Synthesis and linear optical properties of nanoscopic gold particle pair structures

Article abstract of DOI:10.1021/jp992176x

Pair particle nanostructures were prepared in porous anodic aluminum oxide films via a modified template synthesis procedure. By varying the interparticle distance of centrosymmetric gold spherical particles (radius ca. 32 nm), the linear UV/visible polarization spectra of the gold particle/porous alumina film composites can be altered. The resulting spectra can be explained qualitatively in terms of particle electromagnetic interactions within an individual pore through simple quasistatic limit scattering models. Modified long wavelength approximation and discrete dipole approximation methods yield similar calculated spectra, and neither approach quantitatively describes the experimental spectra. Discrepancies are interpreted to arise in part from a distribution of interparticle spacings in the experimental system. Rod, rod-pair, and rod-sphere gold nanoparticle structures were also synthesized. Possible evidence for interaction between the two members of the rod-sphere structure is seen in the experimental spectra. The polarization spectra of the rod-containing systems are also only qualitatively described by quasistatic limit models.

Full text of DOI:10.1021/jp992176x

11398
J. Phys. Chem. B 1999, 103, 11398-11406
Synthesis and Linear Optical Properties of Nanoscopic Gold Particle Pair Structures
Marie L. Sandrock and Colby A. Foss, Jr.*
Department of Chemistry, Georgetown UniVersity, Washington, District of Columbia 20057
ReceiVed: June 28, 1999; In Final Form: September 16, 1999
Pair particle nanostructures were prepared in porous anodic aluminum oxide films via a modified template
synthesis procedure. By varying the interparticle distance of centrosymmetric gold spherical particles (radius
ca. 32 nm), the linear UV/visible polarization spectra of the gold particle/porous alumina film composites
can be altered. The resulting spectra can be explained qualitatively in terms of particle electromagnetic
interactions within an individual pore through simple quasistatic limit scattering models. Modified long
wavelength approximation and discrete dipole approximation methods yield similar calculated spectra, and
neither approach quantitatively describes the experimental spectra. Discrepancies are interpreted to arise in
part from a distribution of interparticle spacings in the experimental system. Rod, rod-pair, and rod-sphere
gold nanoparticle structures were also synthesized. Possible evidence for interaction between the two members
of the rod-sphere structure is seen in the experimental spectra. The polarization spectra of the rod-containing
systems are also only qualitatively described by quasistatic limit models.
Introduction
Theoretical Considerations
The porous anodic alumina films employed as template hosts
in this study are nonisotropic; they contain parallel arrays of
pores which run normal to the face of the films. The optical
spectra of the resulting gold nanoparticle-alumina composite
materials thus depend strongly on the orientation of the film
and the nanometal particle structures relative to the incident
electric field. For nonnormal incidence under p-polarization
(incident electric field polarized parallel to the plane of
incidence), the plasmon resonance spectra of the present Au
particle/alumina composites exhibit significant dependence on
the incidence angle θ (the angle between the propagation vector
and the sample surface normal); increasing θ typically leads to
a red-shift in the plasmon resonance maximum or to the
appearance of a second long-wavelength resonance. We review
below the simple quasistatic limit treatments that describe, at
least qualitatively, the spectral trends seen in experiment.
Single Particle Scattering Theory: Quasistatic Limit. The
spectral properties of isolated nanoscopic particles can be
understood on the basis of the well-known Clausius-Mossotti
expression for particle polarizability:17,18
It is well known that nanoscopic metal particles can exhibit
strong absorption and/or scattering bands in the visible spectrum.
This effect allows for a wide variety of practical applications
of nanoscopic metal systems in the fine arts,1 surface-enhanced
spectroscopy,2,3 nonlinear optics,4-6 and biochemistry.7 Signifi-
cantly, the size and shape of these particles can be tuned to
produce the desired linear optical effect, such as a specific
spectral band position (λmax) or an optimized local field
enhancement factor for surface-enhanced Raman scattering
studies.
The preparation of nanoscopic metal particles of nonspherical
geometry has attracted much attention in the past decade.
Lithographic methods have been used to prepare uniform arrays
of ellipsoids8 and truncated tetrahedra.9 The template synthesis
methods developed by Moskovits10,11 and Martin12,13 rely on
the pore geometry of porous host materials to direct particle
growth. Bulk solution synthesis of rodlike metal particles has
been accomplished using surfactants that form cylindrical
micelle superstructures in solution.14 All of these synthesis
methods have yielded nanoscopic metal particles whose UV/
visible spectra are characteristic of nonspherical geometries. In
the case of the porous template- or micelle-derived rods, the
extinction spectra show long- and short-axis plasmon resonance
bands which are the signature of such prolate structures.6,12,14,15
We recently extended the template synthesis method to
prepare noncentrosymmetric nanoparticle structures involving
two gold particles6 and gold/silver iodide heterojunction par-
ticles.16 In this paper, we describe how this modified template
synthesis method can be used to prepare a variety of nanoparticle
pair structures such as paired spheres, paired rods, and rod-
sphere pairs. We discuss the plasmon resonance spectra of these
different systems and, in the case of the paired-spheres, the
spectral consequences of interparticle spacing. The experimental
results are compared with spectra predicted from simple
quasistatic theory and more advanced electrodynamic treatments.
