Anal. Chem.

2005, 77, 7462-7471

Technical Notes

Deposition Method for Preparing SERS-Active Gold

Nanoparticle Substrates
Kiang Wei Kho,†,‡ Ze Xiang Shen,‡ Hua Chun Zeng,§ Khee Chee Soo,† and Malini Olivo*,†

Division of Medical Sciences, National Cancer Centre, 11 Hospital Drive, 169610, Singapore, Department of Physics, and
Department of Chemical and Biomolecular Engineering, Faculty of Engineering, National University of Singapore,
Lower Kent Ridge Road, Singapore

Surface-enhanced Raman scattering or SERS, discovered intensities.2 Experimental and theoretical studies have shown that
some 20 years ago, has recently become a promising tool an enhancement in the order of from 103 to 106 is attainable in
for routine biofluid assays in a clinical setting. Many SERS,8 giving a detection limit up to the subpicogram range.8,9
attempts have been made to produce cheap and repro- As such, SERS has recently become a promising analytical tool
ducible SERS-active substrates. In this study, we report in the medical field1 and has been used for the study of trace
on the fabrication of SERS-active substrates through the analytes (e.g., urea, critic acid) and chemical changes in blood
convective assembly of gold (Au) particles on electrostati- and urine.10,11
cally charged glass slides. We show that, by a proper For all the SERS-active metallic surfaces that have been
control of the initial particle concentration in an evaporat- developed and studied, it has been amply demonstrated that
ing Au suspension droplet, it is possible to obtain a closely surfaces consisting of closely packed but not aggregated colloidal
packed colloidal film capable of generating SERS activity. arrays are particularly enhancing, owing primarily to the inter-
Finally, AFM and SERS measurements of the resulting particle plasmon resonance.6 Thus, many attempts have been
films reveal comparability in performance with previous made to prepare closely packed colloidal films. Currently, available
silane-immobilized Au colloidal films. The minimum fabrication techniques encompass vapor deposition of nanopar-
electromagnetic enhancement factor of our films is esti- ticles on silica posts,12 fabrication of periodic arrays of nanopar-
mated to be about 2 × 104. ticles via nanosphere lithography,13 electron beam lithography,14
self-assembly of nanoparticles on a chemically functionalized solid
Surface-enhanced Raman scattering (SERS) has been exten- surface,6 and silver enhancement of seed particles that have been
sively studied since Van Duyne and Creighton’s discovery of predeposited on glass substrates.15 While these substrates have
Raman scattering enhancement of molecules adsorbed on rough- been shown to exhibit large Raman enhancement and good SERS
ened metallic surfaces some 20 years ago.1-6 Subsequent experi- reproducibility, most are, unfortunately, too laborious and expen-
mental and theoretical studies have attributed this observation to sive to fabricate in large quantities. For instance, nanosphere
the light-induced surface plasmon resonance (SPR) on the metallic lithography requires the use of a thermal evaporator equipped
surfaces. The result of SPR is the creation of “hot” electric field with an accurate thickness monitor so as to ensure proper
spots on the curved surfaces as well as in the troughs between deposition thickness of metal over the nanosphere mask precoated
nanoprotrusions.7 Consequently, molecules that are trapped within on the substrate.6 Electron beam lithography is an alternative,14
or situated in the vicinity of these “hot” zones would experience but this is expensive and labor intensive. To this end, Grabar and
a strong excitation field and in turn emit amplified Raman co-workers have devised a simpler and cheaper approach in which

†
National Cancer Centre. (8) Van-Duyne, R. P.; Kurt, L. H.; Robert, I. A. Chem. Phys. Lett. 1986, 126,
‡
Department of Physics, National University of Singapore. 190-196.
§
Department of Chemical and Biomolecular Engineering, National Uni- (9) Dou, X.; Takama, T.; Yamaguchi, Y.; Yamamoto, H. Anal. Chem. 1997,
versity of Singapore. 69, 1492-1495.
(1) Jeanmaire, D. L.; Van-Duyne, R. P. J. Electroanal. Chem. 1977, 84, 1-20. (10) Oneal, P. D.; Cote, G. L.; Motamedi, M.; Chen, J.; Lin, W. C. J. Biomed.
(2) Albrecht, M. G.; Creighton, J. A. J. Am. Chem. Soc. 1977, 99, 5215-5217. Opt. 2003, 8 (1), 33-39.
(3) Bjernald, E. J.; Foldes-Papp, Z.; Kall, M.; Rigler, R. J. Phys. Chem. B 2002, (11) Sulk, R.; Chan, C.; Guicheteau, J.; Gomez, C.; Heyns, J. B. B.; Corcoran,
106 (6), 1213-1218. R.; Carron, K. J. Raman Spectrosc. 1999, 30 (9), 853-859.
(4) Mulvaney, P.; He, L.; Natan, M. J.; Keating, C. D. J. Raman Spectrosc. 2003, (12) Vo-Dinh, T.; Hiromoto, M. Y. K.; Begun, G. M.; Moody, R. L. Anal. Chem.
34 (2), 163-171. 1984, 56 (9), 1667-1670.
(5) Podstawka, E.; Ozaki, Y.; Proniewicz, L. M. Appl. Spectrosc. 2004, 58 (5), (13) Haynes, C. L.; Van-Duyne, R. P. J. Phys. Chem. B 2001, 105 (24), 5599-
570-580. 5611.
(6) Grabar, K. C.; Freeman, R. G.; Hommer, M. B.; Natan, M. J. Anal. Chem. (14) Felidj, N.; Aubard, J.; Levi, G.; Krenn, J. R.; Hohenau, A.; Schider, G.; Leitner,
1995, 67 (4), 735-743. A.; Aussenegg, F. R. Appl Phys Lett. 2003, 82 (18), 3095-3097.
(7) Sanchez-Gil, J. A.; Garcia-Ramos, J. V.; Mendez, E. R. Opt. Express 2002, (15) Li, X.; Xu, W.; Jia, H.; Wang, X.; Zhao, B.; Li, B.; Ozaki, Y. Appl. Spectrosc.
10 (17), 879-886. 2004, 58 (1), 26-32.

