Hunting for the Active Sites of Surface-Enhanced Raman Scattering: A New Strategy Based on Single Silver Particles

Article abstract of DOI:10.1021/jp962135q

Three major findings are reported here.First, a new technique based on lithographic modification with a scanning tunneling microscope (STM) has been developed for fabricating single Ag particles on atomically flat Au(111) surfaces.Second, surface-enhanced Raman scattering (SERS) from one isolated Ag particle has been observed and quantified for the first time.The enhancement factor due to one Ag particle of about 1 μm diameter is on the order of 1E4.This factor includes contributions from all forms of chemical enhancement, which are smaller than 1E2.Finally, strong evidence suggests the importance of clusters containing closely-spaced particles.Extra enhancement due to clusters of randomly distributed Ag particles is around 1E2.We believe that the extra enhancement is spatially localized near the gap sites between two particles in proximity.These gap sites from an important class of SERS-active sites.

Full text of DOI:10.1021/jp962135q

632
J. Phys. Chem. B 1997, 101, 632-638
Hunting for the Active Sites of Surface-Enhanced Raman Scattering: A New Strategy
Based on Single Silver Particles
†
Tong Xiao, Qi Ye, and Li Sun*
Department of Chemistry, UniVersity of Minnesota, Minneapolis, Minnesota 55455
X
ReceiVed: July 16, 1996; In Final Form: October 9, 1996
Three major findings are reported here. First, a new technique based on lithographic modification with a
scanning tunneling microscope (STM) has been developed for fabricating single Ag particles on atomically
flat Au(111) surfaces. Second, surface-enhanced Raman scattering (SERS) from one isolated Ag particle
has been observed and quantified for the first time. The enhancement factor due to one Ag particle of about
4
1
µm diameter is on the order of 10 . This factor includes contributions from all forms of chemical
2
enhancement, which are smaller than 10 . Finally, strong evidence suggests the importance of clusters
containing closely-spaced particles. Extra enhancement due to clusters of randomly distributed Ag particles
is around 10 . We believe that the extra enhancement is spatially localized near the gap sites between two
2
particles in proximity. These gap sites form an important class of SERS-active sites.
Introduction
ideal.25 More importantly, in the above two examples, the
detected SERS signal results from a large number of parti-
We report here a new strategy for revealing the correlation
between the surface morphology of a SERS substrate and the
corresponding enhancement factor. The key of this strategy
involves fabrication and characterization of single Ag nanopar-
ticles on an atomically flat surface through a combined use of
three techniques: scanning probe microscopy, electrochemistry,
and surface Raman spectroscopy. The new strategy has already
allowed us to observe SERS from one single Ag particle and
to gain a useful insight into the role of interparticle interactions.
We believe that this work points to a promising new direction
for quantitatively evaluating the importance of various compet-
ing SERS mechanisms.
cles; thus, interparticle interactions cannot be easily elimi-
nated or accurately quantified, especially when the spatial
distribution of a particle ensemble is random. Interparticle
interactions are more defined in arrays of regularly-spaced
particles fabricated either by photolithography or by template
27-30
methods.
However, the particle density reported so far may
be still too high to completely avoid the effects of interparticle
interactions and to isolate the intrinsic enhancement due to
individual particles.
Obviously, for mechanistic studies it is desirable to observe
SERS from one single particle. Toward this goal, Van Duyne
and Haller have developed a SERS microprobe.3
1,32
How-
Since the discovery of SERS about 20 years ago,1,2 many
theories have been proposed to explain this phenomenon which
involves nearly a million-fold enhancement of the Raman
ever, the detection efficiency of their PMT/scanning spectrom-
eter setup was inadequate. In addition, at the time of their
investigation, techniques for fabricating and characterizing single
particles were neither well-developed nor widely available.
We have overcome all of these technical difficulties in this
study, thus observed SERS from one single particle for the first
scattering cross section for molecules adsorbed on the surface
of metal nanoparticles.3
-6
Two major types of enhancement
are now recognized: the long-range electromagnetic (EM) field
7
,8
9,10
enhancement and the short-range chemical enhancement.
33
time.
Since the chemical enhancement does not require the presence
of nanoparticles, it can be studied independently using atomi-
Experimental Section
1
1,12
cally flat surfaces.
However, verifying the predictions of
an EM theory has been difficult because it is difficult to fabricate
Chemicals and Materials. Water of 18.2 MΩ cm resistivity
(Milli-Q Plus, Millipore) was used in all electrochemical
measurements. n-Hexadecanethiol (distilled twice from a 92%
product before use), phosphorus tribromide (99%), thiourea
(99+%), and trans-4-stilbenemethanol were obtained from
Aldrich Chemical Co. Silica gel (J. T. Baker Inc.) and organic
solvents of reagent grade or better were used as received.
Synthesis of trans-4-Mercaptomethyl Stilbene. trans-4-
Mercaptomethyl stilbene (t-4MMS) was synthesized via a
1
3-20
stable and well-defined SERS-active surface morphology.
To solve this problem, several new methods for producing
21-26
roughened surfaces have been developed.