4πꢀoab2
ꢀm - ꢀh
3Lx ꢀm + κꢀh
R )
(1)
where ꢀm and ꢀh are the complex dielectric constants of the
particle and host medium, respectively, and ꢀo is the permittivity
of vacuum. A rodlike structure can be reasonably approximated
as an ellipsoid of revolution, with semimajor axis a and
semiminor axis b. The parameters Lx and κ are the depolarization
and screening factors, respectively. They depend only on particle
shape and orientation in the incident field and are related to
each other via κ ) Lx-1 - 1. The calculation of extinction cross-
section Cext from the particle polarizability is straightforward.17,18
Interacting Particle Pairs. Quasistatic limit expressions for
two-particle and larger clusters have been described by Kreibig
and Vollmer.19 The derivation of the expression for an axisym-
metric structure of two particles is simple and worthy of a brief
* Corresponding author e-mail fossc@gusun.georgetown.edu.
10.1021/jp992176x CCC: $18.00 © 1999 American Chemical Society
Published on Web 12/03/1999

Nanoscopic Gold Particle Pair Structures
J. Phys. Chem. B, Vol. 103, No. 51, 1999 11399
summary here. The electric dipole moments of the two particles
are given by
µ1 ) ꢀhR1 E1
µ2 ) ꢀhR2 E2
(2a)
(2b)
where E1 and E2 are the electric fields experienced by particles
1 and 2, respectively. The field at each particle may be assumed
to be the sum of the applied field E in the material and a
contribution arising from the other particle. For example, the
fields at particles 1 and 2 can expressed as
µ2
ꢀhꢀod12
E1 ) E + g
E2 ) E + g
(3a)
(3b)
3
3
µ1
ꢀhꢀod12
where d12 is the center-to-center distance between the two
particles and g is a geometric factor that depends on the
orientation of the particle pair in the applied field. Substitution
of eqs 2 into eqs 3, and subsequent combination of 3a and 3b
leads to two expressions which describe the relationship between
the applied field and the actual field at a given particle:
Figure 1. Schematic of gold nanoparticle pair synthesis. One face of
the porous anodic alumina film (1) is sputtered with Ag (2). A silver
foundation is electrochemically deposited (3). Additional layers of Au,
Ag spacer, and Au again are electrochemically deposited onto the silver
foundation (4). The Ag foundation and spacers are removed by nitric
acid etching (5) to yield the Au pair structures. The letters s, c, and n
refer to spheres, centrosymmetric, and noncentrosymmetric structures,
respectively.
R1R2
R2
E1 1 - g2
) E 1 + g
(4a)
(4b)
6
3
(
)
)
(
)
)
d12
d12
R1R2
R1
E2 1 - g2
) E 1 + g
method of Furneax et al.21 The details of our procedure have
been given elsewhere.13,15 In this study, we employed porous
alumina films prepared via anodization of aluminum foil
(Aldrich, 99.999%) at 20 V in 6% sulfuric acid at 0 °C. The
mean pore diameters of these films were determined (via TEM
analysis) to be 32 ( 1 nm. Typical film thicknesses were in
the 40-60 µm range.
Gold Nanoparticle Preparation. The porous alumina films
were used as a template to synthesize gold rods, rod-pair
structures, paired rod-sphere structures, and paired spheres with
two different distances of separation. To create a conductive
foundation upon which subsequent electrochemical deposition
would occur, the barrier side of the porous alumina films were
sputtered with ca. 45 nm of silver using an Anatech Hummer
10.2 plasma deposition device.
Electrochemical deposition of Ag and Au were done using a
three-electrode cell equipped with an Ag/AgCl reference
electrode, a platinum mesh counter electrode, and the Ag-
sputtered porous alumina film as the working electrode (area
3.14 cm2). A 0.80-1.27 C/cm2 foundation of silver was
electrodeposited over the 3.14 cm2 film area potentiostatically
(-0.6 V vs Ag/AgCl) from a silver thiocyanate (Ag(I)) plating
solution.22 This foundation ensured that the subsequent Au
deposition would occur at a depth in the film where the pore
structure is well-defined. Au layers were deposited potentio-
statically (-0.9V vs Ag/AgCl) from an Au(I) plating solution
(Orotemp 24, Technic, Inc.).
