Gold nanorod extinction spectra

Cite as: J. Appl. Phys. 99, 123504 (2006); https://doi.org/10.1063/1.2203212

Submitted: 07 November 2005 . Accepted: 10 April 2006 . Published Online: 16 June 2006

Stuart W. Prescott, and Paul Mulvaney

ARTICLES YOU MAY BE INTERESTED IN

Erratum: “Gold nanorod extinction spectra” [J. Appl. Phys. 99, 123504 (2006)]
Journal of Applied Physics 103, 119901 (2008); https://doi.org/10.1063/1.2936980

Determination of the aspect ratio statistical distribution of gold nanorods in solution from a
theoretical fit of the observed inhomogeneously broadened longitudinal plasmon resonance
absorption spectrum
Journal of Applied Physics 100, 044324 (2006); https://doi.org/10.1063/1.2244520

Discrete dipole approximation simulations of gold nanorod optical properties: Choice of input
parameters and comparison with experiment
Journal of Applied Physics 105, 102032 (2009); https://doi.org/10.1063/1.3116139

J. Appl. Phys. 99, 123504 (2006); https://doi.org/10.1063/1.2203212 99, 123504

© 2006 American Institute of Physics.

 JOURNAL OF APPLIED PHYSICS 99, 123504 共2006兲

Gold nanorod extinction spectra

Stuart W. Prescott and Paul Mulvaneya兲
Particulate Fluids Processing Centre, School of Chemistry, University of Melbourne, VIC 3010 Australia
共Received 7 November 2005; accepted 10 April 2006; published online 16 June 2006兲
Geometrical factors suitable for use in Mie–Gans theory are derived from discrete dipole
aproximation generated spectra for a number of pseudonanorods permitting the rapid calculation of
extinction spectra. It is shown that the rod width, rod end-cap geometry 共flat, oblate spheroid, and
sphere兲 and the rod size distribution all have a significant effect on the position of the peak
absorbance. Moreover, it is shown that spectrometric characterization of nanorods is possible given
an independent measure of rod width and suitable assumptions regarding the end-cap geometry and
the form of the rod-length distribution; under such conditions the full width half maximum and the
extinction peak uniquely determine the average rod length and the breadth of the distribution.
© 2006 American Institute of Physics. 关DOI: 10.1063/1.2203212兴