For example,
silver-coated latex spheres as well as immobilized monodis-
persed colloidal particles are morphologically more stable than
an electrochemically roughened surface.1
4,15,24
Although the stability is improved, the above substrates are
still not ideal for probing the SERS mechanisms. For example,
a silver overlayer on latex spheres has a complex geometry
which may not be amenable to theoretical modeling. The size
of monodispersed colloidal particles is limited to be less than
34
thiouronium intermediate PhCHdCHPhCH SH. Transforma-
2
tion via the thiolacetate intermediate was also attempted, but
3
5
the yield was lower (10%).
1
00 nm, and the control of interparticle spacing is far from
trans-4-Bromomethyl Stilbene (1). To a stirred solution of
trans-4-stilbenemethanol (0.43 g, 2 mmol) in CH2Cl2 (50 mL)
was added phosphorus tribromide (0.32 g, 1.18 mmol) in
CH2Cl2 (20 mL). After reaction at 0 °C under N2 for 10 h, the
mixture was quenched by 100 mL of water. The organic layer
was washed sequentially with 5% NaHCO3 (50 mL) and brine
†
Present address: Department of Chemistry, University of Rochester,
Rochester, NY 14627.
Author to whom correspondence should be addressed. E-mail:
sun@chemsun.chem.umn.edu.
*
X
Abstract published in AdVance ACS Abstracts, January 1, 1997.
S1089-5647(96)02135-9 CCC: $14.00 © 1997 American Chemical Society

Hunting for the Active Sites of SERS
J. Phys. Chem. B, Vol. 101, No. 4, 1997 633
Figure 1. Schematic diagram of the Raman microscopic system.
(100 mL). Drying over CaCl2 and removal of the solvent gave
0
.44 g of crude product. Elution on a silica gel column with
n-hexane/chloroform (6:1) yielded 0.43 g (80%) of 1 as a white
1
powder: H NMR (CDCl3) δ 7.4-7.5 (m, 9H), 7.1 (d, 2H),
4
.5 (s, 2H).
trans-4-Mercaptomethyl Stilbene (2). A mixture of 1 from
the above step, thiourea (0.12 g, 11.6 mmol) in degassed ethanol
75 mL) was stirred under N2 for 9 h at room temperature. After
Figure 2. Fabrication of single Ag particles for SERS measurements.
A self-assembled monolayer (SAM) of hexadecanethiol (open circles
with linear tails) on Au(111) was prepared by soaking an Au sample
in a 1 mM ethanolic thiol solution for 48 h. The SAM resist was
patterned via STM lithography and then developed further by electro-
(
solvent removal, the remaining white solid (a thiouronium salt)
was refluxed in a degassed solution of NaOH (0.13 g, 3.12
mmol; 50 mL water) for 5 h. The reaction mixture was cooled
to 0 °C, acidified to pH 2 with 6 N HCl, extracted with
chloroform (2 × 40 mL), and dried over MgSO4. Flash
chemical anodization in a solution of 0.1 M Na
at 300 mV (vs an Ag/AgCl/3 M NaCl reference electrode) for 10 s.
After transfer to another electrolyte containing 1 mM AgClO and 0.1
M HClO , a single Ag particle was deposited onto the lithographic
window at 470 mV (vs an Ag/AgCl/saturated-KCl/0.1 M HClO
2 4
HPO and 0.1 M NaCN
4
chromatography (5:1 n-hexane/chloroform) afforded 0.17 g
4
1
(
(
47%) of 2 as a white solid: H NMR [(CD3)2SO] δ 7.6-7.3
4
m, 9H), 7.2 (d, 2H), 3.7 (d, 2H), 2.9 (t, 1H); Raman (647.1
reference electrode) for 3 min. The Ag particle was covered with a
SAM of Raman probe (solid filled circles) after soaking in an ethanolic
solution of t-4MMS for 20 min.
-
1
nm excitation) 2561 cm (S-H stretching vibration).
Surface Raman Spectroscopy. The electrochemical and
Raman spectroscopic measurement systems have been described
previously.36 The Raman system was modified slightly to adapt
sample holder, similar to a previous design,42 was used to aid
sample transfer between three major instruments, namely, the
potentiostat, the Raman spectrograph, and the scanning probe
microscope.
to the present work (Figure 1). A 10×, 0.21 NA microscope
objective (Nikon) was installed to focus the laser to a diffraction-
limited spot at a 0° incident angle. The same objective was
also used to collect Raman photons via a cube beam splitter
Results and Discussion
(Edmond Scientific, Barrington, NJ). Routine data acquisition
parameters were 400 µm slit width, 2 mW laser power, and 50
s integration time. Samples were moved relative to the laser
spot with an xyz translational stage (Newport, Irvine, CA) of
A. Fabrication and Analysis of Single Ag Particles
Cyanide Etching and Silver Deposition by Electrochemical
Methods. Initially, we hoped to use STM lithography to create
a microeletrode and grow one Ag particle from it. However,
this approach was not successful because there was not enough
contrast between the Ag deposition rate at the tip-etched area
and the rates elsewhere. We noticed that the threshold potential
for Ag deposition onto etched areas varied randomly for different
samples from 0 to 400 mV, a range more negative than the
theoretical threshold of 425 mV for bulk deposition from a 1
0
.5 µm resolution.