6
3
(
(
d12
d12
The average polarizability R of the particle pair is defined
according to the relations
1
2
1
(µ1 + µ2) ) ꢀ (R1E1 + R2E2) ) ꢀh R E
(5)
(6)
h2
Combining eqs 4 and 5 leads to
R2
R1
R1 1 + g
+ R2 1 + g
3
3
(
)
(
)
d12
d12
1
2
R )
R1R2
1 - g2
6
(
)
d12
To obtain the extinction spectrum for the particle pair
structure, the average polarizability is inserted into the standard
C
ext expressions. For the case of an electric field incident along
the particle pair axis (longitudinal polarization), the geometric
factor g is equal to 1/2π. For an electric field incident
perpendicular to the axis of rotation (transverse polarization),
g ) -1/4π.20 For intermediate incidence angles θ, the average
extinction is the weighted sum of the longitudinal and transverse
polarization cases:18
Cext(θ) ) Cext,| sin2(θ) + Cext, cos2(θ)
(7)
The electrochemical template synthesis scheme for the paired
spheres, paired rods and rod-sphere pairs is shown in Figure
1. For closely spaced paired sphere structures, the first Ag
deposition was followed by deposition of 0.03 C/cm2 of Au(I),
0.02C/cm2 of Ag(I), and a second Au(I) deposition of 0.03
C/cm2. The Ag foundation and spacer structures were then
removed by soaking the films in nitric acid (EM Sciences, 69-
71%) for twenty minutes. The films were then rinsed in
deionized water, ethanol, and acetone, and then allowed to dry
Equation 7 pertains specifically to the case where the
rotational axes of the particles are perpendicular to the surface
planes of the thin film sample and the incident field is
p-polarized (vide infra).
Experimental Section
Porous Aluminum Oxide Synthesis. Porous aluminum oxide
films were prepared using the anodization/voltage reduction

11400 J. Phys. Chem. B, Vol. 103, No. 51, 1999
Sandrock and Foss
containing the Au nanostructures were immersed in resin and
microtomed (to ca. 70 nm thickness). The microtomed samples
were then placed on 200 mesh Cu grids (EM Sciences) for
examination in a JEOL 1200 EX microscope. Details of sample
preparation and TEM analysis have been described elsewhere.6
Spectroscopic Analysis of the Nanoparticle Composites.
Linear polarization analysis of the composite films was per-
formed using a Hitachi U-3501 spectrophotometer equipped with
a model 210-2130 polarizer accessory. Extinction spectra were
obtained between λ ) 350 nm and λ ) 800 nm at normal
incidence (defined as θ ) 0° and corresponding to the case
where the incident electric field is perpendicular to the pore
axes) and at incidence angles θ ) 20°, 30°, and 45°, under
p-polarization (the incident electric field polarized parallel to
the plane of incidence). A schematic of the sample/polarizer
configuration is shown in Figure 2.
Figure 2. Schematic of polarization spectroscopy setup. The sample
film is rotated about an axis perpendicular to the page. The angle of
incidence is indicated by θ. p-Polarization corresponds to an incident
electric field polarized in the plane of the page. s-Polarization
corresponds to an electric field oscillating normal to the page.
Results and Discussion
at room temperature. The films were then placed in an oven
for 30 min at 130 °C to anneal the Au layers to a spherical
geometry. For the paired spheres with larger spacing, the
procedure was the same except that 0.03 C/cm2 of Ag(I) was
deposited as the spacer layer.
The centrosymmetric paired rod structures were prepared in
a similar fashion. After the initial Ag foundation was deposited,
two layers of Au(I) were deposited (0.19 C/cm2), separated by
an Ag spacer layer formed via deposition of 0.03 C/cm2 Ag(I).
The rod-sphere structures were formed via deposition of 0.03
C/cm2, 0.03 C/cm2, and 0.16 C/cm2 of Au(I), Ag(I), and Au(I),
respectively. The Ag layers were removed via HNO3 etching
and rinsed in deionized water, ethanol, and acetone. Oven
annealing was not performed on rod-containing composites.
Nanoparticle Characterization. Transmission electron mi-
croscopy (TEM) was used to characterize the dimensions of
the template-synthesized gold nanostructures. The alumina films
Transmission Electron Microscopy. Figure 3 shows the
TEM images of the paired-sphere structures in porous alumina.
