Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 115 (2013) 588–594

Contents lists available at SciVerse ScienceDirect

Spectrochimica Acta Part A: Molecular and

Biomolecular Spectroscopy
journal homepage: www.elsevier.com/locate/saa

Useful multivariate kinetic analysis: Size determination based on

cystein-induced aggregation of gold nanoparticles
Faride Rabbani a, Mohammad Reza Hormozi Nezhad b,c,⇑, Hamid Abdollahi a,⇑
a
Faculty of Chemistry, Institute for Advanced Studies in Basic Sciences (IASBS), 45195-1159 Zanjan, Iran
b
Department of Chemistry, Sharif University of Technology, Tehran 11155-9516, Iran
c
Institute for Nanoscience and Nanotechnology, Sharif University of Technology, Tehran, Iran

h i g h l i g h t s g r a p h i c a l a b s t r a c t

 Size effect of Au NP on aggregation

was used to Au NP size
determination.
 A multivariate calibration model was
constructed for Au NP size
determination.
 The average sizes of unknown
samples were determine using this
method.
 We showed the direct size and
concentration effects of Au NP on
aggregation.
 The pH effect on aggregation is more
significant for lower Au NP
concentration.

a r t i c l e i n f o a b s t r a c t

Article history: This study describes spectrometric monitored kinetic processes to determine the size of citrate-capped
Received 2 March 2013 Au nanoparticles (Au NPs) based on aggregation induced by L-cysteine (L-Cys) as a molecular linker.
Received in revised form 11 June 2013 The Au NPs association process is thoroughly dependent on pH, concentration and size of nanoparticles.
Accepted 24 June 2013
Size dependency of aggregation inspirits to determine the average diameters of Au NPs. For this aim the
Available online 2 July 2013
procedure is achieved in aqueous medium at pH 7 (phosphate buffer), and multivariate data including
kinetic spectra of Au NPs are collected during aggregation process. Subsequently partial least squares
Keywords:
(PLS) modeling is carried out analyzing the obtained data. The model is built on the basis of relation
Gold nanoparticles
Size determination
between the kinetics behavior of aggregation and different Au NPs sizes. Training the model was per-
Multivariate analysis formed using latent variables (LVs) of the original data. The analytical performance of the model was
Aggregation process characterized by relative standard error. The proposed method was applied to determination of size in
unknown samples. The predicted sizes of unknown samples that obtained by the introduced method
are interestingly in agreement with the sizes measured by Transmission Electron Microscopy (TEM)
images and Dynamic Light Scattering (DLS) measurement.
Ó 2013 Elsevier B.V. All rights reserved.

Introduction

Au nanoparticles (Au NPs) with high stability among the other

⇑ Corresponding authors. Address: Department of Chemistry, Sharif University of
Technology, Tehran 11155-9516, Iran (M.R.H. Nezhad).
E-mail addresses: hormozi@sharif.edu (M.R. Hormozi Nezhad), abd@iasbs.ac.ir
(H. Abdollahi).
metal nanoparticles have great interest due to their using in many
fields such as sensors, biosensors, medicine, catalysis and many
emerging areas of nanotechnology [1–3]. Recent interest in nano-
particles stems from the fact that, nano-sized materials exhibit

1386-1425/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.saa.2013.06.090
 F. Rabbani et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 115 (2013) 588–594 589