I. INTRODUCTION Understanding the scattering spectra of these nanopar-

ticles is nontrivial; however, it is an essential step in the
Recent experiments have demonstrated the synthesis of development of optical applications for these materials. Ini-
nonspherical gold and silver nanoparticles. Particle mor- tial modeling efforts have sought to explain the various fea-
phologies that have been synthesized include hemispheri-
tures of the extinction spectra of gold nanorods through Mie–
cally capped cylinders 共nanorods兲,1–3 cubes,4,5 triangular
Gans theory,9 discrete dipole approximation 共DDA兲
prisms,4,6 tetrahedra,5 and hexagonal prisms.4 The techniques
modeling,11 and surface integral modeling. These models
utilized for these syntheses include various electrochemical
have had limited success in replicating the experimentally
reactions, the use of particular surfactant systems, and the
determined spectra with significant features such as ␭max of
addition of trace amounts of other metals. The unusual opti-
SPlong not accurately predicted.
cal properties of some nanoparticles 共e.g., gold nanopar-
Recently, Brioude et al.11 concluded that the Mie–Gans
ticles兲 are well known, while the more recently developed
range of particle morphologies have correspondingly differ- ellipsoid model was able to adequately account for the vari-
ent scattering spectra. ous peaks in the extinction spectra of gold nanorods. Indeed,
These structures have interesting optical properties, with DDA simulations permitted Brioude et al.11 to assign the
the scattering behavior in the UV-visible-near infrared 共NIR兲 peaks in the spectra to the transverse and longitudinal bands
region depending on the orientation of the nanoparticle with and also led the authors to conclude that a multipolar reso-
respect to the incident light, as well as the polarization of the nance was possible in these rods. However, Brioude et al.11
incident light;5,7 for example, aligned gold nanorods illumi- also claimed that the Mie–Gans ellipsoid model quantita-
nated with polarized white light appear either red or blue tively predicted the experimentally determined peak in the
depending on the direction of polarization.7 Here, approaches absorption spectra, ␭max, a claim that is not supported by the
to simulating the UV-visible-NIR scattering spectra of gold published experimental data.9 Recent experimental1,2,9 and
nanorods will be discussed 共in particular, for the longitudinal theoretical11,12 works have also shown that ␭max varies for
plasmon band, SPlong兲, with a view to understanding the re- constant rod width and varying rod length and also for con-
lationships between particle shape and experimental spectra. stant rod length and varying rod width. With analogy to Mie–
While the UV-vis-NIR spectrum of a gold sol is rela- Gans theory, it has previously been concluded that the aspect
tively featureless, significant information can be determined ratio alone determines the position of ␭max for gold
from it. For example, the breadth of the particle size distri- nanorods.11
bution of a set of gold spheres may be estimated from the Here, we present detailed simulation results for SPlong
position of ␭max along with the width of the extinction peak. that challenge this analysis, quantitatively considering the
Similarly, for a sample of identical prolate spheroids, the effect of rod width and aspect ratio on the spectra, showing
aspect ratio of the particles 共i.e., the length of the semimajor that aspect ratio alone cannot be used to predict the experi-
axis divided by the length of the semiminor axis兲 is uniquely mentally determined ␭max. Additionally, it is shown that the
determined by considering ␭max alone; within the confines of morphology of the particle itself is important; it is demon-
Mie–Gans theory, the spectrum is a single parameter func- strated here that nanorods cannot be approximated by ellip-
tion depending only on the particle aspect ratio and not on soids and that the end-cap geometry of the nanorod is
the particle width and length independently.8,9 For this rea- important.
son, it has been common to treat nanorods as spheroids, re- In some experimental reports of the the synthesis of gold
gardless of the obvious differences in shape.9,10 nanorods, the rod length distribution, as calculated by pains-
taking transmission electron microscopy 共TEM兲 measure-
FAX: ⫹61 3 9347 5180; electronic mail: [email protected]
a兲
ments, has been presented 共e.g., Yu et al.1兲. In such cases, it

0021-8979/2006/99共12兲/123504/7/$23.00 99, 123504-1 © 2006 American Institute of Physics

123504-2 S. W. Prescott and P. Mulvaney J. Appl. Phys. 99, 123504 共2006兲

is seen that the rod length distribution is reasonably broad, 2␲V␧3/2

m ␧⬙
with the standard deviation being typically between 10% and Qext共␭兲 = , 共1兲
GL ␭ 兵␧⬘ + 关共1 − L兲/L兴␧m其2 + 共␧⬙兲2
2
25% of the mean value. A distribution of rod sizes in a
sample would be expected to change the extinction spectrum where G is the particle cross-sectional area projected onto a
compared with a monodisperse sample with the same plane that is perpendicular to the incident radiation 共i.e., the
number-average dimensions. If the extinction efficiency of area of the “shadow” that would be naïvely cast by the par-
different sized particles were the same, then the effect would ticle given its orientation兲.
only be to broaden the measured extinction peak without For nonspheroidal particles such as the nanorods under
shifting ␭max. However, since longer rods have a greater ex- consideration here, a spectrum Qextfit
共L ; ␭兲 may be fitted to the
tinction cross section 共see, e.g., Brioude et al.11兲, a distribu- DDA-generated Qext共␭兲 points using a simplex algorithm
tion of rod lengths would be expected to shift ␭max towards varying L. Due to the way DDSCAT calculates Qext, it is also
longer wavelengths. necessary to include a constant of proportionality, Geff, in
It is shown here that a paradigm shift in the modeling this fitting procedure. By calculating the extinction spectra of
approach is required to accurately predict features such as various aspect ratio rods by DDA, then fitting appropriate L
␭max. Of particular importance in comparing the experimen- and Geff to each of these spectra, an empirical functional
tally determined extinction spectra with those obtained from form for L共r , R兲 and Geff共r , R兲 maybe determined. Indeed, a
modeling efforts is the degree to which the pseudonanorods quadratic polynomial in R is found to describe 共1 − L兲 / L to
used in the model reflect those synthesized in the laboratory. good accuracy and a cubic polynomial may be used to de-
Indeed, the exact shape of the particle and the fact that the scribe Geff. Values for L and 共1 − L兲 / L are tabulated in the
particle is a member of an ensemble with a particle size Appendix for various particle shapes and sizes. It should be
distribution are shown to have significant effects on the noted that full width at half maximum 共FWHM兲 calculated
spectra. by Mie–Gans theory using these values for L tends to be
somewhat less than that produced by the DDA simulations
for particles with r ⲏ 10 nm.
II. THEORETICAL APPROACH