These images are cross-sectional. For the system in which 0.02
C/cm2 Ag(I) were deposited to form a spacer layer, the average
Au sphere diameter is 35 ( 4 nm. The surface-to-surface
separation distance is 20 ( 10 nm. For the pair system in which
0.03 C/cm2 Ag(I) was deposited as the spacer, the average
spacing is 84 ( 6 nm. The average sphere diameter was found
to be 30 ( 2 nm.23
Figure 4 depicts cross-sectional TEM images of both cen-
trosymmetric and noncentrosymmetric pair structures and a
single-Au rod composite prepared for spectroscopic comparison
(Figure 4A). The single rod structures have an average length
a ) 133 ( 20 nm and diameter b ) 31 ( 4 nm. The individual
rods in the centrosymmetric rod-pair structures (Figure 4B)
have an average length a ) 70 ( 10 nm and a diameter b ) 30
( 3 nm. The interparticle spacing is 70 ( 20 nm. The
Figure 3. Transmission electron microscope images of 0.03 C/cm2 Au centrosymmetric Au pair particles in an alumina host. (A) 16 nm radius Au
particles with smaller interparticle spacing after deposition of 0.02 C/cm2 Ag(I) spacer. (B) 16 nm radius Au particles with larger spacing after
deposition of 0.03 C/cm2 Ag(I) spacer. Scale bar in both cases corresponds to 50 nm.

Nanoscopic Gold Particle Pair Structures
J. Phys. Chem. B, Vol. 103, No. 51, 1999 11401
Figure 4. Transmission electron microscope images. (A) 0.19 C/cm2 Au centrosymmetric particles in alumina host. (B) 0.19 C/cm2 Au centrosymmetric
pair particles in alumina (0.10 C/cm2 Au(I) deposited for both segments). (C) 0.19 C/cm2 Au noncentrosymmetric pair particles in alumina host
(0.16 C/cm2 Au(I) deposited for larger segment, 0.03 C/cm2 Au(I) deposited for small segment). Scale bar in all cases corresponds to 50 nm.
noncentrosymmetric pair structures (Figure 4C) are composed
of a rodlike segment (a ) 100 ( 10 nm, b ) 33 ( 4 nm) and
a sphere-like segment (a ) 34 ( 5 nm, b ) 33 ( 4 nm). The
separation distance between the rod and sphere segments is 57
( 13 nm.
Effect of Interparticle Spacing in Paired-Sphere Struc-
tures. UV/vis polarization spectra of the Au/Al2O3 composite
films were measured using the sample configuration shown in
Figure 2. The spectra were obtained at normal incidence and at
incidence angles θ ) 20°, 30°, and 45°, with the electric field
polarized in the plane of incidence (p-polarization). Figure 5
shows the spectra of the paired sphere structures whose TEM
images are shown in Figure 3. For the Au pairs with large
interparticle spacing (5A), the plasmon resonance maximum λmax
occurs at 548 nm for all incidence angles. For the pair structures
with smaller separation distance (5B), λmax occurs at ca. 539

11402 J. Phys. Chem. B, Vol. 103, No. 51, 1999
Sandrock and Foss
θ, but a λmax at normal incidence that is lower than that of
isolated 32 nm Au spheres in alumina.
We first compared the experimental spectra with those
calculated for 32 nm Au sphere pairs using the quasistatic limit
treatment for R given by eq 6. Optical data for gold were taken
from Johnson and Christy,24 and the host oxide refractive index
was assumed to be 1.33. When the interparticle spacing d12 is
assumed to be 114 nm, the plasmon resonance band shows no
dependence on the incidence angle. At 52 nm spacing, increasing
θ results in a red-shifted plasmon resonance maximum, a result
in qualitative accord with experiment. However, the absolute
values for λmax are underestimated relative to experiment at all
incidence angles. Furthermore, for the 52 nm separation pair,
total difference ∆λmax () λmax (45°) - λmax(0°)) is ca. 20 nm
in experiment but only ca. 6 nm in the spectra simulated using
eq 6.
Calculations based on eq 6 assume that the Au particles are
negligibly small relative to the incident wavelength. Of course,
we have not achieved this condition in experiment. To address
the finite size of the Au particles, we calculated polarization
spectra for 32 nm Au spheres using the well-known modified
long wavelength approximation (MLWA). The MLWA involves
replacing the particle size independent depolarization factor Lx
in eq 1 with the size-dependent quantity25
1
3
2
9
Lx,eff ) Lx - k2 a2 - ik3a3
(8)
where k is the wave vector ()2πn˜h/λ) and a is the sphere radius.
Figures 7A and B show extinction spectra calculated for the
Au spheres separated by 84 and 20 nm gaps, which correspond
to d12 values of ca. 114 and 52 nm, respectively.26 At large
separation (Figure 7A), the dependence of λmax on θ is still
negligible, though the λmax values themselves are red-shifted
slightly relative to the simple quasistatic result. For d12 ) 52
nm, ∆λmax is ca. 10 nm, compared to 6 nm in the quasistatic
calculation. In the case of contacting Au spheres, the θ-depen-
dence of the extinction intensity and ∆λmax are both much more
pronounced, with the latter quantity ca. 30 nm.