unique optical, electronic and magnetic properties which depend chemometrics methods is partial least squares (PLS) modeling that
on their shape, size, and local environment. The foundation of relating two data matrices, X (dedependent variables) and Y
nanotechnology research is based on the size, distribution and (inpendent variables), by a linear multivariate model [33]. So in
shape of nanoparticles because these parameters are the most this experiment, the PLS modeling was used for the multivariate
important characteristics of the systems [4]. Usually, the standard calibration of the UV–Vis data in order to relate the Au NPs sizes
methods for determination of the metallic nanoparticles size are to aggregation kinetic process.
transmission electron microscopy (TEM), for nanoparticles larger
than 2 nm [5,6] and mass spectrometry for clusters smaller than
Experimental
1.5 nm [7,8]. Both require sophisticated and highly expensive
instrumentation. Therefore, alternative methods for the accurate
Reagents and solutions
determination of the NPs size are essential. When nanoparticles
made of silver and gold dispersed in liquid media, these nanopar-
Hydrogen tetrachloroaurate (HAuCl4, 99%), Sodium citrate (99%)
ticles exhibit a strong UV–Vis extinction band that is not present in
and L-cysteine (97%), were purchased from Sigma–Aldrich. Diso-
the spectrum of the bulk metal. This extinction band results when
dium hydrogen phosphate and potassium dihydrogen phosphate,
the incident photon frequency is resonant with the collective exci-
which used to prepare phosphate buffer (pH 5, 7) were purchased
tation of the conduction electrons, and is known as the surface
from Merck. Water was purified with a Milli-Q water system.
plasmon resonance (SPR) [9,10]. The resonance frequency and
intensity of this SPR are sensitive to the size, shape, dielectric prop-
erties and local environment of the nanoparticles [11–15]. As the Instrumentation and software
size of Au NPs increases, the color of the solution varies from red
to pink, and the SPR of the Au NPs undergoes a red shift with The absorption spectrum of each solution was recorded with a
increasing size as predicted by Mie theory [16]. UV–Vis s-2100 spectrophotometer (Scinco). Transmission electron
Controlled aggregation of Au NPs leads to a change in their opti- microscopy (TEM) was carried out on a Hitachi H-9000NAR trans-
cal properties (i.e. a red shift in surface plasmon band) which is a mission electron microscope, operating at 200 kV. Dynamic light
topic of considerable scientific interest in nanotechnology and scattering (DLS) was performed on a standard laser light scattering
materials science, as it can be exploited for various applications spectrometer (BI-200SM) equipped with aBI-9000 AT to determine
[17,18]. Different studies showed that the change in the localized the hydrodynamic diameter for one set of nanoparticles in solution.
surface plasmon band is related to the aggregated size of Au NPs Each original kinetic data was arranged into a matrix format Y
[19–21]. Typically the Au NPs can be readily modified with thiol- (I  J), where I is the number of rows (measured spectra over
containing biomolecules because the thiol group exhibits a strong times) and the J is the number of columns (measured wave-
interaction on the metal surface [22]. Generally, modified Au NPs lengths). The data were analyzed using MATLAB software, version
aggregate when the terminal amino group of a biomolecule forms 6.5 (The MathWorks), with ‘PLS Toolbox’, version 2.0.