7462 Analytical Chemistry, Vol. 77, No. 22, November 15, 2005 10.1021/ac050437v CCC: $30.25 © 2005 American Chemical Society
Published on Web 10/08/2005
nanoparticles are self-assembled on chemically functionalized significantly improving the film’s stability and hence the SERS
(with organosilanes) glass surfaces.6 While this particular tech- performance. No oxidizing cleaning solution is needed in this case
nique is relatively simpler, it is unfortunately compromised by the as the glass slide used here is a commercially available product.
requirement that the glass slides be thoroughly cleaned with a Uniform particle layers can be readily identified after the solvent
highly oxidizing cleaning solution (e.g., piranha solution) before has completely evaporated. A comparison of the AFM micrographs
derivatization with colloid can be carried out. Extreme care must for our films and those previously reported for chemically
be taken when handling the solution as it normally reacts immobilized colloidal films reveals qualitatively similar surface
explosively with organic substances such as acetone.6 Moreover, morphology in that the majority of the constituent particles are
the solution can also generate toxic waste products, which, in closely packed but not aggregated. Subsequent SERS measure-
some cases, can be carcinogenic to humans. Last, the solution ments of our films showed comparable SERS reproducibility and
must be prepared freshly each time, rendering it too costly to enhancement factor to those chemically immobilized colloidal
scale up for large numbers. As such, we seek to find a safer and films. Therefore, we believe that our deposition technique may
cheaper alternative to fabricating SERS-active closely packed be the simplest alternative for preparing SERS-active substrates.
colloidal films. More specifically, we ask whether a stable layer
of closely packed Au particles can be obtained through the EXPERIMENTAL SECTION
convective assembly of suspended Au colloid. While convection- Materials. A 15 nm Au hydrosol (EMGC.15, particle concen-
induced ordered packing of dielectric particles (e.g., latex) with tration ) 1.4 × 1012 particles per mL, %CV < 10%) was purchased
sizes between 42 nm and 1.1 µm has been extensively studied,16-18 from British Bio-cell International and used as received. This
convective assembly of small, charged-stabilized metallic colloids colloid was stored at 4 °C if not in use. Ethanol was purchased
(5-15 nm) has been shown to be more problematic due to the from Merck. Distilled water (18 MΩ) was obtained from a
particles’ large Hamaker constant,6,18-22 which makes them Millipore Mili-Q water purification system. SuperFrostPlus mi-
relatively unstable. Thus, arrays of closely packed metallic particles croscope glass slides were purchased from VWR Scientific. Plastic
can form only if the salt content in the solvent is sufficiently low containers for saliva collection (15 mL) and centrifuge tubes (1.5
or completely removed and/or the particle is protected with a mL) were purchased from Eppendorf. Crystal violet (CV) was
layer of surfactant molecules or ions. Examples of a closely packed purchased from Sigma-Aldrich and silica gel from Chemicod.
Au colloidal film would be those reported in the studies by Giersig Au Hydrosol Preparation. Au hydrosols containing different
and Mulvaney23,24 and Zhao et al.,21 who deposited Au particles particle concentrations were prepared as follows: To increase the
on TEM grids via electrophoretic means and convective assembly, particle concentration to 1.9 × 1012 particles per mL, 1 mL of the
respectively. Unfortunately, as has been pointed out by Michael, Au hydrosol was pipetted into a 2 mL centrifuge tube and spun at
colloidal films prepared on TEM grids are not stable and would 14 000 rpm for 5-10 min. About 300 µL of the supernatant was
tend to break or crack during handling, owing to the softness of then discarded before resuspending the Au particles with a vortex
the grid.24 Furthermore, particles adsorbed on these grids could mixer. To decrease the particle concentration to 3.5 × 1011
be resuspended very easily, since they are not immobilized particles per mL, 250 µL of the Au hydrosol was pipetted in a 2
chemically.23 As such, the above colloidal films are unapt for SERS mL centrifuge tube, and then distilled water was added to a final
applications. On the other hand, Prevo and Velev studied the volume of 1 mL.
deposition of a metallic colloidal layer on a solid glass substrate SERS Substrate Preparation. SuperFrostPlus microscope
via convective assembly of 15 nm Au suspension in the wetting glass slides were rinsed thoroughly with 70% v/v aqueous solution
film of a moving meniscus.18 Yet, in contrast to Zhao et al.’s report, of ethanol for 1 min. The slides were then rinsed 4 times, to
Prevo and Velev did not observe any organized lattice or closely remove traces of ethanol, with fresh distilled water each time.