Scanning Probe Microscopy. All STM and scanning force
microscopic (SFM) images were acquired with a Nanoscope
III microscope (Digital Instruments, Santa Barbara, CA).
Vibration isolation was accomplished with the “bungee”-cord
3
7
design.
The STM tips were mechanically cut from a 0.010-in.
diameter Pt/Ir (80/20 w/w) wire while the SFM tips were
pyramidal silicon nitride cantilevers with a spring constant of
+
mM Ag solution. This suggests that the thiols in tip-etched
0
.58 N/m. Typical STM imaging conditions were 300 mV bias
area have not been completely removed, which would lead to
large, and perhaps irreproducible, threshold overpotential. One
consequence of large overpotential is that Ag is also deposited
at the adventitious defect sites. We have attempted to remove
the residue thiols in the tip-etched area by oxidizing it at 1100
mV prior to Ag deposition. However, this anodic excursion
enhances Ag nucleation at the adventitious defects as well,
resulting in no or even decreased contrast in Ag deposition rate
(Figure 3). The Ag particle density is so high that it is
impossible to find features due to STM lithography.
voltage, 0.11 nA constant tunneling current, and 2 Hz scan rate.
Constant force mode and a 2-Hz scan rate were used in SFM
imaging, but the exact force was unknown because the force-
vs-distance plot was not recorded.
Typical conditions for STM lithography were 3-5 V bias
voltage, 0.11 nA tunneling current, 30-40 Hz scan rate, and 3
min etching time.
4
3
38,39
All the STM lithographic operations were
performed in air without humidity control although we found
later that humidity is a critical factor.40
Sample Preparation and Measurements. Au(111) facets
We then tried to remove the thiol residue through electro-
chemically controlled cyanide etching. Unlike Ag deposition,
cyanide etching exhibits a large contrast in etching rate: no
pits are observed in the area not etched by the STM tip.
Consequently, cyanide etching dramatically improves the con-
on an Au sphere were prepared according to literature
procedures.3
8-42
The remaining fabrication steps are illustrated
in Figure 2. Ag particles with adsorbed probe molecules were
first characterized by SERS, and then by SFM. A home-made

634 J. Phys. Chem. B, Vol. 101, No. 4, 1997
Xiao et al.
Figure 4. SFM image of a string of mesa-like Ag deposits at four
tip-etched and cyanide-cleaned 1 µm × 1 µm windows. Ag was
deposited at 500 mV for 20 s. The z-scale is 10 nm.
Figure 5. SFM image of a string of Ag particles deposited at seven
tip-etched and cyanide-cleaned 500 nm × 500 nm windows. Ag was
deposited at 470 mV for 3 min. The z-scale is 1 µm. The Ag#5 particle
was knocked off the surface by the imaging SFM tip.
particles can be seen in the area not etched by the STM tip.
Longer deposition time simply results in a larger particle size
4
4
(Figure 5).
Size and Shape Analysis of Ag particles. As seen in Figure
6
, a close-up view of an individual particle does not reveal its
45,46
true topography because the SFM tip has a finite cone angle.
According to a simple geometrical model, R′ obtained from a
line cross section should be equal to R, the half cone angle of
the pyramidal SFM tip (Figure 7). The calculated half cone
angle is 35°, assuming that each face of the pyramid is molded
from a Si(111) plane. This agrees approximately with the
observed R′ of 32° ( 3° obtained from line cross sections, such
as the one shown in Figure 6B.
The relatively dull SFM tip makes it difficult to determine
the shape of each particle. Nevertheless, an Ag particle may
be approximated as a spherical protrusion with its center located
at a distance q away from the Au surface (Figure 7). The
Figure 3. (A) SFM image of clusters of Ag particles deposited onto
adventitious defect sites within a hexadecanethiol SAM on Au(111).
The deposition potential was 200 mV, and the deposition time was 10
s. (B) Smaller scan area. (C) Line profile shown in B. The z scale in
both A and B is 500 nm.
4
7
trast in the Ag deposition rate. There exists a window of
threshold potential within which Ag deposition occurs at the
tip-etched, cyanide-cleaned area but not at the adventitious
defect sites. Figure 4 shows an SFM image of Ag particles
produced with this method. At a full z-scale of 10 nm, no Ag

Hunting for the Active Sites of SERS
J. Phys. Chem. B, Vol. 101, No. 4, 1997 635
The radius obtained with eq 1 is subject to a large uncertainty,
however. On the basis of sequential images of the same particle,
we observed a 5% error in h and d, and a 10% error in R′.
Error propagation results in even larger error in radii. For
example, the Ag#3 particle in Figure 5 shows r ) 370 ( 110
nm and q ) 420 ( 120 nm; the Ag#4 particle shows r ) 590
(
120 nm and q ) 450 ( 130 nm. Thus, these particles are
approximately spherical (r ≈ q) although the geometry below
the top hemisphere cannot be obtained from an SFM image.
Similar analysis of the Ag particles in Figure 3B reveals a
hemispherical shape with typical r ) 310 ( 66 nm and q )
-
30 ( 70 nm. The hemispherical shape is reasonable since,
in this case, Ag was deposited under a large overpotential that
4
8
favors diffusion-controlled growth.
