COMMUNICATION
www.rsc.org/loc | Lab on a Chip
Accelerated synthesis of titanium oxide nanostructures using
microfluidic chips{
Ben F. Cottam, Siva Krishnadasan, Andrew J. deMello, John C. deMello and Milo S. P. Shaffer*
Received 3rd November 2006, Accepted 15th December 2006
First published as an Advance Article on the web 22nd December 2006
DOI: 10.1039/b616068a
The synthesis of one-dimensional titanium oxide nanostructures has been accelerated by performing the reaction in a
microfluidic environment as opposed to a classical batch
process.
In recent years, the synthesis of nanoscale rods and tubes of
titanium dioxide (TiO2) has attracted much attention, motivated
by the wide range of applications in dye-sensitised solar cells,1–3
photocatalysis,4,5 sensors6 and photochromic devices.7 Equiaxed,
nanoparticulate titania is used commercially on a vast scale but is
typically polycrystalline, and contains a mixture of sizes and even
phases. In order to produce more optimal materials, considerable
efforts have been made to synthesise high-purity, monodispersed,
single crystal TiO2 particles.8–15 Recently, it has emerged2,5 that
one-dimensional (1-D) or high aspect ratio TiO2 nanoparticles are
particularly interesting; the high anisotropy, network forming
abilities, and high surface-to-volume ratio offer advantages in
many applications. Various synthetic routes have been explored;
the most relevant example is the two-phase synthesis, by Cozzoli
et al.,10 of TiO2 nanorods by rapid hydrolysis of titanium
tetraisopropoxide (TTIP), using oleic acid (OLEA) as a surfactant
and solvent.
The use of continuous-flow, microfluidic reactors as environments in which to perform synthetic processes confers several
advantages over conventional ‘macroscale’ techniques. In simple
terms, the high surface area-to-volume ratios and reduced
diffusional dimensions that characterize microfluidic systems allow
reactions to be performed in a rapid and controllable manner.
Local variations in reaction conditions (such as concentration and
temperature) are minimized and therefore control of both
nucleation and particle growth can be used to improve
monodispersity and efficiency in nanoparticle synthesis.16–23 Such
improvements have been reported for both metallic and inorganic
nanoparticles,17,18 including gold,19 CdS,20 CdSe21 and TiO2.22,23
Wang et al.22 synthesised anatase TiO2 nanoparticles by means of
a two-phase reaction, however the products were not extensively
characterised. To date, there have been no reported attempts to
synthesise high aspect ratio, 1-D nanostructures using microfluidic
systems.
Employing the hydrolysis of TTIP technique developed by
Cozzoli et al., we report a comparison of the TiO2 nanostructures
obtained in classical batch reactions with those produced in
Department of Chemistry, Imperial College London, South Kensington,
London, UK. E-mail: [email protected]; Fax: +44 2075945801;
Tel: +44 0275945825
{ Electronic supplementary information (ESI) available: Yield data and
experimental setup. Raman, XRD and thermogravimetric analysis data.
Additional TEM images. See DOI: 10.1039/b616068a
This journal is ß The Royal Society of Chemistry 2007
microfluidic channels. In a typical batch synthesis, 50 g of oleic
acid (OLEA, Aldrich tech y90%) was dried under vacuum, at
120 uC and for 1.5 h, after which it was cooled to 90 uC under
nitrogen. 2.33 mL of titanium tetraisopropoxide (TTIP, Aldrich,
97%) (8.18 mmol) was then added and, upon stirring for 5 min,
turned yellow (solution A). 29.7 mL of a 0.5 M trimethylamine
N-oxide dehydrate (TMAO, Fluka, purum ¢ 99%) (aq.) (solution
B) solution was rapidly injected into solution A and stirred
vigorously for 22 h, under nitrogen. Aliquots were removed after 1,
1.5, 2, 3, 4 and 6 h of the reaction. Processing was identical for all
samples: upon addition of ethanol a white precipitate was obtained
which was then centrifuged (4000 rpm) and washed several times
in ethanol. No product was isolated from the aliquots removed
after 1 and 1.5 h.