packed particles; instead the particles formed micrometer-sized The slides were stored in water until needed. To fabricate a Au
aggregates with ill-defined dimensions,18 which is likely a result colloidal film, a Au hydrosol droplet of about 150 µL in volume
of high particle concentration in the colloid meniscus. was pipetted onto the surface of a cleaned slide. The slide was
Here, we show that through a proper control of the initial subsequently dried in a desiccator for 24 h, followed by vigorous
particle concentration in a drying colloidal droplet, it is possible rinsing in distilled water, and finally dried with a dust blower. All
to obtain a stable, closely packed colloidal film capable of SERS of the prepared Au colloidal films were then kept in distilled water
activity. In addition, we use an electrostatically charged glass slide or a dust-free environment until needed.
as the supporting substrate. This allows the Au particles to be Preparation of CV Solutions. CV solutions with concentra-
firmly immobilized via electrostatic attraction forces, thereby tions of 1 µM, 100 µM, and 1 mM were prepared in phosphate-
buffered saline (PBS) pH 7.2.
(16) Micheletto, R.; Fukuda, H.; Ohtsu, M. Langmuir 1995, 11, 3333-3336. Instruments. Optical transmission measurements of the
(17) Ng, V.; Lee, Y. V.; Chen, B. T.; Adeyeye, A. O. Nanotechnology 2002, 13, resultant Au colloidal films were performed using a Shimadzu
554-558.
(18) Prevo, B. G.; Velev, O. D. Langmuir 2004, 20 (6), 2099-2107. UV-2401 PC monochromator system. A cleaned SuperFrostPlus
(19) Dimon, P.; Sinha, S. K.; Weitz, D. A.; Safinya, C. R.; Smith, G. S.; Varady, microscope slide was used as the reference sample. Each slide
W. A.; Lindsay, H. M. Phys. Rev. Lett. 1986, 57, 595-598. bearing a Au film was placed in the illumination light path and
(20) Weitz, D. A.; Lin, M. Y.; Sandroff, C. J. Surf. Sci. 1985, 158 (1-3), 147-
164. secured with blue-tack. The light beam was passing through the
(21) Zhao, S.-Y.; Wang, S.; Kimura, K. Langmuir 2004, 20 (5), 1977-1979. slide perpendicularly. Care was taken to ensure that the blue-
(22) Diao, J. J.; Qiu, F. S.; Chen, G. D.; Reeves, M. E. J. Phys. D: Appl. Phys. tack was not in the light path.
2003, 36 (3), L25 - L27.
(23) Giersig, M.; Mulvaney, P. Langmuir 1993, 9 (12), 3408-3413. Atomic force microscopy (AFM) measurements were per-
(24) Giersig, M.; Mulvaney, P. J. Phys. Chem. 1993, 97 (24), 6334-6336. formed under ambient conditions using a Digital Instrument
Analytical Chemistry, Vol. 77, No. 22, November 15, 2005 7463
DI3000 Nanoscope III in tapping mode (247.9 kHz) with a typical closely packed particles are normally SERS-active.6,15,27,28 They
lateral and vertical resolution of about 5 and 1 nm, respectively. derive their SERS activity mainly from the crevices between two
The horizontal scanning rate is 1 Hz. adjacent particles, in which trapped analyte molecules experience
The Raman experiments were carried out with a modified enhanced electromagnetic field due to interparticle surface plas-
micro-Raman system, in which an Olympus microscope with a mon resonance.3,29-31 A number of methods have been devised
color closed circuit television (CCTV) system is coupled to a Spex to achieve close particle spacing as discussed earlier.4,6,15,27,28 In
1704 spectrometer that is equipped with a liquid nitrogen cooled this study, we investigate colloidal films deposited by evaporating
CCD detector.25 In this modified system, the coupling optics are Au suspension droplets on electrostatically charged glass slides.
arranged in a box that is linked to the modified optical microscope The typical image of circular colloidal films obtainable when the
by a mechanical arm and fixed to the Spex spectrometer. The solvent has completely evaporated is shown in Figure 1a. These
laser light (632.8 nm) is introduced from the back of the box, films exhibit a characteristic “coffee-ring” feature. A magnified
after passing through a plasma filter, and is then directed into a image, taken using a stereomicroscope, of a portion of the one of
microscope via a notch filter, which acts as a reflection mirror to the films is shown in Figure 1b. One can clearly see two distinct
the laser light, but to the signal returning from the sample, it acts features, a ring and a uniform colloidal Au layer, indicating that
as a very effective rejection filter to the Rayleigh (laser) line. This there are two different stages in the evaporation process. Initially,
filter prevents any backscattered laser from entering the spec- the ring is formed when the contact line (the intersection at which
trometer, and hence from interfering with the Raman signals to the droplet surface touches the solid substrate) is pinned down
be collected, but allows Raman signals to transmit with little by nanoparticles jammed between the liquid surface and the glass