Discussions on the Single Particle Fabrication Technique.
The single particle fabrication technique outlined here has three
distinct advantages for SERS mechanistic studies. First, this
technique allows unambiguous isolation of a single particle for
SERS measurements. The region surrounding each particle is
an atomically flat Au(111) surface which is known to support
little or no EM enhancement. Furthermore, Raman probe
molecules are excluded from the flat Au(111) region which has
been protected with the SAM of hexadecanethiol. In short, any
Raman signal from the probe molecules originates from the
surface of one isolated Ag particle. This kind of isolation is
important since the laser spot size is usually larger than the
particle size, which means that it is inevitable that a significant
portion of laser energy will be incident on the surface surround-
ing the particle (Vide infra). Second, the size and the shape of
the particle can be controlled by controlling the size and the
shape of the lithographic area and by controlling the Faradaic
charge passed during the Ag deposition step. Due to the high
spatial resolution of STM lithography, we expect that particles
as small as 10 nm can be fabricated and characterized. Finally,
the interparticle spacing, thus the degree of interactions among
the localized EM field around each particle (Vide infra), can
also be controlled by controlling the etched pit spacings during
STM lithography. In this study, we have demonstrated only
the first advantage.
Figure 6. SFM image and a line profile of the Ag#3 particle shown
in Figure 5. The z-scale in A is 2 µm.
Combining electrochemistry and scanning probe microscopy
for fabricating and characterizing nanometer-scale features is a
very useful concept that has been demonstrated in this work
38,49-51
and elsewhere.
For example, Li et al. reported a method
for depositing Ag nanoparticles on an atomically flat graphite
5
0,51
surface.
This method also has the aforementioned three
advantages; therefore, it is potentially useful for single-particle
SERS studies as well. One drawback is that the laser beam
may overheat the graphite surface and cause various problems
such as heat-induced desorption or morphology changes.
Since STM lithography relies on sequential tip movement to
create patterns, its throughput is lower than those of parallel
Figure 7. Illustration of a spherical particle being imaged by a
pyramidal SFM tip. The dashed line represents the observed topography.
Symbols are defined in the text.
motivation behind this model is to obtain the radius of curvature,
a parameter of great significance in characterizing the roughness
scale of a SERS-active substrate. After a simple derivation
based on geometric algebra, we obtain an equation which relates
the radius of curvature r with three parameters that can be
conveniently determined from an SFM image, namely, the half
base width d, the height h, and the cone angle R′:
5
2,53
54,55
29,56
lithography based on masks,
stamps,
or templates.
However, the spatial resolution of STM lithography is much
higher, and it is less prone to dust particle contamination. These
are desirable characteristics for single particle fabrication which
only demands low-throughput operations.
B. Surface-Enhanced Raman Scattering from Single Ag
Particles
1
r )
(d cos R - h sin R) - rtip
(1)
1
- sin R
Basic Equations of Surface Raman Spectroscopy. Quan-
titative analysis of SERS is the key to a better understanding
of the SERS mechanisms. The integrated surface Raman
where R ) R′ ) arctan(d/h). Since the tip radius of curvature
rtip is much smaller than the particle’s radius, the last term can
be ignored. The value q (q ) h - r) gives us a rough estimate
of particle shape. For example, h ) 2r and q ) r correspond
to a sphere whereas h ) r and q ) 0 correspond to a hemisphere.
-
1
3
intensity ISR in photon s can be expressed as
ISR ) σ′ Ωn P η
(2)
SR
S L

636 J. Phys. Chem. B, Vol. 101, No. 4, 1997
Xiao et al.
where σ′SR is the differential surface Raman scattering cross
2
-1
-1
section in cm sr molecule ; Ω is the solid angle of the
-
2
collecting lens; ns is the surface coverage in molecule cm ;
-1
PL is the incident laser power in photon s ; and η is the overall
detection efficiency which includes the transmission efficiency
of the imaging optics and the quantum yield of the CCD
57
detector. Similarly, the normal Raman intensity INR from bulk
molecular scatters embedded in a thin and transparent film is
INR ) σ′ Ωn P η
(3)
NR
F L
where σ′NRis the differential normal (or unenhanced) Raman
scattering cross section and nF the number of molecules per
geometric area of the thin film.
The surface enhancement factor KEF is defined as
σ′
I n
SR F
ISR
SR
KEF
)
)
)
(4)
σ′
INRnS σ′ Ωn P η
NR S L
NR
where Ω, η, and PL are all assumed to be constants for both
surface and bulk measurements. If σN′ R is unknown, it can be
determined by comparing its Raman intensity with a standard
sample whose σN′ R is known:
B
NR
A
F
I
I
n
n
B
NR
A
NR
σ
)
σ
(5)
A
NR
B
F
Again, Ω, η, and PL should be kept constant. To ensure a
constant η, the frequencies of the two Raman lines should be
near each other.
-
3
Typical parameters used in this study are η ) 1.7 × 10
58
2
2
59
according to eq 3, Ω ) πNA /(1 - NA ) ) 0.145, σ′NR )
-
28
2
-1
2
.7 × 10
cm sr for the CdC stretching vibrations of
Figure 8. SERS spectra of t-4MMS adsorbed on the single Ag particles
shown in Figure 5. The height of each box represents an intensity scale
of 10 counts per second per mW of incident laser power. The
background spectrum was obtained at a particle-free spot: around the
lower-right corner of the image in Figure 5.