TiO2 nanorods were produced with a yield of y85% after 6 h
and y88% after 22 h (see ESI{ for full yield plot). Nanorods
formed after 6 h were found to be 25.9 ¡ 5.4 nm in length and
3.3 ¡ 0.7 nm in diameter; Fig. 1A shows a high-resolution
transmission electron microscope (HRTEM) image of a typical
structure. Only 1D structures were observed after a reaction time
of 6 h. When the reaction was carried out for 22 h, branched
nanorods with Y- and H-shaped structures were observed; a
typical H-shaped branch is shown in Fig. 1b (further examples in
ESI{). The branches themselves are of a similar diameter
(y3.3 nm) to the individual nanorods. The fast Fourier transform
(FFT) (Fig. 1B, inset) is consistent with single-crystal anatase with
the [001] direction parallel to the rod axis. The majority of the faces
appear to be {101} type.
Under these conditions, branched structures had not previously
been reported, Cozzoli et al. having stated that ‘‘prolonged heating
[over 6 h] did not further influence the particle growth.’’10 Since the
TiO2 yield only increases slightly between 6 h (85%) and 22 h
Fig. 1 HRTEM images of typical nanostructures formed after (A) 6 h
and (B) 22 h (inset shows the FFT). Both scale bars are 10 nm.
Lab Chip, 2007, 7, 167–169 | 167
(88%), these branched structures must be formed by the oriented
attachment of individual rods,8 rather than continued growth.11
Oriented attachment of rods is driven by a net reduction in surface
energy, and is thought to involve a topotactic attachment of
primary particles at the {101} surfaces, in the absence of sufficient
mobility for wholesale coarsening of the structure.
Raman spectroscopy, powder X-ray diffraction (XRD) and
thermogravimetric analysis (TGA) were performed on samples
removed from the reaction after 3, 6, and 22 h (see ESI{); the
Raman and XRD analyses confirm that the anatase polymorph is
present in all samples (the slight features in the XRD patterns at
29u and 45u are thought to be due to a small fraction of partially
condensed hydrous TiO2) and indicate an increase in crystallinity
for longer reactions. In addition, the TGA data indicate a lower
surfactant content with increasing reaction time (38.1% organic
content after 6 h reaction time, compared with 27.1% after 22 h);
this result is consistent with a reduction in surface area either due
to the oriented attachment process or the generation of a smoother
surface as crystallinity improves.
An analogous reaction was subsequently performed using a
microfluidic process, the reaction solutions being prepared in the
same manner as in the bulk experiments as described above.
Solutions A and B were concurrently pumped into a two-input/
one-output y-shaped microfluidic chip, a schematic of which is
shown in Fig. 2 (channel dimensions: 60 mm depth 6 100 mm
width 6 40 cm length, corresponding to a volume of y17.0 mL;
details of chip fabrication can be found in the ESI{) at 1.2 mL min21
and 0.6 mL min21 respectively (this input ratio of 2 : 1 matched the
reactant ratio in the bulk reactions).
At this flow rate the reactants reside for an average of 10 min in
the microchannels. The chip was placed in an oven, set at 90 uC;
the use of an oven rather than a more conventional hot plate
provided improved thermal stability and reproducibility. The flow
regime in the chip was plug-flow, with y100 mm plugs of each
solution alternating along the length of the microchannel. When
the reaction solution exited the channel it was passed to a two-way
stream splitter and the product was collected in air (unquenched)
or quenched by cooling using dry ice, which is sufficient to freeze
both the water and oleic acid phases. The reaction solutions were
collected for 20 h. The product was then quickly isolated, by
precipitation and washing, as described above.
Both the quenched and unquenched solutions resulted in
broom-like aggregates of anatase nanorods (Fig. 3) with a yield
Fig. 2 Schematic diagram of the microfluidic chip used for the synthesis
of TiO2 nanostructures.