DDA simulations have been described elsewhere in sig- B. Effects of particle shape
nificant detail,11,13,14 so here we will only provide the briefest
of introductions. The first requirement of these DDA simula- The exact shape of the gold nanorods being produced is
tions is to represent the continuum scattering target as a fi- not as yet well defined. Various TEM studies seem to indi-
nite, cubic array of polarizable points. These points are then cate that a spherically capped cylinder is a reasonable ap-
able to interact with each other and the incident electromag- proximation to the correct geometry; however, analysis of
netic radiation 共EMR兲, with the DDA algorithm seeking a these TEM images also indicates that many particles have
self-consistent solution for the dipole polarizations and the end caps that are better approximated by an oblate spheroid.
local electric field. Constructing pseudonanorods once more using a cylinder as
In this work, the DDA package DDSCAT 6.0 was used, the center of the particle but this time with two oblate sphe-
provided by Draine and Flatau.15 To facilitate comparison roids as the end caps, one has two limiting cases for the
with experimental data, all spectra reported here are for well- particle shape, as shown in Fig. 1 共note that the aspect ratio
dispersed particles in water, with in vacuo wavelengths. The of the particle, R, is still the total length of the rod divided by
dielectric data for gold were taken from the experimental the total width of the rod兲. The effect of the end-cap shape
data of Weaver et al.16 using standard methods to account for may be studied by varying the eccentricity e of the oblate
the surface damping in small particles.14 spheroid 共for the dimensions shown in Fig. 1, e2 = 1 − b2 / a2兲
within the limits of the spherical end caps 共with e = 0兲 and no
end caps 共i.e., just a cylinder, with e → 1兲.
A. Describing spectra with geometrical factors
Using Mie–Gans theory, for an ellipsoid with the electric C. Modeling particle size distributions
field of the incident radiation parallel to one of its principle
axes, the electrostatic approximation permits the extinction Spectra for samples with a distribution of particle sizes
efficiency spectrum Qext共␭兲 to be written in terms of a geo- may be synthesized by discretizing the rod size distribution
metrical factor L.8 While for prolate spheroids, oblate sphe- into a number of bins and adding an appropriately weighted
roids, and spheres analytic, expressions for L have been de- spectrum for each of these bins. For a particle size distribu-
rived, this is not the case for most particle geometries. The tion Ni normalized such that 兺iNi = 1, one has
focus of this work is SPlong, hence Lz is abbreviated as L 1
throughout; the geometrical factors for the minor axes may Qext共␭兲 = 兺 NiQext
具G典 i
i
共␭兲Gi , 共2兲
be trivially calculated using the relationship Lx + Ly + Lz = 1
with Lx = Ly by symmetry.8 i
where Qext are the extinction efficiencies for particles falling
For a generalized particle of volume V with an in vacuo into bin i, Gi are the cross-sectional areas of the particles
dielectric function of ␧共␭兲 = ␧⬘共␭兲 + i␧⬙共␭兲, that is dispersed in projected onto a plane perpendicular to the direction of
a nonabsorbing medium with dielectric ␧m, one obtains the propagation, and 具G典 is the cross-sectional area of the aver-
following expression for the extinction efficiency Qext共␭兲: age particle.
123504-3 S. W. Prescott and P. Mulvaney J. Appl. Phys. 99, 123504 共2006兲

FIG. 1. Schematics of the different particle morphologies under consider-

ation including a prolate spheroid, a spherically capped cylinder, an ellip-
soidally capped cylinder, and a cylinder. Each particle has the same aspect
ratio R.