60
the probe molecule (t-4MMS) according to eq 5, and nS )
14
4
.61 × 10 molecules/cm2 for a closely-packed SAM of
6
1
t-4MMS. When calculating the enhancement factor due to
one single particle using eq 4, one has to realize that the effec-
tive laser power is much smaller than the actual power. This
is because the laser beam size is often larger than the size
of a single particle. The correction factor can be found
numerically if we assume that the particle is at the center of
the beam and that the beam has a TEM00 Gaussian cross section
at its focal plane:32
Figure 8 shows the surface Raman spectra obtained from
individual Ag particles depicted in Figure 5. The bands at 1587
-
1
and 1626 cm can be assigned to the phenyl ring CdC
stretching vibration and the ethylene CdC stretching vibrations,
64
respectively. Although weak, the signals are clearly distin-
guishable from the background noise. Since the Ag particles
are of different sizes, their SERS intensities also differ. Smaller
diameter usually gives smaller intensity; however, exceptions
to this trend are obvious, such as the case of Ag#3 vs Ag#1,
indicating that the shape factor or the local roughness may be
the causes.
2
PL
2
2
2
0
I(r) )
exp(-2r /r )
(6)
πr0
where PL is the total laser power. The measured spot radius
r0, defined as the radial distance at which the intensity falls to
-2
of the value at the center, is 4.3 ( 0.2 µm.62
Using eq 4, we can now estimate the enhancement factor
e
65
due to one single Ag particle. For example, the enhancement
Correlation between Enhancement Factor and Particle
4
4
factor is 3.5 × 10 for the Ag#3 particle and 1.9 × 10 for the
Ag#4 particle. These values include contributions from all
forms of chemical as well as EM enhancement due to local
roughness at a much smaller scale than the particle size.
However, these contributions are less than 250 as discussed
shortly before.
Morphology. Under optimized experimental conditions, we did
not observe surface Raman signal from any Ag particle shown
in Figure 4. These particles, having a width of about 1.1 µm
and a height of about 10 nm, resemble a “mesa” structure. The
null detection does not imply no enhancement; rather, there exits
an upper enhancement limit which is primarily determined by
the detection efficiency η. After considering the correction
factor for the incident power (about 4%), we find that the upper
limit of the enhancement factor is about 250.6 This value
probably results from chemical enhancement alone since most
mesa surface is “flat” (about 1 nm RMS roughness). EM
enhancement at this roughness scale may also be possible, but
its contribution, capped by 250, is still insignificant when
The above enhancement factors for single Ag particles are 2
orders of magnitude lower than the value measured with an Ag
3
3
electrode roughened electrochemically. We have initially
attributed this discrepancy to the nonideal shape of the Ag
particle. The Ag #3 particle has an aspect ratio of about 1
whereas a typical EM theory predicts that an ideal Ag particle
should be an ellipsoid with an aspect ratio larger than 1.66
However, a more careful analysis of our data obtained with
6
compared with a typical overall enhancement of 1 × 10 .

Hunting for the Active Sites of SERS
J. Phys. Chem. B, Vol. 101, No. 4, 1997 637
bution of isolated particles from that of clusters is difficult. Our
single particle SERS experiments avoid this difficulty and make
quantitative evaluations possible. Furthermore, we have shown
a one-to-one correlation between the observed Raman signal
from a single particle and its high-resolution topographical
image. Therefore, the statistical averaging effect present in
1
4-17,24,27
previous investigations
is completely avoided.
Intuitively, we believe that maximum perturbation of the local
EM field occurs at the “gap sites” where molecules are flanked
by two particles. This hypothesis agrees with the theoretical
results by Gersten and Nitzan who showed that molecules
located between a two-sphere cluster experience a stronger local
EM field and resonance at a lower frequency than those for a
6
9
single sphere. Thus, these “gap sites” are an important class
of SERS-active sites. With further refinement of our single
particle technique, we will fabricate two-particle clusters which
will allow direct comparison with the Gersten and Nitzan model.
Not every EM theory gives the same prediction for the effects
of particle clustering. For example, Laor and Schatz showed
that a cluster of hemispheroids on a perfect conductor will give
a smaller enhancement factor and a broader resonance profile
Figure 9. (A) Typical SERS spectrum of t-4MMS adsorbed on
randomly distributed Ag particles shown in Figure 3. (B) SERS
spectrum of t-4MMS adsorbed on the Ag#3 particle shown in Figure
7
0
than those for a single particle. Although a broader profile
7
1
5
. The y-scale is magnified 100-fold.
has supporting experimental evidence, the smaller enhance-
ment factor clearly contradicts to our observation.
another sample revealed something that demands a more critical
examination of the above interpretation.