168 | Lab Chip, 2007, 7, 167–169
Fig. 3 HRTEM image of a bunched TiO2 nanostructure composed of
anatase nanorods, collected after quenching in dry ice. Insets display the
SAED pattern and a low resolution image of the nanostructure.
of 35%; the similarity of the products suggests that they grow on
chip. The primary nanorods are of the same diameter, y3.3 nm as
those observed in the batch reactions. The selected area electron
diffraction pattern (SAED) confirms that the aggregates are
anatase and are aligned to within y30u. The constituent nanorods
appear to be single crystals.
XRD spectra of the TiO2 products obtained from batch and
micro-scale reactions are compared in Fig. 4A. The solid
spectrum is the sample obtained from a batch reaction after 6 h
(analysed on a standard powder XRD diffractometer); the peaks
can be indexed to the standard anatase pattern. The dashed
spectrum represents a quenched microfluidic chip sample
(analysed at a mg scale using a single crystal diffractometer), it
also displays the principal anatase peaks. Note the relatively
sharp (004) peak in the batch sample which indicates that the
axis of the rods lies parallel to [001], as observed in the electron
microscopy and in earlier work.10 Systematic broadening, due to
the difficulties associated with handling a small sample, obscures
the effect for the microfluidic product.
Fig. 4 (A) XRD patterns of bulk (solid line) and microfluidic (dashed)
TiO2 nanorods synthesised without quenching, displayed with a standard
anatase diffraction file on the base-line. (B) Raman spectra comparing
(dashed) a drop of reaction solution directly after exiting the microfluidic
chip and (solid) the reaction solution after a 6 h batch reaction.
This journal is ß The Royal Society of Chemistry 2007
In order to confirm further that the nanorod synthesis occurs on
chip, ‘wet’ Raman analysis was also performed on a drop of
reaction solution immediately upon exit (Fig. 4B, dashed). This
spectrum was compared to that obtained from a solution obtained
from a 6 h batch reaction (Fig. 4B, solid). Since neither sample had
been washed or processed in any way, oleic acid peaks (*, y240,
595 and 720 cm21) are still present. The three principle anatase
peaks (A, y397, 510 and 650 cm21) can just be observed. Due to
the similarity of these spectra and of the various products, it can be
concluded that anatase nanorods are being indeed synthesised on
the chip, and that the synthesis is continuous in nature.
When compared, the reaction rate is significantly higher on-chip
than in a batch reaction. After only 10 min residence on the chip,
crystalline nanorods are observed, whilst no sample can be isolated
before y90 min of a batch reaction. It is thought this increase in
reaction rate is predominantly due to the rapid mixing conditions
in the microchannels. In a bulk reaction, the phase domains
(OLEA/TTIP and TMAO (aq)) are millimetre-sized and complete
mixing of the reagents is relatively slow. On the ‘chip’, the linear
dimensions of the domains are at least an order of magnitude
smaller, resulting in a higher interfacial area, and short diffusion
distances. The rapid mixing may be particularly significant, as the
rate of reagent addition was previously reported to control product
morphology.10 Another explanation for the increased reaction rate
on-chip is that, due to the high heat transfer, the reaction solutions
reach the reaction temperature almost immediately. In the batch
reactions, the addition of solution B (at room temperature), lowers
the temperature of the mixture significantly. However, since the
mixture returns to the intended reaction temperature within a few
minutes, this effect cannot explain the observed change in overall
rate, but may affect nucleation. In the present case, rapid
nucleation at the large areas of interface is likely to lead to local
high concentrations of nanorods that subsequently aggregate.
Unlike the slow, oriented attachment observed at long times in the
batch reaction, the bunched aggregates obtained on chip are not
crystallographically coherent, but are loosely aligned along their
common axis.
We have presented a comparison of the growth of anatase TiO2
nanorods in microfluidic and bulk processes. It has not been
possible to assess fully any effects on monodispersity due to the
formation of bunched agglomerates in the microfluidic product.