attenuation. The returned signals then pass through a second surface. Since the contact line cannot shrink at this stage, an
notch filter which is used to further improve the Rayleigh rejection. outward flow of liquid from the interior is induced, as a result, to
The Raman signal is then focused onto the entrance slit of the replenish the evaporative losses along the perimeter of the
Spex spectrometer by a coated singlet focusing lens of 50 mm droplet.18,32-34 Owing to the viscosity of the solvent, the flow carries
focal length. The inclusion of the CCTV system allows both the the suspended particles toward the pinned contact line, bringing
laser beam and white light to be viewed directly from the monitor. about accumulation of particles there, i.e., ring formation. But as
A biological specimen can also be viewed directly from the the volume of the drying droplet continues to reduce so does the
monitor. A 20 × 0.4 NA objective lens purchased from Olympus contact angle. Eventually, the surface tension will balance out the
was used in this study. pinning force, causing the contact line to pull off from the ring.
SERS Measurements. A droplet (approximately 30 µL) of a This is the start of the second stage, where the contact line slowly
CV solution is pipetted onto a dried Au colloidal film. By adjusting retracts toward the drop center, creating a continuous colloidal
the microscope stage, the 633 nm laser was focused, via a 20 × layer that extends from the ring deposit. We shall show later that
0.4 NA objective lens, through the sample solution and onto the by properly controlling the initial particle concentration in the
sample-film interface. Spectrum acquisition was then started drying droplet, a closely packed but nonaggregated colloidal layer
immediately. Laser intensity at the focal spot was about 2 mW. can be obtained. Note that the deposit pattern observed in the
The spectral resolution was 1 cm-1, and an integration time of 20 current experiment is consistent with that reported by Deegan et
s was used for all Raman measurements. al. in their study of dried liquid droplets containing small latex
The focus spot size was defined as the area enclosed by the beads.33
first dark ring of the airy disk, the diameter of which can be Further analysis of our films shows that a good film reproduc-
calculated using the formula, 1.22λ/η sin R,26 where λ is the laser ibility is attainable with the current deposition technique. One can
wavelength, η is the refractive index of the solvent, and R is the readily obtain a uniform colloidal film of at least 6 mm2 in size.
half-angle of the focusing light cone. The diameter and area of This can be seen from Figure 1b. Note that the image has been
the focus spot are 1.48 µm and 1.72 µm2, respectively. processed to correct for any intensity variation arising from uneven
Unenhanced Raman Measurements. A droplet (approxi- illumination over the imaged area. From the histogram shown in
mately 30 µL) of a 100 µM CV solution is pipetted onto a cleaned Figure 1c, the mean intensity of the enclosed area is calculated
microscope glass slide. By adjusting the microscope stage, the to be 198.66 (mean background intensity ) 206.94) with a standard
633 nm laser was focused, via the objective lens, onto the liquid deviation (STD) of about 2.92, which is comparable to the 2.41
surface of the droplet. Spectrum acquisition was then started calculated for the uncoated glass surface (on the left of the ring).
immediately. Laser intensity at the focal spot was 2 mW. The We attribute these low STD values to noise in the image. A large
spectral resolution was 1 cm-1, and the integration time was 20 s. (27) Olson, L. G.; Lo, Y.-S.; Beebe, T. P., Jr.; Harris, J. M. Anal. Chem. 2001,
Again by making use of the diffraction theory, we calculated a 73, 4268-4276.
probe volume of about 15 fL at the focus. Therefore, about 1.5 × (28) Zhang, J.; Li, X.; Liu, K.; Cin, Z.; Zhao, B.; Yang, B. J. Colloid Interface Sci.
2002, 255, 115-118.
10-18 mol (15 fL × 100 µM) of CV molecules was being probed (29) Mabuchi, M.; Takenaka, T.; Fujiyoshi, Y.; Uyeda, N. Surf. Sci. 1982, 119
in the unenhanced Raman measurements. (2-3), 150-158.
(30) Blatchford, C. G.; Campbell, J. R.; Creighton, J. A. Surf. Sci. 1982, 120
(2), 435-455.
RESULTS AND DISCUSSION (31) Ahem, A. M.; Gavell, R. L. Langmuir 1995, 7 (2), 254-261.
Macroscopic and Microscopic Characterization of Au (32) Robert, J. G. Contact Angle, Wettability, and Adhesion; Fawkes, F. M., Ed.;
Colloidal Films. Metal colloidal substrates containing layers of Advances in Chemistry Series 43; American Chemical Society: Washington,
DC, 1964; p 112.
(25) Qin, L.; Shen, Z. X.; Tang, S. H.; Kuok, M. H. Asian J. Spectrosc. 1997, 1, (33) Deegan, R. D.; Bakajin, O.; Dupont, T. F.; Huber, G.; Nagel, S. R.; Witten,
121. T. A. Nature 1997, 389, 827-829.
(26) Max, B.; Emil, W. Principles of Optics; Pergamon Press: London, 1964. (34) Deegan, R. D. Ph.D. Thesis in Physics, University of Chicago, 1998; p 40.