Figure 5. Plasmon resonance spectra of 16 nm radius Au sphere pair
systems (p-polarization). (A) Large interparticle spacing (d12 ca. 114
nm). (B) Smaller interparticle spacing (d12 ca. 52 nm). In both figures,
an increasing overall extinction corresponds to increasing incidence
angle θ (0, 20, 30, and 45 degrees). All spectra were obtained in double-
beam mode with no material in the reference beam.
Another factor in the experiment, theory discrepancy is the
distribution of interparticle separations. An average surface-
surface separation of 20 ( 10 nm implies a significant fraction
of pairs with smaller separation distances whose λmax depen-
dence on θ is stronger than that shown in Figure 7B. For
example Figure 7C shows polarization spectra calculated for
the extreme case of 32 nm Au spheres in contact (d12 ) 32
nm). The net change in the extinction maximum ∆λmax is 24
nm, a number comparable to experiment. Conversely, the
population fraction that involves larger separations will show a
weaker dependence of λmax on θ. We find that the calculated
extinction cross section increases with decreasing interparticle
separation. This implies that the measured spectrum, which is
the sum of the extinction of all particle pairs, may tend to be
dominated by the more closely spaced particles. Indeed, the
experimental plasmon resonance bands broaden and become
asymmetric as the incidence angle increases, an observation
consistent with a distribution of interparticle separations.
Introducing particle size corrections into the simple quasistatic
theory yields only a slight improvement in the theory-
experiment comparison. Indeed, regardless of particle size
corrections, eq 6 assumes that neighboring particles can be
treated as point dipole field sources and necessarily avoids the
details of surface charge distribution. Furthermore, only dipole
induction modes are considered; quadrupole and higher multi-
pole modes are ignored. Finally, the finite size and center-to-
Figure 6. Experimental λmax (nanometers) versus incidence angle θ
(p-polarization) for Au sphere pair systems. Open circles: d12 ) 114
nm. Closed circles: d12 ) 52 nm.
nm at normal incidence and progressively red-shifts with
increasing incidence angle. At θ ) 45°, λmax ) 558 nm. The
width of the plasmon resonance band also increases with
increasing θ. In both sets of spectra, the overall extinction
intensity increases with increasing incidence angle. This ob-
servation is consistent with an increasing effective optical path
length through the porous aluminum oxide. The λmax trend with
incidence angle for the two Au sphere-pair structures is shown
in Figure 6. The Au sphere structures with an average separation
distance of 20 nm show not only a significant dependence on

Nanoscopic Gold Particle Pair Structures
J. Phys. Chem. B, Vol. 103, No. 51, 1999 11403
collection of dipole elements exhibits features that correspond
to quadrupole and higher multipole induction modes. A thorough
evaluation of the DDA applied to nanoscopic metal particles,
including comparisons with Mie theory and the MLWA, is given
by Jensen et al.30
We have applied the DDA to 32 nm gold particle pairs at
114, 54, and 32 nm center-to-center separations. The host
medium real refractive index was assumed to be 1.33. The
number of polarizable elements assumed for each particles was
large (1791 to 14 398), the optimal number depending on the
interparticle spacing.31 Extinction spectra were calculated for
the limiting cases where the incident electric field was polarized
parallel and perpendicular to the pair axis. Spectra corresponding
to incidence angles θ ) 0°, 20°, 30°, and 45° were then
produced using eq 7. The results of the DDA calculations are
shown in Figure 8. They are surprisingly similar to the MLWA
results in terms of peak position and intensity trends. In the
case of the spheres in contact, the DDA spectra show an increase
in extinction at long wavelengths, a feature not seen in the
quasistatic and MLWA results. This long wavelength feature
in the DDA result may be due to a second Au particle resonance
in the near-infrared spectrum. Calculations by Jensen et al. show
two major resonance bands for contacting 30 nm silver spheres
in a vacuum, one near 400 nm and another near 800 nm. The
behavior of two contacting gold spheres should be similar but
with both resonances at longer wavelengths. We should note,
however, that we do not see evidence of a second resonance in
our experimental UV/vis spectra.
For gold spheres in close proximity, all of the theoretical
treatments considered predict a spectroscopic red-shift in λmax
with increasing incidence angle θ. While this result is in
qualitative accord with experiment, the ubiquitous prediction
of a large increase in extinction intensity accompanying the red-
shift is not seen in the experimental data. It should be noted
that all of the spectral simulations are based on so-called single-
particle scattering models; extinction is assumed to arise solely
from the gold particle structures and not from absorption and/
or reflection losses due to the host material. In the experimental
systems, the net effect of reflection and absorption losses from
the alumina host is an increasing vertical displacement of the
plasmon resonance bands with increasing incidence angle. Since
the spectrum of anodic aluminum oxide is featureless between
ca. 400 and 800 nm,13 this displacement is roughly constant
over the visible spectral range and cannot act to selectively
diminish the long-axis plasmon resonance bands (or selectively
enhance the short-axis bands).