This inconsistency, however, may not necessarily imply a
fundamental difference between Laor and Schatz’s theory and
Gersten and Nitzan’s theory. For example, besides the obvious
difference in the geometric configurations of their models, Laor
and Schatz averaged the EM field over the entire particle surface
while Gersten and Nitzan focused only on the special site
between two particles. This implies that the EM enhancement
is not uniformly distributed around a particle surface and that
the special “gap sites” are far more important than other sites
in determining the overall enhancement factor. Our observation
of the extra enhancement by clusters concurs with this picture
qualitatively. Obviously, more quantitative comparisons will
require refinements in both theoretical and experimental cluster
geometries so that they converge to a common one.
The sample of interest is the one shown in Figure 3, in which
a much higher Ag particle density can be seen. The shape of
these particles is hemispherical, which is probably less ideal
than a spherical shape. However, the SERS signal from this
sample is much more intense than the signal from one single
particle, even if one includes the factor that more than one
particles are illuminated by the focused laser beam (Figure 9).
Our suspicion is verified after a quantitative calculation of
the enhancement factor. Since the average particle density does
not change significantly over an area defined by the laser beam
size, we can use an effective surface coverage given by
2
n ) 2πr nMONONP
(7)
S
Conclusions
2
where 2πr is the surface area of a hemispherical particle; nMONO
is the closely-packed monolayer coverage; and NP is the
average particle density. Both r (average of 0.3 µm) and NP
We have demonstrated here a new technique for fabricating
single Ag nanoparticles. The technique shows great potential
for controlling the size, the shape, and the spatial configuration
of model SERS particles.
We have detected for the first time surface-enhanced Raman
scattering due to one isolated single Ag particle. The total
enhancement factor due to one single Ag particle of about 1
µm diameter is about 10 , which includes contributions from
all forms of chemical and EM enhancement at roughness scales
smaller than the dimension of the particle.
Finally, we quantitatively evaluated the role of interparticle
interactions in SERS. Extra enhancement of about 60 is possible
for Ag clusters of random spatial configuration. We speculate
that the extra enhancement is due to the modulation of the local
EM fields from individual particles. In particular, the gap sites
between two particles may be more important to SERS than
other sites. This possibility is significant because it will provide
a valuable guideline for optimizing the surface morphology, thus
the enhancement factor, of a practical SERS substrate. After
further refinement, the single particle fabrication technique
presented here should lead us to a more quantitative understand-
ing of SERS-active sites.
8
2
(
1.1 × 10 particles/cm ) can be readily measured from Figure
3
×
. Using eq 4, we can show that the enhancement factor is 1.1
6
10 , which is about 60 times larger than the average
4
enhancement factor of 1.9 × 10 for a single Ag particle of
comparable size.
4
This result means that the SERS signal due to a cluster of
closely-spaced particles is not simply the sum of signals from
each individual particles. Rather, an extra enhancement due to
interactions between these particles is present. We think the
interactions involve perturbation of the local EM fields sur-
rounding individual particles. This view has been speculated
previously based on the effects of colloid aggregation on
1
8,27,28,67,68
SERS.
For example, Blatchford et al. observed that
SERS intensity from pyridine adsorbed on Au colloidal particles
increases as the degree of aggregation increases.
6
7
They
attributed this phenomenon to the excitation of the longitudinal
EM resonance of string-like clusters, which produces an intense
red-shifted absorption band near 650 nm.
However, previous experimental evidence is only qualitative
evidence since the enhancement factor due to one isolated
particle is virtually unknown. In previous SERS studies, the
observed SERS signal results from a large number of particles
which always include clusters; therefore, separating the contri-
Acknowledgment. L.S. gratefully acknowledges financial
support from National Science Foundation (CHE-9531243). This

638 J. Phys. Chem. B, Vol. 101, No. 4, 1997
Xiao et al.
work was initially supported by the University of Minnesota in
the form of start-up funds. We thank Professor Timothy Lodge
for loaning us a high precision translational stage and Mr.
Mengcheng Chen for assistance in SFM imaging.
(37) Sonnenfeld, R.; Schneir, J.; Hansma, P. K. In Modern Aspects of
Electrochemistry; R. E. White, J. O’M. Bockris, B. E. Conway, Eds.;
Plenum: New York, 1990.
(
38) Ross, C. B.; Sun, L.; Crooks, R. M. Langmuir 1993, 9, 632-636.
(39) Schoer, J. K.; Ross, C. B.; Crooks, R. M.; Corbitt, T. S.; Hampden-
Smith, M. J. Langmuir 1994, 10, 615-618.
(
40) Schoer, J. K.; Zamborini, F. P.; Crooks, R. M. J. Phys. Chem. 1996,
References and Notes
1
00, 11086-11091.
(
(
(
41) Hsu, T. Ultramicroscopy 1983, 11, 167.
42) Schott, J. H.; White, H. S. Langmuir 1992, 8, 1955-1960.
43) Johnson, D. C.; LaCourse, W. R. Anal. Chem. 1990, 62, 589A-
(
1) Jeanmaire, D. L.; Van Duyne, R. P. J. Electroanal. Chem. 1977,
8
4, 1-20.
(2) Albrecht, M. G.; Creighton, J. A. J. Am. Chem. Soc. 1977, 99,
5
96A.
5
215-5217.
(44) The Ag particles shown in Figures 4 and 5 were deposited at 500
(
3) Van Duyne, R. P. In Chemical and Biochemical Applications of
and 470 mV, respectively. These values are higher than the theoretical
threshold of 425 mV. We think that the difference is caused by liquid
junction potential and other factors associated with a non-ideal reference
electrode.