In synthesizing spectra for the longitudinal plasmon

band of the gold nanorods, the particle size distribution is for
different rod lengths with a given rod width; for rods of
radius r, and aspect ratio Ri, and Gi = ␲r2 + 4r2共Ri − 1兲. With
appropriate values for L and Geff calculated from DDA simu-
lations as described above, it is straightforward to generate a
spectrum for the particle size distribution using
Eq. 共2兲.

III. NUMERICAL EXPERIMENTS

A. Geometrical factors
DDA simulations were performed to calculate the ex-
tinction spectra of the longitudinal plasmon band of spheri-
cally capped cylinders with rod radii r 苸 兵5 , 10, 20, 40其 nm
and aspect ratios R 苸 关1 , 8兴. The peak position, ␭max, of the
plasmon response was calculated from the DDA spectra, FIG. 2. 共Color online兲 DDA-derived parameters for spherically capped cyl-
along with the geometrical factor L, that describes this shape, inders 共spherocylinders兲 with various rod widths and aspect ratios: 共a兲 ex-
tinction spectra, 共b兲 extinction maxima, and 共c兲 Mie–Gans geometrical fac-
as shown in Fig. 2. Data for the prolate spheroid of the same
tors for SPlong.
aspect ratio are shown in Fig. 2 for comparison. All simula-
tions were performed with a cubic lattice of around 1 nm
giving a total of between 6000 and 615 000 dipoles per tar- ity, R 苸 兵0 , 0.5, 0.8, 0.95其 as well as for a plain cylinder. The
get; under these simulation conditions, both Mie theory and influence of the shape of the end-cap shape on L is shown in
the DDA simulations for spherical particles of various sizes Fig. 3.
gave the same values for ␭max and FWHM.
C. Rod length distribution
For gold nanorods that are 10 nm in radius and have
B. Rod end caps
共number average兲 aspect ratios R 苸 兵2 , 4 , 6 , 8其, the effect of
As before, extinction spectra were calculated for the lon- the breadth of the rod-length distribution was studied. For
gitudinal plasmon band of pseudonanorods with varying each average aspect ratio, 具R典, Qext共␭兲 was calculated for a
shaped end caps as shown in Fig. 1. Spectra were generated Gaussian distribution about R with a range of standard de-
for particles with r = 5 nm, R 苸 关1 , 8兴, and end-cap eccentric- viations; the standard deviation is expressed as a percentage
123504-4 S. W. Prescott and P. Mulvaney J. Appl. Phys. 99, 123504 共2006兲

FIG. 3. 共Color online兲 DDA-derived parameters for different shapes of par-

ticle end cap including the limiting cases of the plain cylinder and the
spherically capped cylinder 共spherocylinder兲 along with cylinders capped
with oblate spheroids of varying eccentricity e: 共a兲 extinction maximum
␭max and 共b兲 Mie–Gans geometrical factors L.

of R; spectra are shown in Fig. 4, along with the peak posi-

tion, ␭max, and FWHM of these peaks. For a given relative
standard deviation, the redshift increases approximately lin-
early for longer rods, with a shift of 12 nm from a standard
deviation of 15% for the R = 6 rods.