Polarization Spectra of Rods, Rod Pairs and Rod-Sphere
Pairs. Figure 9 shows the UV/vis spectra of various Au particle/
alumina composites at four incidence angles under p-polariza-
tion. While the absolute amount of gold deposited in the three
composites is similar, the manner in which the metal is
distributed in the composite has a direct effect on the plasmon
resonance spectra. Figure 9A shows the polarization spectra for
the composite in which the gold is deposited in a single step to
yield long rods. At normal incidence, a single band centered at
516 nm corresponding to the rod’s short-axis polarization is
observed. As the incidence angle θ is increased, a second band
corresponding to the long-axis resonance appears, the λmax of
which is centered in the near-infrared spectrum and out of the
working range of the film polarizer.
Figure 7. Plasmon resonance spectra of 16 nm radius gold sphere
pairs calculated for various incidence angles under p-polarization using
MLWA model. (A) d12 ) 114 nm. (B) d12 ) 52 nm. (C) d12 ) 32 nm
(spheres touching). Increasing overall extinction corresponds to the
incidence angle θ progression: 0, 20, 30, and 45 degrees.
center separation distances between two nanoscale particles
results in their experiencing different phases of the incident
oscillating field.19,27 This issue is also ignored in models based
on eq 6.
A more rigorous treatment of the extinction spectrum of
interacting particle pairs and other complex structures is clearly
needed and is fortunately available in the so-called discrete
dipole approximation (DDA).28,29 The DDA treats each particle
as a cubic array of polarizable elements, each element capable
of producing only a dipolar response to the applied field.
However, the extinction spectrum of a particle treated as a large
Subdividing the gold deposit into two 70 nm long segments
separated by a 70 nm gap leads to the spectra shown in Figure
9B. At normal incidence the spectrum shows a single band
centered at ca. 517 nm. As θ is increased, the long-axis

11404 J. Phys. Chem. B, Vol. 103, No. 51, 1999
Sandrock and Foss
Figure 9. UV/visible p-polarization spectra at incidence angles 0, 20,
30, and 45 degrees. (A) Centrosymmetric Au particle/alumina com-
posite. (B) Centrosymmetric Au pair particle/alumina composite. (C)
Noncentrosymmetric pair particle/alumina composite. Incidence angles
are indicated next to curves.
Figure 8. Plasmon resonance spectra of 16 nm radius gold sphere
pairs calculated using the DDA model. Interparticle spacings for A, B
and C as given in Figure 7. The number of polarizable elements assumed
wavelength than in the cases of the other two rod-containing
composites. We interpret this as resulting from a large extinction
contribution from the sphere member of the pair. As the
incidence angle is increased, a long axis resonance appears,
centered at about 725 nm. While the general trends with θ are
similar to the other rod-containing composites, the rod-sphere
pair system shows an additional feature. As the incidence angle
increases, a shoulder band develops on the long-wavelength
slope of the resonance centered at 520 nm.
The polarization spectra of the gold rod, rod-pair, and rod-
sphere pair composites can be explained qualitatively on the
basis of the quasistatic theory outlined earlier. At normal
incidence, the electric field is directed exclusively across the
short axis of the gold rods. In this orientation, the screening
for each sphere are: A: 1791; B: 4224; C: 14398. Note that Qext
C
)
ext/πa2.
resonance centered at 660 nm appears. The λmax values for the
long-axis plasmon resonance band is close to that observed for
gold rods of similar dimensions (but not present as pairs)
embedded and oriented in polyethylene.15 This suggests that
the 70 nm gap is large enough to preclude strong interparticle
interactions and that the two segments behave essentially as
independent rods.
Figure 9C shows the polarization spectra of the Au/alumina
composites in which the deposited gold is distributed into
asymmetric rod-sphere structures. At normal incidence, the
single plasmon resonance band maximum occurs at a longer

Nanoscopic Gold Particle Pair Structures
J. Phys. Chem. B, Vol. 103, No. 51, 1999 11405
long-axis resonance band is much higher than in experiment,
and the theoretical λmax values are less than those seen in Figure
9. A similar discrepancy was noted in our earlier work on
oriented gold rods in polyethylene.15 The large apparent short-
axis extinction intensity may arise from spherical Au impurity
particles, as seen in solution-phase surfactant-based syntheses.14
However, we do not observe a significant number of spherical
particles in the TEM images (except those which have been
intentionally synthesized). Also, in the Au rod/polyethylene
study, which employed an identical template synthesis proce-
dure, the polarization spectra did not indicate a large number
of spherical Au particles.32
One interpretation of the shoulder feature seen in the rod-
sphere composite (Figure 9C) is that it arises from a red-shift
of the sphere resonance due to interactions with the nearby rod.