Lasers; C. B. Moore, Ed.; Academic Press: New York, 1979; p 101
4) Surface Enhanced Raman Scattering; R. K. Chang, T. E. Furtak,
Eds.; Plenum Press: New York, 1982.
5) Otto, A. In Light Scattering in Solids; M. Cardona, G. Guntherodt,
Eds.; Springer-Verlag: Heidelberg, 1984; p 289.
(
(
(
(
(
45) Butt, H.-J.; Berharz, B. Langmuir 1995, 11, 4735-4741.
46) Keller, D. Surf. Sci. 1991, 253, 353.
(
(
(
(
6) Moskovits, M. ReV. Mod. Phys. 1985, 57, 783.
7) Schatz, G. C. Acc. Chem. Res. 1984, 17, 370-376.
8) Kerker, M. Acc. Chem. Res. 1984, 17, 271-277.
9) Ueba, H. Surf. Sci. 1983, 131, 347-366.
47) Albrecht, T. R.; Akamine, S.; Carver, T. E.; Quate, C. F. J. Vac.
Sci. Technol. A 1990, 8, 3386-3396.
48) Sondag-Huethorst, J. A. M.; Fokkink, L. G. J. Langmuir 1995, 11,
823-4831.
49) Sun, L.; Crooks, R. M. Langmuir 1993, 9, 1951-1954.
(50) Li, W.; Virtanen, J. A.; Penner, R. M. Langmuir 1995, 11, 4361-
4365.
(51) Li, W.; Virtanen, J. A.; Penner, R. M. J. Phys. Chem. 1994, 98,
11751.
(52) Chan, K. C.; Kim, T.; Schoer, J. K.; Crooks, R. M. J. Am. Chem.
(
4
(
(
10) Otto, A. J. Raman Spectrosc. 1991, 22, 743-752.
11) Campion, A.; Ivanecky, J. E., III; Child, C. M.; Foster, M. J. Am.
(
Chem. Soc. 1995, 117, 11807-11808.
12) Campion, A. J. Electron Spectrosc. Relat. Phenom. 1983, 29, 397-
00.
13) Here, the surface morphology refers to the size, shape, and spatial
(
4
(
distribution of metal particles that are either supported on a substrate, such
as Ag islands on a glass slide, or merged with a substrate, such as protrusions
on an electrochemically roughened electrode surface. A related term, namely,
Soc. 1995, 117, 5875-5876.
(53) Berggren, K. K.; Bard, A.; Wilbur, J. L.; Gillaspy, J. D.; Helg, A.
G.; McClelland, J. J.; Rolston, S. L.; Phillips, W. D.; Prentiss, M.;
Whitesides, G. M. Science 1995, 269, 1255-1257.
“
surface roughness”, has a more narrow meaning: it only defines an average
particle size and does not necessarily include a connotation of spatial
distribution.
(54) Jackman, R. J.; Wilbur, J. L.; Whitesides, G. M. Science 1995,
269, 664-666.
(
14) Van Duyne, R. P.; Hulteen, J. C.; Treichel, D. A. J. Chem. Phys.
993, 99, 2101-2115.
15) Schueler, P. A.; Ives, J. T.; DeLaCroix, F.; Lacy, W. B.; Becker,
P. A.; Li, J.; Caldwell, K. D.; Drake, B.; Harris, J. M. Anal. Chem. 1993,
1
(55) Chou, S. Y.; Krauss, P. R.; Renstrom, P. J. Science 1996, 272,
85-87.
(
(56) Masuda, H.; Fukuda, K. Science 1995, 268, 1466-1468.
(57) Note that the surface coverage nS refers to the number of molecules
per geometric surface area since the area term originates from the cross
sectional area of the laser beam. In addition, if the sample surface is
homogeneous over a scale of beam size, then IR depends only on the laser
power PL, but not on the area of illumination.
6
5, 3177-3186.
(
(
(
16) Roark, S. E.; Rowlen, K. L. Anal. Chem. 1994, 66, 261-270.
17) Otsuka, I.; Iwasaki, T. J. Vac. Sci. Technol. A 1990, 8, 530-533.
18) Sun, L. Ph.D. Thesis, Northwestern University, Evanston, IL,
1
990.
(
(
(
19) Furtak, T. E. J. Electroanal. Chem. 1983, 150, 375-388.
(58) In eq 3, INR was measured as the integrated intensity of the 992
-
1
-30
2
20) Pockrand, I.; Otto, A. Solid State Commun. 1981, 38, 1159-1163.
cm band from a 1-mm thick benzene film, and σ′NR ) 7.1 × 10
cm
-
1
21) Buncick, M. C.; Warmack, R. J.; Ferrell, T. L. J. Opt. Soc. Am. B
sr as reported in Trulson, M. O.; Mathies, R. A. J. Chem. Phys. 1986,
84, 2068-2074.
1
987, 4, 927-933.
22) Moody, R. L.; Vo-Dinh, T.; Fletcher, W. H. Appl. Spectrosc. 1987,
1, 966-970.