IV. DISCUSSION
In Fig. 5, we compare the DDA calculations with recent FIG. 4. 共Color online兲 Predicted extinction characteristics for various
experimental data from Pérez-Juste et al.7 on the SPlong band Gaussian distributions of rods with a given number-average aspect ratio and
of gold rods grown in aqueous solution using cetyl trimethyl varying standard deviation, showing the redshift and broadening of the ex-
ammonium bromide 共CTAB兲 as a template. For these rods, tinction peak of SPlong due to broadening the rod-length distribution: 共a兲
extinction spectra, 共b兲 extinction maximum, and 共c兲 full width at half
the length and aspect ratio had been determined by analysis maximum.
of several hundred particles by TEM, and the standard de-
viation in both the length and width were known. The SPlong
peak wavelength positions were measured with a precision pure cylinders despite the fact that the electron microscopy
of ±1 nm. images suggest that a spherical cap is present. The actual
Figure 5 shows the results for prolate spheroids, spheri- cylinder length for such calculations is best approximated
cally capped cylinders, and long 共flat兲 cylinders in a medium using the length determined by TEM, without taking into
with refractive index of 1.33. Note that even for long rods account any cap curvature. As shown in Fig. 4, this counter-
with aspect ratios greater than 5, the end cap geometry is intuitive result is coincidental with the effects of the rod-
critical in determining the SPlong position. The results pre- length distribution.
sented here allow us to conclude that the SPlong band for Previous work has suggested two different crystallo-
these small gold rods are not well modeled as ellipsoids. The graphic morphologies for the rods. In one case the longitu-
best fit with experiment is obtained when they are treated as dinal growth direction of the rod was identified as 具112典 and
123504-5 S. W. Prescott and P. Mulvaney J. Appl. Phys. 99, 123504 共2006兲

before the aspect ratio is greater than 2. The optical proper-

ties of the smallest, growing spherical nuclei will therefore
lie between the two curves for the “Mie–Gans spheroid” and
“long cylinder” shown in Fig. 5.
Finally, we have presented here clear evidence for the
importance of including polydispersity in the calculated
spectra. Peak plasmon energies are affected considerably by
the effects of polydispersity. However, if the shape can be
identified unambiguously, and hence the correct geometrical
factors are known a priori, the width and peak extinction can
be used to model the polydispersity, at least for known par-
ticle length distributions such as the small Gaussian distribu-
tions shown in Fig. 4.

FIG. 5. 共Color online兲 Comparison of the position of the extinction maxi-

mum of SPlong as predicted by DDA simulations for various particle shapes V. CONCLUSIONS
with experimental results of Pérez-Juste et al. 共Ref. 7兲 for 22 nm diameter
nanoparticles. The use of DDA modeling of the extinction spectra of
gold nanorods has demonstrated that the effect of a distribu-
the rod had an overall pentagonal cross section.17 In the other tion in nanorod lengths leads to a redshift in ␭max; moreover,
for realistic distribution widths, the shift is of similar magni-
TEM study, the sidewalls were proposed to be the more ob-
tude to the difference between the experimentally determined
vious 兵100其 and 兵110其 facets while the end caps were gener-
and DDA-simulated values for ␭max. Furthermore, the aspect
ated from 兵111其 and 兵110其 facets along with 兵100其 for the end
ratio alone does not uniquely determine the extinction spec-
face.18 This generates a more or less octagonal cross section. trum of a nanorod even for a monodisperse sample of rods.
It seems the end-cap geometry is important for SPlong reso- In contrast to the case of an ellipsoid, the position of ␭max is
nances 共as shown in Fig. 3兲, while the cross section plays a also a strong function of the rod width with the direct con-
less significant role 共as evidenced by various experimental sequence that trying to “fix” the ellipsoid model so that it
data for SPshort, which is in good agreement with Mie–Gans agrees with experiment 共e.g., Link et al.10 and Yan et al.19兲 is
theory for a spheroidal particle兲. a futile exercise. We demonstrate that a fresh approach to
The data here indicate that the SPlong mode is extremely modeling the extinction of nanorods is required to obtain
sensitive to the end-cap geometry, a factor not recognized in quantitative agreement with experiment. It is further noted
initial work on rod synthesis. For example, Link et al. and that, with an independent measure of the rod width 共from
Yan et al. proposed that the discrepancy was due to a surface TEM or AFM兲, the average rod length and the the width of
layer with a modified refractive index due to the presence of the rod length distribution may be uniquely determined from
either ordered water or surfactant molecules.10,19 Brioude the position of ␭max and the width of the extinction peak with
et al.11 actually considered the observed shifts to be consis- only allowable assumptions regarding the shape of the rod
tent with the ellipsoid model. While it is true that the depen- size distribution.
dence of ␭max on the aspect ratio is similar, there is a drastic
difference of up to 100 nm in the absolute peak position,
depending on the cap geometry. ACKNOWLEDGMENTS
The results indicate that for more complex geometries
such as prisms and cubes, small truncations may well play a The Particulate Fluids Processing Centre was established
and supported under the Australian Research Centre’s Re-
determining factor in the peak position of the plasmon reso-
search Centres Programme. Support of the Victorian Institute
nances. These results lead to the simple conclusion that it is
for Chemical Sciences High Performance Computing Facil-
very difficult to identify the morphology of a small metal
ity is gratefully acknowledged. The DDSCAT program was
particle directly from the extinction spectrum, without inde-
developed by Professor Bruce T. Draine 共Princeton Univer-
pendent structural information gleaned from TEM or atomic sity Observatory兲 and Dr. Piotr J. Flatau 共University of Cali-
force microscopy 共AFM兲. fornia兲.
In the limit of very short rods and cylinders, the DDA
calculations give two different limiting plasmon band posi-
tions. The ellipsoid converges on a sphere as the eccentricity
APPENDIX
tends to 1, with a plasmon peak wavelength identical to that
for a small sphere. The cylinder converges to a slightly Tabulated here are the geometrical factors L for various
longer wavelength. Experimentally, the core seed particles nanoparticle morphologies. As described in the main text,
are spheres, and hence belong on the “ellipsoid” dispersion these geometrical factors may be used in Mie–Gans theory to
line initially. As they grow into spheres, the dispersion curve calculate Qext共␭兲 using Eq. 共1兲. Since the actual calculation
will move up onto the “flat cylinder” line, and it does this of Eq. 共1兲 relies on 共1 − L兲 / L rather than L, this value is also
123504-6 S. W. Prescott and P. Mulvaney J. Appl. Phys. 99, 123504 共2006兲