Since this feature is not reproduced in the quasistatic model
spectra (Figure 10C), an alternate interpretation is that it is a
quadrupole mode of the rod component. We should note,
however, that based on previous T-matrix scattering simulations
and experimental data for large aspect ratio gold rods (16 nm
radius) such modes are expected to be weak, at least for isolated
particles.15 Simulations based on the MLWA grossly overesti-
mated the λmax values for both the long- and short-axis
resonances of the rod pair and rod-sphere systems. DDA
calculations on these structures are in progress.
Conclusions
Porous alumina based template synthesis can be used to
prepare paired metal nanoparticle structures. The average
spacing between the members can also be controlled, though
the distribution of spacings is large. In the case of the sphere
pairs, the interparticle spacing has an observable effect on the
measured spectra. When the average surface-to-surface distance
is about three particle diameters, an increase in incidence angle
has no effect on the plasmon resonance peak position. At the
smaller interparticle spacing (corresponding to less than one
particle diameter), an increase in incidence angle causes a red-
shift and broadening of the resonance band.
Rod pair and rod-sphere pair structures can also be achieved
with the template method. The particular structures prepared in
this study show polarization spectra generally similar to those
observed for gold rods oriented in polyethylene. The interparticle
spacing in the rod pair system is apparently large enough to
allow each rod to respond independently to the applied field.
The spectra corresponding to the rod-sphere pair system may
indicate some electromagnetic interaction between the members.
The experimental spectra of the sphere pair systems were
compared with simulations based on simple quasistatic, MLWA,
and DDA models. For the sphere pair dimensions considered
here, the MLWA and DDA calculations yield similar results.
While incidence angle dependent trends were reproduced in all
three models, none of these approaches yielded quantitative
agreement with experiment, either in terms of peak position or
shape. Discrepancies in peak shape may arise in part from the
distribution of interparticle spacings.
Figure 10. Quasistatic limit model spectra for rod-containing com-
posites under p-polarization at incidence angles 0, 20, 30, and 45
degrees. (A) Rods. (B) Rod pairs (d12 ) 70 nm). C. Rod-sphere pairs
(d12 ) 57 nm). Increasing overall extinction corresponds to increasing
incidence angle θ.
parameter κ is less than 2. The screening parameter κ for E
parallel to the long axis is larger than that corresponding to the
short axis, and thus the plasmon resonance condition (the
wavelength at which ꢀm ) -κꢀh) occurs at a longer wavelength.
As the incidence angle increases under p-polarization, the
component of the electric field along the long axis of the
particles also increases, and the long wavelength band becomes
more pronounced.
Figure 10 shows the polarization spectra calculated using eq
6 and the average dimensions and spacings of the three gold
particle/alumina composites obtained from TEM images. The
qualitative trend in band position is the same as seen in
experiment. However, in all cases, the relative intensity of the
Quasistatic limit simulations of the spectra of the rod-
containing composites also agreed only qualitatively with
experiment. The relative intensities of the long axis extinctions
were always larger in theory than in experiment. Subtle spectral
features observed in the rod-sphere pair experimental spectra
could not be reproduced with quasistatic model.
Acknowledgment. This material is based on work supported
by the National Science Foundation under Grant DMR 9625151.

11406 J. Phys. Chem. B, Vol. 103, No. 51, 1999
Sandrock and Foss
(19) Kreibig, U.; Vollmer, M. Optical Properties of Metal Clusters,
Springer-Verlag: Berlin, 1995.
(20) Schmitt, J.; Machtle, P.; Eck, D.; Mohwald, H.; Helm, C. A.
Langmuir 1999, 15, 3256.
(21) Furneaux, R. C.; Rigby, W. R.; Davidson, A. P. Nature 1989, 337,
147.
(22) Shumilova, N. A.; Zutaeva, E. V. In Encyclopedia of Electrochem-
istry of the Elements; Bard, A. J., Ed.; Marcel-Dekker: New York, 1978,
M.L.S also acknowledges the financial support of the ARCS
Foundation. The authors are grateful to Prof. George Schatz
and Lance Kelley for useful discussion and advice regarding
the DDA theory and execution of the DDSCAT Program. The
DDSCAT program was obtained from B. T. Draine and P. J.
Flatau at ftp://astro.princeton.edu/draine/scat/ddscat/ver5a9. Elec-
tron microscopy support was provided by the Lombardi Cancer
Center Microscopy and Imaging Shared Resource (U.S. Public
Health Service Grant 2P30-CA-51008).
8.
(23) The dimensions of the particles were determined from TEM images.
The average length, diameter, spacing distance, and standard deviations
were determined from measurements of thirteen to twenty individual
structures.