hydrogen bonding with the carboxyl group of another biomole-
cules on an adjacent nanoparticle [23,24]. Further studies have Measurements
been carried out on the effect of nanoparticles size on biomolecules
induced aggregation and to understand the kinetics of the aggrega- Absorption Spectra were collected over the range of 400–
tion process [25–31]. 950 nm. Briefly, Au NP solution and buffer solution were quantita-
In this paper Au NPs with different sizes were prepared to tively mixed and was allowed to react for 10 min, afterward, a
investigate their aggregation kinetic depending on their average quantitative amount of L-Cys was added, and the reaction was
size. Citrate-capped Au NPs prepared by Frens method have been monitored via UV–Vis. Final concentration of L-Cys used for the
induced to aggregate upon addition of the L-cysteine (L-Cys), which study was 34.3 lM, and final concentrations of phosphate was
can bind to Au NPs via a thiol group. We took the advantage of size- 7.6 and 3.8 mM at pH 7 and 5, respectively. Different concentra-
dependent nanoparticle aggregation to construct a calibration tions of two buffers were used, because when pH tends to the
model for determination of Au NPs size, as the first application of acidic range, lower ionic strength is needed to induce aggregation
aggregation kinetic for determining Au NPs size. In previous works of Au NPs.
[25–31], there was not any calibration model to determine the size
of nanoparticles based on biomolecule-induced aggregation. As Synthesis of colloidal Au NPs
mentioned before in order to determine size of nanoparticles e.g.
gold nanoparticles, several methods [5–8] are employed; besides The citrate-capped Au NPs was prepared following the well-
of their time consuming and high costs aspects, these methods documented Frens method [34] to obtain monodispersed colloidal
need considerable concentration of particles and also sample prep-
aration should be considered. But in present work we will show ul-
tra-low concentration of Au NPs will be used to build calibration Table 1
Details for the size-selective synthesis of Au NPs by Turkevich method.
model based on biomolecule-induced aggregation which makes
this approach as a practical process for Au NPs size. Herein a ki- Set Amount of Amount of Color Average diameter (nm)
netic data matrix was obtained per each Au NPs sample allowing HAuCl4 trisodium
Observeda Reportedb
solution citrate solution
exploit the advantage of multivariate data analysis. In multivariate
(12.5 mM, ml) (1%, ml)
data, the objects (independent variables) are described by many
A 2 2 Red 15 16
variables (dependent variables). Multivariate calibration is differ-
B 2 1.75 Red – 20
ent from univariate calibration because it uses more than one C 2 1.5 Red 24.5 25
dependent variable. This leads to new chances, e.g. using full spec- D 2 1.25 Pinkish – 32
tra rather than the signal at a single wavelength. Using the entire red
spectral information may generally, cause to better results [32]. E 2 1 Pink – 41
F 2 0.8 Pink – 55
Chemometrics presents methods that allow the analysis of multi-
a
variate data where, the enormous amount of data is compressed Observed average diameter in present work by TEM image.
b
to meaningful information [32]. A common type of quantitative Expected average diameter according to Ref. [30].
590 F. Rabbani et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 115 (2013) 588–594