TABLE I. Geometrical factors for 5 nm radius cylinders with various end geometries.

Sphere e = 0.5 e = 0.8 e = 0.95 Cylinder

R L 共1 − L兲 / L L 共1 − L兲 / L L 共1 − L兲 / L L 共1 − L兲 / L L 共1 − L兲 / L

1 0.311 2.2 0.307 2.3 0.295 2.4 0.268 2.7 0.222 3.5
2 0.163 5.1 0.159 5.3 0.151 5.6 0.140 6.1 0.126 6.9
3 0.099 9.1 0.098 9.2 0.094 9.6 0.089 10.2 0.083 11.1
4 0.068 13.6 0.067 13.8 0.065 14.3 0.063 15.0 0.059 15.9
5 0.050 18.9 0.050 19.1 0.048 19.6 0.047 20.4 0.045 21.4
6 0.039 24.8 0.038 25.1 0.038 25.7 0.036 26.5 0.035 27.6
7 0.031 31.4 0.031 31.7 0.030 32.4 0.029 33.3 0.028 34.5
8 0.025 38.7 0.025 39.1 0.025 39.8 0.024 40.8 0.023 42.1

TABLE II. Summary of quadratic parameters to fit geometrical factors for 5 nm radius capped cylinders: 共1
− L兲 / L = aR2 + bR + c.

Type a b c

Sphere 0.3504 2.0809 −0.3035

e = 0.5 0.3487 2.1350 −0.2918
e = 0.8 0.3418 2.2680 −0.2491
e = 0.95 0.3388 2.3795 0.0267
Cylinder 0.3455 2.3994 0.7611

TABLE III. Geometrical factors for 10 nm radius cylinders with various end geometries.