References and Notes
(24) Johnson, P. B.; Christy, R. W. Phys. ReV. B 1972, 6, 4370.
(1) Lam, D. M. K.; Rossiter, B. W. Sci. Am. 1991, NoVember, 80.
(2) Chumanov, G.; Sokolov, K.; Gregory, B. W.; Cotton, T. M. J. Phys.
Chem. 1995, 99, 9466.
(3) Moskovits, M. ReV. Mod. Phys. 1985, 57, 783.
(4) Hache, F.; Klein, R. D.; Flytzanis, C. J. Opt. Soc. Am. B 1986, 3,
1647.
(5) Flytzanis C.; Hache, F.; Klein, M. C.; Richard, D.; Roussingnol,
P. In Progress in Optics; Wolf, E., Ed., North-Holland: Amsterdam, 1991,
29, 323.
(6) Sandrock, M. L.; Pibel, C. D.; Geiger, F. M.; Foss, C. A., Jr. J.
Phys. Chem. B. 1999, 103, 2668.
(7) Storhoff, J. J.; Elghanian, R.; Mucic, R. C.; Mirkin, C. A.; Letsinger,
R. L. J. Am. Chem. Soc. 1998, 120, 1959.
(8) Liao, P. F. In Surface Enhanced Raman Scattering; Chang, R. K.,
Furtak, T. E., Eds.; Plenum Press: New York, 1982.
(9) Hulteen, J. C.; van Duyne, R. P. J. Vac. Sci. Technol. 1995, 13,
1553.
(10) Preston, C. K.; Moskovits, M. J. Phys. Chem. 1988, 92, 2957.
(11) Preston, C. K.; Moskovits, M. J. Phys. Chem. 1993, 97, 8405.
(12) Tierney, M. J.; Martin, C. R. J. Phys. Chem. 1989, 93, 2878.
(13) Foss, C. A., Jr.; Hornyak, G. L.; Stockert, J. A.; Martin, C. R. J.
Phys. Chem. 1994, 98, 2963.
(14) Yu, Y.; Chang, S.; Lee, C.; Wang, C. R. C. J. Phys. Chem. B 1997,
101, 6661.
(25) (a) Meier, M.; Wokaun, A. Opt. Lett. 1983, 8, 851. (b) Zeman, E.
J.; Schatz, G. C. In Proceedings of the 17th Jerusalem Symposium; Pullman,
B., Jortner, J., Nitzan, A., Gerber, B., Eds.; Reidel: Dordrecht, Holland,
1984; p 413.
(26) For all MLWA and DDA simulations, Cext or Qext values were
calculated at 4 nm intervals from 350 to 800 nm and plotted as continuous
curves using Excel software.
(27) Granqvist, C. G.; Hunderi, O. Phys. ReV. B 1977, 16, 1353.
(28) Purcell, G. M.; Pennypacker, C. R. Astrophys. J. 1973, 186, 705.
(29) Draine, B. T.; Flatau, P. J. J. Opt. Soc. Am. A 1994, 11, 1491.
(30) Jensen, T.; Kelly, L.; Lazarides, A.; Schatz, G. C. J. Cluster Sci.
1999, 10, 295.
(31) In the computation of extinction spectra using the DDSCAT
program, the selection of an insufficient number of polarizable elements
results in spectral oscillations on the long wavelength side of the plasmon
resonance maximum. For each of the spectra in Figure 8, the number of
polarizable elements was increased until the oscillations disappeared. We
find that the number of polarizable elements required to prevent such
oscillations increases as the interparticle spacing decreases.
(32) In the study reported in ref 15, porous aluminum oxide films
containing Au rods were impregnated with polyethylene and then exposed
to base to dissolve all of the alumina. The Au rods were then mechanically
oriented in the plane of the polyethylene film surface. Normal incidence
plasmon resonance spectra were obtained at various polarization angles
relative to the direction of rod orientation. Significantly, under conditions
of parallel polarization, the spectra of the Au rod systems indicated long
wavelength resonance bands arising from the particles’ long axes. Short
wavelength resonances were either absent or present only at very low
intensity (consistent with either imperfect orientation of the rod-like particles,
or a very small number of spherical Au impurity particles).
(15) Al-Rawashdeh, N. R.; Sandrock, M. L.; Seugling, C. J.; Foss, C.
A., Jr. J. Phys. Chem. B 1998, 102, 361.
(16) El-Kouedi, M.; Sandrock, M. L.; Seugling, C. J.; Foss, C. A., Jr.
Chem Mater. 1998, 10, 3287.
(17) Van de Hulst, H. C. Light Scattering by Small Particles; Dover:
New York, 1981.
(18) Bohren, C. F.; Huffman, D. R. Absorption and Scattering of Light
by Small Particles; Wiley: New York, 1983.