However, reaction rates were found to be increased approximately
tenfold under microfluidic conditions. The use of a number of
microfluidic chips in parallel may, therefore, offer increased
production rates as well as new product assemblies. Currently,
This journal is ß The Royal Society of Chemistry 2007
the products are terminated with oleic acid groups, but these can
be substituted by other groups as desired.10 The study also shows
that high aspect ratio particles can be synthesised on chip without
blocking, and hence that the method is potentially applicable to a
wide variety of 1-D nanomaterial synthesis. Ideally, conditions will
be adjusted to produce individual nanorods, however, aligned,
high surface area bundles may, in fact, be useful for some
applications; for example, in dye-sensitised solar cells, aligned
aggregates may offer improved electron transport and high surface
areas.
Notes and references
1 B. O’Regan and M. Grätzel, Nature, 1991, 353, 737.
2 A. Petrella, M. Tamborra, P. D. Cozzoli, M. L. Curri, M. Stricolli,
P. Cosma, G. M. Farinola, F. Babudri, F. Naso and A. Agostiano, Thin
Solid Films, 2004, 451–452, 64.
3 J. Jiu, S. Isoda, F. Wang and M. Adachi, J. Phys. Chem. B, 2006, 110,
2087.
4 J. Joo, S. G. Kwon, T. Yu, M. Cho, J. Lee, J. Yoon and T. Hyeon,
J. Phys. Chem. B, 2005, 109, 15297.
5 P. D. Cozzoli, R. Comparelli, E. Fanizza, M. L. Curri and A. Agostiano,
Mater. Sci. Eng., C, 2003, 23, 707.
6 Y. Zhu, J. Shi, Z. Zhang, C. Zhang and X. Zhang, Anal. Chem., 2002,
74, 120.
7 K. Iuchi, Y. Ohko, T. Tatsuma and F. Fujishima, Chem. Mater., 2004,
16, 1166.
8 R. L. Penn and J. F. Banfield, Geochim. Cosmochim. Acta, 1999, 63,
1549.
9 Y. C. Zhu and C. X. Ding, Nanostruct. Mater., 1999, 11, 427.
10 P. D. Cozzoli, A. Kornowski and A. Weller, J. Am. Chem. Soc, 2003,
125, 14539.
11 Y.-w. Jun, M. F. Casula, J.-H. Sim, S. Y. Kim, J. Cheon and
A. P. Alivisatos, J. Am. Chem. Soc., 2003, 125, 15981.
12 Q. Huang and L. Gao, Chem. Lett., 2003, 32, 638.
13 Z. Zhang, X. Zhong, S. Liu, D. Li and M. Han, Angew. Chem., Int. Ed.,
2005, 44, 3466.
14 K. Yu, J. Zhao, X. Zhao, X. Ding, Y. Zhu and Z. Wang, Mater. Lett.,
2005, 59, 2676.
15 S. Yang and L. Gao, Chem. Lett., 2005, 34, 964.
16 S. Xu, Z. H. Nie, M. Seo, P. C. Lewis, E. Kumacheva, H. A. Stone,
P. Garstecki, D. B. Weibel, I. Gitlin and G. M. Whitesides, Angew.
Chem., Int. Ed., 2005, 44, 724.
17 A. J. deMello and J. C. deMello, Lab Chip, 2004, 4, 11N.
18 A. J. deMello, Nature, 2006, 442, 394.
19 J. Wagner and J. M. Kohler, Nano Lett., 2005, 5, 685.
20 J. B. Edel, R. Fortt, J. C. deMello and A. J. deMello, Chem. Commun.,
2002, 1136.
21 E. M. Chan, A. P. Alivisatos and R. A. Mathies, J. Am. Chem. Soc.,
2005, 127, 13854.
22 H. Wang, H. Nakamura, M. Uehara, M. Miyazaki and H. Maeda,
Chem. Commun., 2002, 1462.
23 M. Takagi, T. Maki, M. Miyahara and K. Mae, Chem. Eng. J., 2004,
101, 269.
Lab Chip, 2007, 7, 167–169 | 169