Sphere e = 0.5 e = 0.8 e = 0.95 Cylinder

R L 共1 − L兲 / L L 共1 − L兲 / L L 共1 − L兲 / L L 共1 − L兲 / L L 共1 − L兲 / L

1 0.310 2.2 0.303 2.3 0.291 2.4 0.262 2.8 0.221 3.5
2 0.156 5.4 0.151 5.6 0.143 6.0 0.132 6.5 0.119 7.4
3 0.092 9.8 0.091 10.0 0.087 10.5 0.083 11.1 0.077 11.9
4 0.062 15.0 0.061 15.3 0.059 15.8 0.057 16.6 0.054 17.6
5 0.045 21.1 0.045 21.4 0.043 22.0 0.042 22.8 0.040 23.9
6 0.034 28.1 0.034 28.4 0.033 29.1 0.032 30.0 0.031 31.4
7 0.027 35.9 0.027 36.3 0.026 37.1 0.026 38.1 0.025 39.6
8 0.022 44.8 0.022 45.2 0.021 46.0 0.021 47.1 0.020 48.7

TABLE IV. Summary of quadratic parameters to fit geometrical factors for 10 nm radius capped cylinders:
共1 − L兲 / L = aR2 + bR + c.

Type a b c

Sphere 0.4532 2.0100 −0.3072

e = 0.5 0.4495 2.0856 −0.2903
e = 0.8 0.4420 2.2417 −0.2489
e = 0.95 0.4378 2.3786 0.0144
Cylinder 0.4473 2.4262 0.6937

TABLE V. Geometrical factors for 15 nm radius cylinders with various end geometries.

Sphere e = 0.5 e = 0.8 e = 0.95 Cylinder

R L 共1 − L兲 / L L 共1 − L兲 / L L 共1 − L兲 / L L 共1 − L兲 / L L 共1 − L兲 / L

1 0.303 2.3 0.295 2.4 0.282 2.5 0.253 2.9 0.303 2.3
2 0.144 6.0 0.140 6.1 0.132 6.6 0.122 7.2 0.144 6.0
3 0.083 11.1 0.081 11.3 0.078 11.9 0.073 12.6 0.083 11.1
4 0.055 17.3 0.054 17.7 0.052 18.4 0.049 19.3 0.055 17.3
5 0.039 24.9 0.038 25.3 0.037 26.0 0.036 27.1 0.039 24.9
6 0.029 33.7 0.029 34.1 0.028 35.0 0.027 36.3 0.029 33.7
7 0.022 43.8 0.022 44.3 0.022 45.2 0.021 46.6 0.022 43.8
8 0.018 55.1 0.018 55.8 0.017 56.8 0.017 58.2 0.018 55.1
123504-7 S. W. Prescott and P. Mulvaney J. Appl. Phys. 99, 123504 共2006兲

TABLE VI. Summary of quadratic parameters to fit geometrical factors for 15 nm radius capped cylinders:
共1 − L兲 / L = aR2 + bR + c.

Type a b c

Sphere 0.6337 1.8484 −0.2111

e = 0.5 0.6365 1.8929 −0.1371
e = 0.8 0.6230 2.1265 −0.1679
e = 0.95 0.6136 2.3586 0.0120
Cylinder 0.6337 1.8484 −0.2111

TABLE VII. Geometrical factors for 20 nm radius cylinders with various end geometries.

Sphere e = 0.5 e = 0.8 e = 0.95 Cylinder

R L 共1 − L兲 / L L 共1 − L兲 / L L 共1 − L兲 / L L 共1 − L兲 / L L 共1 − L兲 / L

1 0.292 2.4 0.283 2.5 0.269 2.7 0.241 3.1 0.197 4.1
2 0.131 6.6 0.126 6.9 0.119 7.4 0.110 8.1 0.097 9.3
3 0.072 12.9 0.070 13.3 0.067 14.0 0.062 15.0 0.058 16.1
4 0.046 20.7 0.045 21.2 0.043 22.1 0.041 23.2 0.039 24.4
5 0.032 30.3 0.031 30.9 0.030 31.9 0.029 33.2 0.027 35.8
6 0.023 41.8 0.023 42.6 0.022 43.9 0.022 45.1 0.021 47.4
7 0.018 54.8 0.018 55.2 0.017 57.2 0.017 58.7 0.016 61.2
8 0.014 70.2 0.014 71.2 0.014 72.9 0.013 74.6 0.013 76.8

TABLE VIII. Summary of quadratic parameters to fit geometrical factors for 20 nm radius capped cylinders:
共1 − L兲 / L = aR2 + bR + c.