Au NPs over a wide size range. The sizes of Au NPs in this reductive
synthesis were controlled changing the amount of citrate to
HAuCl4 ratio. For example, a standard procedure for the prepara-
tion 32 nm gold colloid is as follows. A 50 ml aqueous solution of
HAuCl4 (0.5 mM) was heated to boiling with vigorous stirring
and 1.25 ml of trisodium citrate (1%) was added all at once. The
yellow solution turned clear, dark blue and then a pinkish red color
within a few minutes. The solution was heated for an additional
15 min after the red color was observed. The solution was then re-
moved from heat and kept stirring for 15 min. After that the solu-
tion was set aside to cool down to room temperature. Details for
the size-selective synthesis for different sets of Au NPs are also
summarized in Table 1.
Results and discussion
L-Cysteine-induced aggregation and UV–Visible absorbance
characterization
Au NPs have been prepared by Frens or Turkevich method,
which offered us the ease of achieving Au NPs over a wide range
of sizes. This was possible by varying [Au (III)]/[citrate] ratio during
the reduction process. This method is very often used because the
surface of the nanoparticles can be functionalized by a variety of
ligands via the displacement of the citrate capping ligands from
the Au nanoparticle surface. Upon addition of L-Cys to the cit-
rate-stabilized Au NPs, at the fixed ion strength and pH, the color
of the solutions becomes purple to blue indicating the formation
of aggregates amongst the Au NPs. The aggregation of citrate-stabi-
lized Au NPs was induced by L-Cys as a molecular linker. For the
association of Au NPs, the Cys-Au NPs interaction involves: (1)
the adsorption of L-Cys on the Au NP surface, upon displacement
with the original citrate-capping molecules and (2) the zwitter-
ion-type electrostatic interactions between the L-Cys groups bound
to the Au NP surface [35]. The changes in the SPR of the nanopar-
ticles occurring during this process were measured to study the
size effect of Au NPs on the kinetics of the aggregation process.
In our experiment, the cysteine-mediated assembly process was
examined through UV–Vis spectroscopy measurements. As men-
tioned before, the size of the Au NPs was found to affect the kinet-
ics of the assembly process. Fig. 1 shows spectral evolutions of
surface plasmon (SP) bands resulting from the aggregation process
of Au NPs with average diameter of 15, 20, 32, 41 and 55 nm, which
were collected over the course of 30, 30, 40, 60 and 60 min respec-
tively, at pH 7. Upon addition of 34.3 lM L-Cys into the Au NPs