23) Grabar, K. C.; Freeman, R. G.; Hommer, M. B.; Natan, M. J. Anal.
Chem. 1995, 67, 735-743.
24) Freeman, R. G.; Grabar, K. C.; Allison, K. J.; Bright, R. M.; Davis,
(
(59) NA, the numerical aperture, is sin θmax in air, where θmax is the
angle between the optical axis and the marginal ray entering the objective.
4
2
(
In comparison, Ω ) π(tan θmax) .
(60) The integrated intensity of the CdC stretching bands from 1-mm
thick 0.14 M t-4MMS solution was compared with the intensity of the 992
(
-
1
20
J. A.; Guthrie, A. P.; Hommer, M. B.; Jackson, M. A.; Smith, P. C.; Walter,
cm band from 1-mm thick neat benzene. Thus, nF ) 6.74 × 10
2
1
8
2
D. G.; Natan, M. J. Science 1995, 267, 1629-1632.
molecules/cm for benzene, and nF ) 8.32 × 10 molecules/cm for
(25) Grabar, K. C.; Smith, P. C.; Musick, M. D.; Davis, J. A.; Walter,
t-4MMS.
D. G.; Jackson, M. A.; Guthrie, A. P.; Natan, M. J. J. Am. Chem. Soc.
(61) We obtained this value by ignoring the local surface roughness
and assuming a closely-packed SAM of (x3 × x3)R30 lattice on an
Ag(111) surface. The (x3 × x3)R30 lattice distance can be calculated
from the nearest neighbor distance of an Ag fcc crystal (2.89 Å).
(62) The laser spot size was measured following the method in ref 32.
The measurement uncertainty is primarily determined by the resolution of
the one-dimensional micrometer (0.1 µm) used in the measuring apparatus.
(63) We may define the upper limit as the enhancement factor in eq 4
when the noise is comparable to the signal, i.e., ISR ) INOISE. The measured
spectrum baseline noise is about 1 count per second per mW of incident
power.
1
996, 118, 1148-1153.
26) Chumanov, G.; Sokolov, K.; Cotton, T. M. J. Phys. Chem. 1996,
00, 5166-5188.
27) Liao, P. F.; Bergman, J. G.; Chemla, D. S.; Wokaun, A.; Melngailis,
J.; Hawryluk, A. M.; Economou, N. P. Chem. Phys. Lett. 1981, 82, 355.
28) Bergman, J. G.; Chemla, D. S.; Liao, P. F.; Glass, A. M.; Pinczuk,
(
1
(
(
A.; Hart, R. M.; Olson, D. H. Opt. Lett. 1981, 6, 33-35.
(
(
29) Martin, C. R. Science 1994, 266, 1961-1965.
30) Treichel, D. A. Ph.D. Thesis, Northwestern University, Evanston,
IL, 1991.
(
31) Van Duyne, R. P.; Haller, K. L.; Altkorn, R. I. Chem. Phys. Lett.
(64) Baranovic, G.; Meic, Z.; Gusten, H.; Mink, J.; Keresztury, G. J.
Phys. Chem. 1990, 94, 2833-2843.
1
986, 126, 190.
(
32) Haller, K. L. Ph.D. Thesis, Northwestern University, Evanston, IL,
988.
33) SERS from colloidal Ag particles has been observed under a
(65) Since nS refers to the surface coverage per geometric area, it should
be equal to the monolayer coverage multiplied by a geometric factor. The
factor is simply equal to the ratio of the particle’s surface area to its projected
1
(
2
2
spectroscopic microscope. However, the detected signal might result from
a cluster of particles since higher resolution images of the observed spots
were not reported. See, for example, (a) Emory, S. R.; Nie, S., Pittsburgh
Conference, 1996 Meeting Abstract, and (b) Vlckova, B.; Gu, X. J.; Tsai,
D. P.; Moskovits, M. J. Phys. Chem. 1996, 100, 3169-3174.
geometric area, i.e., 4πr /πr ) 4 for spherical particles.
(66) Zeman, E. J.; Schatz, G. C. J. Phys. Chem. 1987, 91, 634-643.
(67) Blatchford, C. G.; Campbell, J. R.; Creighton, J. A. Surf. Sci. 1982,
120, 435-455.
(68) Freeman, R. G.; Hommer, M. B.; Grabar, K. C.; Jackson, M. A.;
Natan, M. J. J. Phys. Chem. 1996, 100, 718-724.
(69) Gersten, J. I.; Nitzan, A. Surf. Sci. 1985, 158, 165-189.
(70) Laor, U.; Schatz, G. C. Chem. Phys. Lett. 1981, 82, 566-570.
(71) Wokaun, A.; Gordon, J. P.; Liao, P. F. Phys. ReV. Lett. 1982, 48,
957-960.
(
34) Wardell, J. L. In The Chemistry of the Thiol Group; S. Patai, Ed.;
Wiley: New York, 1974; p 186.
35) Bain, C. D.; Troughton, E. B.; Tao, Y.-T.; Evall, J.; Whitesides,
G. M.; Nuzzo, R. G. J. Am. Chem. Soc. 1989, 111, 321-335.
36) Tsen, M.; Sun, L. Anal. Chim. Acta 1995, 307, 333-340.
(
(