alternative to body tintingsthe current practice of coloring the
whole pane of glasssas it is an expensive process to implement
on a glass float line. A coating that gave the same intensity in
color, and whose color could be varied by a simple change in
precursor concentration, may make tinted glass much less
expensive to produce. However, we believe that the main
application of this technique lies in the variety of films that
can be produced by varying the precursors and nanoparticles
used. One obstacle that must be overcome is the relatively low
efficiency of incorporation of nanoparticles into the films.
Film Color and Optical Properties. The color and optical
spectra are perhaps the most interesting features of the composite
films produced in this study, and these characteristics are
dominated by the plasmon resonance of the gold nanoparticles
incorporated within the films. Three observations are important
here: the red shifting of the plasmon peak on deposition, the
further red shifting on annealing of the films, and the reflection
peak in the red and near-IR regions. Mie theory is frequently
used to model the plasmon peak in gold and other metallic
nanoparticles. Mie theory is applicable to low particle concen-
trations, including isolated single nanoparticles,38 but breaks
down in very dense colloids, where significant dipole interaction
between the particles occurs.5,6 In this intermediate regime,
between nanoparticulate and bulk phases, the Maxwell-Garnett
(MG) model has been used to predict the plasmon resonance.5
We have considered which model best fits the observations cited
here, and whether the particles in the films behave in isolation
or interact.
Mulvaney et al. studied films of Au@SiO2 core-shell
particles on glass, the gold cores of which showed dipole
interaction.4 In their work, the MG model was used to
successfully predict the optical properties of dense arrays of
gold particles that were kept physically separate by their
surrounding layers of silica. This structure bears a resemblance
to the nanocomposite materials produced in this work, where
gold particles are also separated by a dielectric material. This
is especially true if, as discussed above, the gold particles acquire
a shell of TiO2 during the CVD process. The optical spectra
for both sets of films show similar features. The reflection
Conclusion
Metallic gold nanoparticles can be deposited as thin films
with little change in their size or shape, and without chemical
reaction, under AACVD conditions. They can also be deposited
in conjunction with conventional CVD precursors, resulting in
the formation of continuous, durable nanocomposite films
9
1596 J. AM. CHEM. SOC. VOL. 128, NO. 5, 2006