Fig. 1. Time-dependent absorption spectra of Au NPs aggregation with different diameters of (a) 15 nm, (b) 20 nm, (c)32 nm, (d) 41 nm, (e) 55 nm. Spectra were monitored
for (a and b) 30 min, (c) 40 min, (d and e) 60 min. Data were carried out at [L-Cys] = 34.3 lM, pH = 7 and the same time intervals: (a) [Au15nm] = 1.6 nM, (b) [Au20nm] = 0.67 nM,
(c) [Au32nm] = 0.15 nM, (d) [Au41nm] = 0.066 nM, (e) [Au55nm] = 0.044 nM.
 F. Rabbani et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 115 (2013) 588–594
Fig. 2. Time-dependent absorption spectra of Au20nm aggregates at two concentrations: (a) [Au20nm ] = 0.17 nM; (b) [Au20nm] = 0.84 nM. Spectra were monitored for (a)
50 min and (b) 30 min. Data were carried out at [L-Cys] = 34.3 lM, pH = 7 and the same time intervals.
solutions, the characteristic SP band shifts towards longer wave-
lengths and its intensity decreases with increasing reaction time.
As seen in Fig. 1 each of data sets of Au NPs shows distinctly differ-
ent changes in the UV–Vis spectra as well as different spectral evo-
lution rate which could appears the effect of Au NP size on the
aggregation process kinetic. In view of the fact the Au NPs concen-
trations are not the same in all sets of solutions, therefore any ob-
served changes cannot reveal the effect of Au NPs sizes. More
discussion about Au NPs size and concentration has been left for
next section.
Effect of Au NPs concentration and size
In the beginning of this part, effect of Au NPs concentration on
the aggregation procedure was studied. Fig. 2 shows a typical set of
UV–Vis spectra at two different concentrations of 20 nm diameter
Au NPs (Au20nm), under the same conditions. This figure elaborates
the effect of Au NPs concentration on the kinetic of aggregation
process (Fig. 2a and b). It is clear that the reaction time of associa-
tion is depended on amount of nanoparticles, i.e. the lower concen-
tration needs the higher reaction time to be completed. Hence 30
and 50 min have been selected as optimized time for 0.84 and
0.17 nM Au NPs, respectively. Fig. 2 indicates more rapid and great-
er shift toward longer wavelengths for higher concentration rather
than lower one. In continue the rate constants (k) for the stated
concentrations were calculated by multivariate curve fitting using
the first order reaction model, which are 2.4  102 min1
(0.17 nM) and 2.3  101 min1 (0.84 nM) for the data of Fig. 2a
and b, respectively. The rate constants point a difference of 1 or-
der of magnitude that illustrates the higher concentrations kineti-
cally favored aggregation.
Fig. 3. Absorbance as a function of time for the Au NP aggregates for Au32nm and
Au41nm with the same concentration 0.078 nM. Spectra were monitored for Au32nm
and Au41nm in 50 min. Data were carried out at [L-Cys] = 34.3 lM, pH = 7 and the
same time intervals.
591
The second part of this investigation includes the effect of Au
NPs sizes on the aggregation process. In this case, two Au NPs sam-
ples with different sizes have been chosen which were in the same
concentration. Clearly, effect of size on the aggregation process can
be seen in Fig. 3 which is a plot of absorbance as a function of time
at 650 nm for two Au NPs samples with different sizes. This kinetic
behavior was very similar to that of concentration effect. Typically
the rate of spectral changes in the presence of L-Cys has been in-
creased with size of Au NPs. Similarly, the estimated rate constants
(k) were obtained 1.4  103 min1 and 1.1  102 min1 for
Au32nm and Au41nm, respectively.
Effect of pH
The aggregation process was also dependent on the pH of reac-
tion medium. Fig. 4 shows the SP bands related to the aggregation
of Au20nm and Au55nm at pH 7 and 5. It was found that the spectral
evolution changes of both Au20nm and Au55nm at pH 5 are greater
than that of pH 7. The k values were obtained for Au20nm, which
are 1.4  101 min1 (pH 7) and 3.9  101 min1 (pH 5) (Fig. 4a
and b) as well as for Au55nm which are 4.5  103 min1 (pH 7)
and 4.6 min1 (pH 5) (Fig. 4c and d). These results reveal that the
rate of aggregation decreases at the higher pH, which can be illus-
trated according to the mechanism of aggregation. L-Cys has three
pKa values, including pKa1 (1.92), pKa2 (8.37) and pKa3 (10.70) re-
lated to deprotonating of carboxylic, thiol and amine functional
groups, respectively [36]. At the isoelectric point of cysteine
(pI = 5.02), the –CO2H groups are mostly deprotonated to CO 2
while –NH2 is protonated to NHþ 3 , maximizing the zwitterionic
electrostatic interactions. On the other hand, raising pH to 7.0
causes deprotonating and decreasing the percentage of NHþ 3 . Sub-
sequently zwitterionic electrostatic interactions are no longer at
maximum value. Hence Au NPs repel each other due to negative
charge corresponds to CO 2 [36].
Besides, difference between the k values and shift toward longer
wavelengths are more significant for Au55nm than Au20nm with
decreasing the pH, which means the effect of pH on the association
process is not the same for Au20nm and Au55nm. This result could be
referred to difference in Au NPs concentrations. Considering the