7464 Analytical Chemistry, Vol. 77, No. 22, November 15, 2005
Figure 1. Macroscopic and microscopic images of colloidal films fabricated by drying 150 µL of 15 nm Au suspension droplet on a SuperFrostPlus
microscope slide. (a) Macroscopic image showing three colloidal films. (b) Magnified view showing a portion of one of the films. (c) Histogram
of the enclosed area shown in (b). (d) Magnified image of an unevenly coated area showing archlike structures. (e) Histogram of the closed
area shown in (d). Note that both images shown in (b) and (d) have been processed to correct for intensity variations arising from uneven
illumination over the imaged area.

uniform colloidal region means ease of identification and localiza- Figure 1e shows a histogram derived for an anomalously
tion of areas for carrying out SERS measurements. We will hereby coated region (the enclosed region in Figure 1d). The mean
refer to such a uniform colloidal region as the microscopically intensity is 209.49 with a STD of 7.95, which is about 3 times as
uniform colloidal area. large as that of the uniform Au colloidal film.
Also of interest are the inevitable anomalously coated regions Surface Morphologies of Au Colloidal Films. As mentioned
present in some of the films fabricated. A representative micro- above, we are interested in what the effect of the initial particle
scopic image of such regions is shown in Figure 1d. One can concentration of the drying droplets would have on the surface
notice archlike structures (arrowed). Although surfactant-en- morphology of the microscopically uniform colloidal areas. In
hanced Bernald convection can also give rise to a similar struc- particular, we would like to know the optimal particle concentra-
ture,35 we believe a different mechanism has taken place here since tion for generating a colloidal array consisting of closely spaced
the arches are not connected as in the case observed in ref 35. but physically separated particles. To this end, droplets containing
One plausible explanation is that as the contact line retreats, different particle concentrations were dried as delineated above.
different segments along the line move at different speeds, re- Generally, we did not find any significant difference in the general
sulting in convex and concave regions: segments that move feature of the resulting films (a ring deposit along the perimeter
slower become convex, while those that move faster become of the films, a uniform-coated region extending from the ring, and
concave. Consequently, the convex region would deposit faster some regions containing archlike structures) except that those
than the concave region,33 thereby forming a semiring or arch.
derived with lower particle concentrations appear more translucent
Exactly how and when the contact line starts to move in this
and pinkish. The optical properties of the microscopically uniform
fashion is not known at this stage, but it is suspected to have
colloidal areas are studied using UV-vis spectroscopy. Typical
begun at the very initial stage of the evaporation process along
UV-vis spectra are shown in Figure 2. Note that we have taken
regions where the contact line is irregular. This is quite telling
care to ensure that only the uniform areas are illuminated and
since irregular ring deposits are often found near regions contain-
have also confirmed that the ring deposit around each film is not
ing archlike structures (compare the ring deposits in Figure 1,
contributing any detectable effect on the final spectrum. Thus,
parts b and d). The exact mechanism behind the formation of
the spectra shown for the films directly reflect the surface
such an interesting pattern is currently being investigated.
properties of the uniform colloidal layers. The UV-vis spectrum
(35) Van, X. N.; Kathleen, J. S. Phys. Rev. Lett. 2002, 88 (6), 164501-164504. for a solution containing isolated 15 nm diameter colloidal Au
Analytical Chemistry, Vol. 77, No. 22, November 15, 2005 7465
 peak is situated at 520 nm, which is in agreement with the previous
results obtained for a 15 nm spherical Au colloid.6 However, when
the interparticle separation is comparable to the particle diameter
(,λ), interparticle coupling becomes pronounced, and this would
generate a new red-shifted feature in the optical spectrum centered
between 600 and 800 nm. Theoretical and experimental studies
have shown that the intensity and peak of this feature scale with
the decrease in particle separation,15,37 and for particles that are
physically connected (i.e., aggregated), this feature would become
broadened and shifted to a longer wavelength beyond 700 nm
(see spectrum Agr). Red-shifted bands are also seen here in the
spectra (curve A, B, and C) of the films deposited using different
initial particle concentrations. The general trend observed is that
the band shifts to a longer wavelength as the initial particle
concentration increases, indicating ascending packing density. In
spectra C, which is that of the film deposited by drying colloid
droplets containing 1.9 × 1012 particles per mL, the red-shifted
band, which is centered at around 720 nm, is broad, suggesting
the formation of aggregates (compare this with curve Agr). On
the contrary, spectra A and B, which are derived from films
fabricated with initial particle concentrations of 3.5 × 1011 and 1.4
× 1012 particles per mL respectively, exhibit two resonance peaks.
The first resonance band is independent of the particle concentra-
tion and is situated at 520 nm. This band is attributable to the
plasmon resonance in the individual isolated particles. The second
band arises from the particle-particle resonance and is located
at 617 nm for film A and 633 nm for film B. As reported
previously,6,15,27 the presence of such a double-peak feature is an
indication of closely spaced morphology. To verify these results,
AFM analysis was performed on the above colloidal films. The
AFM data are presented in Figure 3. In film A, one can observe
closely spaced but amorphously organized particles with an
average particle-to-particle separation of about 30 nm, which is in
good agreement with the closely packed morphology predicted
above from the UV-vis spectrum. As can be seen from the height
Figure 2. Normalized UV-vis spectra. Sol: 15 nm Au hydrosol; profile, the film thickness is a submonolayer with height variations
Agr: solution containing Au aggregates induced by mixing NaCl with
Au hydrosol. A, B, and C: Normalized UV-vis spectra of dried
comparable to the size of the individual particles. Increasing the
colloidal films prepared from droplets containing 3.5 × 1011, 1.4 × particle concentration of the evaporating droplets increases the
1012, and 1.9 × 1012 particles per mL, respectively. Curves are shifted packing density of the resultant colloidal films. This is shown in
vertically for clarity. Figure 3, parts b and c, which correspond to colloidal layers
obtained with 1.4 × 1012 and 1.9 × 1012 particles per mL,
particles (spectrum denoted as sol) and that for a solution of Au respectively. One can readily see that, at the highest particle
aggregates induced by addition of 10% NaCl to the Au hydrosol concentration, aggregates of ill-defined dimensions are formed.
(spectrum denoted as Agr) are also shown for comparison. As This correlates well with the UV-vis data described above (see
can be seen, there are distinct differences between the spectrum curve C in Figure 2). The surface morphology of the film deposited
derived from the isolated Au particles and those from the other at an intermediate particle concentration (i.e., film B), on the other
samples. This is due mainly to the different levels of coupling hand, is qualitatively different and is interesting in that it consists
between particles. In the isolated sol, which has a particle of interconnected islands. These islands appear to have a “flat top”
concentration of 1.4 × 1012 particles per mL, the average particle in general as indicated by the height profile (lower panel in Figure
separation is estimated to be about 900 nm (.15 nm). With such 3b). We thus ask whether these islands are actually domains of
a large separation, interparticle coupling is extremely weak, and closely packed particles that are too small to be resolved by the
the UV-vis spectrum is predominantly derived from the surface AFM tip. To answer this question, we digitally improve the lateral
plasmon resonance in the individual particles. The physical nature resolution of the image by making use of a band-pass filter as
of this resonance mode, which gives the colloidal Au its charac- described in ref 16. The processed image of the area enclosed by
teristic intense red color, is well understood, as are its dependence the white-dotted box shown in Figure 3b is presented in Figure
on particle size and shape.29,36 In the current case, this resonance 4. Individual particles are now more discernible. One can observe