Type a b c

Sphere 0.8999 1.5674 −0.0328

e = 0.5 0.8963 1.6949 −0.0032
e = 0.8 0.8987 1.9147 −0.0147
e = 0.95 0.8881 2.1617 0.2410
Cylinder 0.8632 2.6373 0.5082

tabulated here. An approximate functional form for 共1 Small Particles 共Wiley, New York, 1983兲.
9
J. Pérez-Juste, I. Pastoriza-Santos, L. M. Liz-Marzán, and P. Mulvaney,
− L兲 / L is also provided for convenience; as this is a purely
Coord. Chem. Rev. 249, 1870 共2005b兲.
empirical relationship, physical meaning should not be read 10
S. Link, M. Mohamed, and M. El-Sayed, J. Phys. Chem. B 103, 3073
into this function. Data are provided for cylinders with 共1999兲.
11
spherical and ellipsoidal square end caps along with cylin- A. Brioude, X. C. Jiang, and M. P. Pileni, J. Phys. Chem. B 109, 13138
ders for rods with r = 5 nm 共Tables I and II兲, r = 10 nm 共2005兲.
12
N. G. Khlebtsov, A. G. Melnikov, V. A. Bogatyrev, L. A. Dykman, A. V.
共Tables III and IV兲, r = 15 nm 共Tables V and VI兲, and r Alekseeva, L. A. Trachuk, and B. N. Khlebtsov, J. Phys. Chem. B 109,
= 20 nm 共Tables VII and VIII兲. 13578 共2005兲.
13
B. T. Draine and P. J. Flatau, J. Opt. Soc. Am. A Opt. Image Sci. Vis 11,
1
Y. Y. Yu, S. S. Chang, C. L. Lee, and C. R. C. Wang, J. Phys. Chem. B 1491 共1994兲.
14
101, 6661 共1997兲. N. Felidj, J. Aubard, and G. Levi, J. Chem. Phys. 111, 1195 共1999兲.
2 15
J. Pérez-Juste, L. M. Liz-Marzán, S. Carnie, D. Y. C. Chan, and P. Mulv- B. T. Draine and P. J. Flatau, DDSCAT 6.0, available online, 2003 http://
aney, Adv. Funct. Mater. 14, 571 共2004兲. www.astro.princeton.edu/⬃draine/
3 16
I. Lisiecki and M. P. Pileni, Langmuir 19, 9486 共2003兲. J. Weaver, C. Krafka, D. Lynch, and E. Koch, Physics Data: Optical
4
T. K. Sau and C. J. Murphy, J. Am. Chem. Soc. 126, 8648 共2004兲. Properties of Metals, Part 2: Noble Metals, Aluminium, Scandium, Yt-
5
M. A. El-Sayed, Acc. Chem. Res. 34, 257 共2001兲. trium, the Lanthanides and the Actinides 共Fach-Informations Zentrum,
6
I. Pastoriza-Santos, D. Gomez, J. Pérez-Juste, L. M. Liz-Marzán, and P. Karlsruhe, 1981兲.
17
Mulvaney, Phys. Chem. Chem. Phys. 6, 5056 共2004兲. P. Gai and M. Harmer, Nano Lett. 2, 771 共2002兲.
7 18
J. Pérez-Juste, B. Rodriguez-Gonzalez, P. Mulvaney, and L. M. Liz- Z. L. Wang, M. B. Mohamed, S. Link, and M. A. El-Sayed, Surf. Sci. 440,
Marzán, Adv. Funct. Mater. 15, 1065 共2005兲. L809 共1999兲.
8 19
C. F. Bohren and D. R. Huffman, Absorption and Scattering of Light by B. Yan, Y. Yang, and Y. Wang, J. Phys. Chem. B 107, 9159 共2003兲.