(36) Bohren, C. F.; Huffman, D. R. Absorption and Scattering of Light by Small (37) Zhao, L.; Kelly, K. L.; Schatz, G. C. J. Phys. Chem. B 2003, 107 (30), 7343-
Particles; John Wiley and Sons: New York, 1983. 7350.

7466 Analytical Chemistry, Vol. 77, No. 22, November 15, 2005
 Figure 4. Digitally processed AFM image showing closely packed
morphology in the film prepared at an initial particle concentration of
1.4 × 1012 particles per mL.

comparable to the individual particle size. If a particle radius of

7.5 nm (15 nm/2) is assumed, this would be translated into an
interparticle spacing of 0-2 nm. Note that this equilibrium
distance is far too small to be explained by diffuse-layer repulsion.24
Thus, the short particle spacing is likely due to the steric
interaction of a stabilizing molecular layer adsorbed on the particle
surface, in this case, the chloride ions. We have to note that the
observation of such a closely packed Au colloidal array, formed
by a drying suspension droplet, is rather suprising, because this
clearly disagrees with the majority of the previous claims that Au
hydrosols are not stable enough to form 2D closely packed
colloidal arrays without forming large aggregates,6,20-22 owing to
their large Hamaker constant. Although particles can be made
more stable with thiol- and amide-based protection layers to allow
the formation of large, ordered 2D colloid lattices, and even 3D
superlattices,38,39-41 capping Au particles with a protective layer
may have compromising effects on their SERS performance: first,
the protective layer can increase the separation between the
absorbate and the particle surface, thereby reducing the SERS
effect3; second, the protective layer may introduce confounding
signals to the final SERS spectrum.3 However, as far as reusability
is concerned, a suitably thin protective layer may still be desirable
to prevent covalent bonding of the absorbate on the metal surface.
This will be investigated in the future as an improvement to our
Au colloidal films.
Film Stability. One of the principal objectives of the current
study is to prepare SERS-active substrates that are suitable for
biofluid analysis. Since biofluids generally contain a certain amount
of salts, it is thus important to study the stability of the Au colloidal
layers under conditions of high ionic strengths. To achieve this,
we immersed the Au colloidal layers in an aqueous solution
containing 10% NaCl, which is about 10 times more salty than
Figure 3. AFM mappings of colloidal films prepared from droplets
containing (a) 3.5 × 1011, (b) 1.4 × 1012, and (c) 1.9 × 1012 particles (38) Harfenist, S. A.; Wang, L. Z.; Whetten, R. L.; Vezmar, I.; Alvarez, M. M.
per mL. Scanned area ) 1 µm × 1 µm. Adv. Mater. 1997, 9 (10), 817-822.
(39) Stoeva, S.; Klabunde, K. J.; Sorensen, C. M.; Dragieva, I. J. Am. Chem. Soc.
hexagonal or square particle arrangements (enclosed in dotted 2002, 124 (10), 2305-231.
(40) He, S.; Yao, J.; Jiang, P.; Shi, D.; Zhang, H.; Xia, S.; Pang, S.; Gao, H.
boxes). Further analysis of this particular image shows a center- Langmuir 2001, 17 (5), 1571-1575.
to-center separation ranging between 15 and 17 nm, which is (41) Brown, L. O.; Hutchison, J. E. J. Phys. Chem. B 2001, 105, 8911-8916.

Analytical Chemistry, Vol. 77, No. 22, November 15, 2005 7467
Figure 5. UV-vis spectra of a Au film before and after immersion in a 10% NaCl solution.

the normal physiological levels found in biofluids.42 No resuspen- reasons. First, films deposited at this concentration (i.e. film B)
sion of particles were observed after 2 weeks. Typical UV-vis have a closely packed surface morphology (see Figure 3b and
absorption spectra measured on these colloidal layers before and Figure 4) identical to that of SERS-active silane-immobilized
after the NaCl treatment are shown in Figure 5. Both spectra are particle arrays.6,27 Second, UV-vis measurement of the films
generally identical, indicating that there is no appreciable mor- indicates an interparticle coupling resonance peak at 633 nm,
phological change in these layers after the treatment. This is which matches that of our current HeNe excitation laser wave-
clearly in contrast to Au hydrosols, which tend to coagulate at length, an important criteria for inducing large SERS activity.44,45
high molarities of salts (see curve Agr in Figure 2) due to the Third, film B has been shown experimentally to be highly Raman
screening of the repulsive interactions between particles by the enhancing compared to the other two (i.e., films A and C), as will
co-ions. The excellent stability of our films toward salty solutions be discussed below.
therefore leads us to believe that the Au particles are bonded To quantify film reproducibility, we calculated the arithmetic
firmly via electrostatic interactions to multiple sites on the charged mean roughness value46 (Rs ) (1/N)∑ |z1 - zav|) for a 1 µm × 1
SuperFrostPlus surface. In fact, in order for all of our SERS µm microscopically uniform area in each of the 18 Au colloidal
experiments described below to be of biological relevance, all films. An Rs value of 5 nm with a 14% variation was obtained,
sample solutions used in the experiments were prepared with a indicating good reproducible surface roughness. The roughness
physiological salt concentration of about 0.2 M. of the SuperFrostPlus glass surface is measured to be about 0.9
Finally, we have to stress that our films are assembled solely nm and thus does not contribute significantly to the calculated Rs
by the convection-induced particle flows in the wetting film near values.
the contact line of the drying droplet and not by spontaneous Due to its high Raman enhancement, film B will be used in
electrostatic adsorption of particles on the glass surface as in the most of our SERS experiments and will hereby be referred to as
case of electrostatic self-assembly,43 since immersing the Super- the Au film unless otherwise explicitly stated.
FrostPlus glass slides in a solution of Au colloid for 2 days did Preresonance SERS Spectra of Crystal Violet on Au Films.
not result in particle attachment. Note also that in no case did we The SERS performance of the Au films was studied by measuring
obtain any stable colloidal film if the suspension drops are dried the Raman signals from CV molecules adsorbed on them.
on normal glass slides, which lack the electrostatic layers. Films Depending on the solvent’s pH, CV generally exists in three
formed on these slides were either partially or completely removed different forms.47 In the present study, the CV was dissolved in a
upon rinsing with distilled water. PBS buffer with a pH value of 7.2 and should therefore exist
Film Reproducibility. A useful deposition method must be predominantly in the carbonium ion (native) form (chemical
reproducible. To this end, we studied the surface roughness of structure of which is shown in the inset of Figure 6a). This was
18 different Au colloidal films fabricated by our deposition method confirmed by the occurrence of a prominent singlet π-π*
using the initial particle concentration of 1.4 × 1019 particles per
mL. This particular concentration was chosen for two main (44) Fornasiero, D.; Grieser, F. J. Chem. Phys. 1987, 5, 3213-3217.
(45) Crieghton, J. A. J. Chem. Soc., Faraday Trans. 1979, 2 (75), 790-798.
(42) Andrei, B.; Elvira, E. M.; Robert, O. D.; Thomas, C. B. Biochemistry 1997, (46) Nowicki, B. Wear 1985, 102, 161-176.
35 (25), 7821-7831. (47) Jessop, J. L. P.; Goldie, S. N.; Scranton, A. B.; Blancherd, G. J. J. Vac. Sci.
(43) Sastry, M. Pure Appl. Chem. 2002, 74 (9), 1621-1630. Technol., B 2002, 20 (1), 219-225.

7468 Analytical Chemistry, Vol. 77, No. 22, November 15, 2005
Figure 6. (a) UV-vis spectra of a 1 µM CV solution. Inset: chemical structure of a native CV. (b) Preresonance SERS spectrum of a 1 µM
CV solution in pH 7.2 PBS buffer (upper). Unenhanced preresonance Raman spectrum of a 100 µM CV solution in pH 7.2 PBS buffer (lower).
Excitation Source: 633 nm; 2 mW; 20 s integration.

electronic absorption peak at 593 nm in the UV-vis spectrum of

the CV solutions (see Figure 6a).47 Thus, at the excitation
wavelength of 633 nm, the CV is preresonantly excited.
Typical unenhanced (bottom curve) and surface-enhanced
(upper curve) preresonance Raman spectra of CV are shown in
Figure 6b. The unenhanced Raman spectrum was derived from
the liquid surface of a droplet containing 100 µM CV in PBS, as
delineated in the Experimental Section. The preresonance SERS
spectrum on the other hand was derived from the surface of a
Au film immersed in a 1 µM CV solution. The integration time
used for all experiments is 20 s. Note that the use of a buffer
solvent is essential. This is to eliminate any pH-dependent variation
in the CV’s SERS spectrum.48 The preresonance SERS spectrum
obtained here is in good agreement with that previously reported
for this particular compound.48
Figure 7. The maximum SERS intensity at 1619 cm-1 (left axis,
(48) Polwart, E.; Keir, R. L.; Davidson, C. M.; Smith, W. E.; Saddler, D. A. Appl. solid line) and the wavelength of the interparticle plasmon resonance
Spectrosc. 2000, 54 (4), 522-527. (right axis, dashed line) vs the initial particle concentrations.

Analytical Chemistry, Vol. 77, No. 22, November 15, 2005 7469
Figure 8. (a) Preresonance SERS spectrum of a 1 µM CV solution in the region of 150-500 cm-1. (b) Preresonance SERS spectrum obtained
from a Au film soaked in 1 mM CV solution for 21 h. Excitation Source: 633 nm; 2 mW; 20 s integration.

On the basis of a triplicate measurement of the 1 µM CV’s resonance is shorter than the excitation wavelength, the SERS
SERS intensities at 1619 cm-1, the SERS reproducibility of our intensity increases with the initial particle concentrations. When
Au films was estimated to be around 14%, which is comparable to the resonance maximum shifts to a wavelength longer than that
that of a silane-immobilized Au colloidal layer. of the excitation, the SERS intensity falls off. Olson et al. have
We also investigated the enhancement of Raman scattering also found this trend in their research on a silane-immobilized
for the Au colloidal arrays fabricated with different initial particle Au colloidal film as a SERS-active substrate.27
concentrations exposed to a 1 µM CV solution. Generally, the Raman Enhancement Factor. The apparent enhancement
excitation wavelength for the maximum SERS enhancement would factor (AEF) for the Au films can be calculated by ratioing the
coincide with the adsorption maximum of the interparticle plasmon intensity of the peak at 1619 cm-1 in the preresonance SERS
resonance.44,45 Such a correlation is studied here for our Au spectrum to that of the corresponding peak in the unenhanced
colloidal arrays. Figure 7 illustrates the maximum SERS intensity Raman spectrum, taking into account that the SERS signal is
at 1619 cm-1 (left axis) and the wavelength of the interparticle derived from a CV solution 100 times more dilute than that used
plasmon resonance maximum (right axis) versus initial particle for the unenhanced Raman measurement. According to the curves
concentrations. It can be seen from Figure 7 that the SERS displayed in Figure 6b, this yields an AEF of approximately 2000.
intensity increases with the increase in the initial particle con- Unfortunately, it is not as easy to estimate the chem-
centrations and then levels down gradually beyond 1.4 × 1019 ical/electromagnetic enhancement factor of the Au films, since
particles per mL as the interparticle plasmon adsorption maximum this will generally require a detailed knowledge of the surface
shifts to a longer wavelength than the excitation wavelength (633 morphology, the binding characteristic (i.e., covalent or nonco-
nm). That is to say, when the wavelength of the plasmon valent), and the binding geometry of the absorbate on the metallic
7470 Analytical Chemistry, Vol. 77, No. 22, November 15, 2005
 plateau has been reached in the adsorption of CV on the Au film
at 1mM. To calculate the minimum electromagnetic enhancement
factor (EFF), we first made the following assumptions: (1) the
Au particles in the array take on a hexagonal packing as shown
in Figure 9, which is the maximum packing density possible for
a spherical particle; (2) only the molecules that are adsorbed in
the crevice (“hot spot”) between Au particles needed to be
considered since they are the main contributors to the measured
SERS signals, as has been reported in the previous electromag-
netic studies of SERS;3,29-31 (3) all available hot spots are occupied
by CV molecules when the film is saturated; (4) owing to the large
molecular size (120 Å2) of the CV and steric interactions,51 each
hot spot is occupied by not more than one CV molecule (see
Figure 9). With these in mind, we then estimated that there are
Figure 9. Hexagonally packed Au colloidal array saturated with CV
molecules. about 3 × 104 SERS-contributing CV molecules (∼5.017 × 10-20
mol) residing within the 1.78 µm2 laser spot. In a similar manner
as in the calculation of the AEF, the minimum EEF value is thus
surface. As such we can only calculate for the minimum enhance-
calculated to be about 2 × 104. We stress that this is only the
ment factor of the Au films. This is achieved in the following
lowest estimation, since the surface morphology of the current
manner:
Au films is not ideally packed and that it is very unlikely that all
First, we establish that CV molecules are normally bonded to
of the available hot spots between particles are occupied by CV
the Au surface by Coulombic and van der Waals interactions.49
molecules even at very high concentration as has been assumed
This is evident by the absence of a Au-N stretching vibrational
in the calculation; the actual EEF should thus be significantly
peak in the 225-231 cm-1 region of the SERS spectrum derived
higher.
from Au films exposed to a 1 µM CV solution (see Figure 8a).
Clearly, this is in contrast to the SERS spectra of DNA nucleosides
CONCLUSION
chemically adsorbed onto Au nanoparticles, where Au-N peaks We have demonstrated a simple convective assembly of 15 nm
were observed indicating that the nucleosides were covalently Au nanoparticles on electrostatically charged glass substrate.
bonded via the lone electron pair in the nitrogen atoms.50 We Uniform colloidal layers can be easily found in the dried films
therefore surmise that there is no chemical enhancement in the and were shown to contain islands of hexagonally spaced particles.
SERS of the CV-Au complex; i.e., the Raman enhancement The average interparticle separation is estimated to be about 17
observed in the present study is purely electromagnetic. Next, nm based on AFM analysis. With the use of CV as a test sample,
we soak a Au film in a 1mM CV solution for 21 h in an attempt to the SERS reproducibility of a properly deposited films was shown
saturate the film’s surface. After rigorous rinsing with PBS solution to be about 14%, which is comparable to that previously reported
for 2 min to wash off unbound CV molecules, SERS measurement for silane-immobilized Au colloids. The minimum EEF is estimated
was carried out immediately with the film remained immersed in to be about 2 × 104.
the PBS solution. The resultant preresonance SERS spectrum in
the region between 1500 and 1750 cm-1 is shown in Figure 8b. ACKNOWLEDGMENT
The intensity of the Raman peak at 1619 cm-1 is about 70 000 We would like to thank Dr. Sow Chorng Haur and Mr. Ong
counts, and further increases in the CV concentration beyond 1 Peng Ming from the Department of Physics, National University
mM did not result in more photon counts, suggesting that a of Singapore for their technical assistance with the AFM mapping.

(49) Liang, J. D.; Burstein, E.; Kobayashi, H. Phys. Rev. Lett. 1986, 57 (14),
1793-1796. Received for review March 14, 2005. Accepted July 27,
(50) Nak, H. J. Bull. Korean Chem. Soc. 2002, 23 (12), 1790-1800. 2005.
(51) Shaw, A. M.; Hannon, T. E.; Li, F.; Zare, R. N. J. Phys. Chem. B 2003,
107, 7070-7075. AC050437V

Analytical Chemistry, Vol. 77, No. 22, November 15, 2005 7471