fact that the same initial concentration of AuCl4 was used to syn-
thesize all Au NPs with different sizes, the Au NPs solutions with
larger particle diameter have lower concentrations than solutions
with smaller diameters. There for the concentration of synthesized
Au20nm is higher than that of Au55nm. Consequently the k values are
in the same order for two data sets of Au20nm (Fig. 4a and b) due to
the effect of Au NPs concentration on the association process.
Hence in the case of Au55nm with lower concentration compared
to the Au20nm, the effect of pH can be exhibited more clearly, and
the rate of spectral evolution decreases significantly at pH 7
(Fig. 4c and d).
592 F. Rabbani et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 115 (2013) 588–594

Fig. 4. Time-dependent absorption spectra of Au NPs aggregation at different pH, (a and b) 20 nm, (c and d) 55 nm. Spectra were monitored for (a and b) 30 min, (c) 60 min,
(d) 30 min. Data were carried out at [L-Cys] = 34.3 lM and the same time intervals: (a) [Au20nm] = 0.67 nM at pH 7, (b) [Au20nm] = 0.67 nM at pH 7, (c) [Au55nm] = 0.044 nM at
pH 5, (d) [Au55nm] = 0.044 nM at pH 5.

Determination of Au NPs sizes based on cysteine-induced aggregation: Accordingly five different sizes of Au NPs (15, 20, 32, 41 and
construction a calibration set 55 nm) have been employed to develop calibration set for size
determination at pH 7. Some of representative data sets were
According to our observations, pH, size and concentration of Au shown in Figs. 1–4. In Table 2, the different samples used in the cal-
NPs influenced the behavior of aggregation kinetics in this system. ibration model are summarized. As mentioned before the aggrega-
So our aim was to demonstrate the potential of nanoparticle aggre- tion kinetics of Au NPs at a fixed pH and ionic strength not only
gation to build a calibration model, which enables the determina- depend on the size of Au NPs but also depend on the concentration
tion of nanoparticle size in different concentrations. For modeling of NPs. Thus 17 calibration samples with varying of the size and
the relation between sizes of NPs and the kinetics behavior of concentration of NPs were used for constructing the calibration
aggregation a calibration set should be constructed by varying model. For example in the case of Au41nm, there were 3 samples
the sizes and concentrations of NPs at a fixed pH. Based on this with different concentration of Au NP. For each sample with cer-
modeling the variation in the optical response can be used for tain concentrations and sizes of Au NPs a multivariate data matrix
determination of Au NP size of an unknown sample. Such calibra- was obtained by monitoring the changes in the UV–Vis spectra
tion can model the relation between NPs sizes and kinetic spectral during the aggregation process over the reaction time.
changes due to aggregation in a certain concentration range. The first step of analyzing was using principal component anal-
ysis (PCA) as a data reduction procedure on each data separately. In
PCA each data matrix was decomposed to scores and loadings ma-
Table 2 trixes using singular value decomposition (SVD). Afterwards the
The sample of different sizes used in the calibration set normalized loading matrix including the most significant loadings
for the size determination of Au NPs. vectors which describe most of the data variation was chosen. Then
Sample number Diameter (nm) loadings vectors were augmented in a long vector. In other word
Sample 1 15
Sample 2 15
Sample 3 20
Table 3
Sample 4 20
Calculated and expected diameter.
Sample 5 20
Sample 6 20 Sample number Expected diameter (nm) Calculated diameter (nm)
Sample 7 32
Sample 1 15 15.2
Sample 8 32
Sample 2 15 14.8
Sample 9 32
Sample 3 20.0 19.7
Sample 10 32
Sample 4 20.0 19.4
Sample 11 32
Sample 5 24.5 24.9
Sample 12 41
Sample 6 24.5 25.8
Sample 13 41
Sample 7 32.0 29.8
Sample 14 41
Sample 8 55.0 54.9
Sample 15 41
Sample 9 55.0 57.5
Sample 16 55
Sample 17 55 Relative root mean square error 3.76%.
 F. Rabbani et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 115 (2013) 588–594
Fig. 5. TEM images of the Au NPs (a) 15 nm (set A in Table 1), and (b) 24.5 nm (set C
in Table 1).
using this straightforward and easy strategy we converted a matrix
to a vector, per each Au NP sample. The number of selected signif-
icant loadings was not the same for all samples. In order to have
the equal dimension for all augmented loading vectors we used
the zeros vectors in the cases with less factors. Eventually obtained
loadings vectors were used to construct the calibration model in-
stead of original data, using PLS regression. It is generally known
that the number of PLS factors (latent variables) is critical param-
eter using the PLS algorithm. In order to determine the optimum
number of factors (latent variables) for the partial least squares
calibration model, the cross validation procedure was applied.
The F statistic was carried out to determine the smallest model
(fewest numbers of factors). The predictive ability of the model
was described in term of relative root mean square error for esti-
mation of size in the data sets. A relative root mean square error
of 2.28% was obtained for this calibration set. Once the model
was constructed, size prediction of Au NPs was performed for
Fig. 6. Size distribution from DLS measurements of set C in Table 1, with average
diameter of 26.7 nm.
593
unknown samples. In the case of unknown sample the same data
compressing was again carried out for 9 unknown samples. Table 3
shows the results of size prediction for unknown samples with cal-
culated relative root mean square error of 3.76%. As can be seen the
predicted sizes of unknown samples were generally in agreement
with the expected size and what was obtained by TEM images. Suc-
cess of the general linear PLS modeling for dependency of average
size to aggregation profiles is an evidence of linearity in considered
system. Interestingly the prediction ability of this model for deter-
mination of average diameter of Au NPs is independent of concen-
tration. TEM images of selected samples of Au NPs can be seen in
Fig 5. For more evidence the diameter of a sample (set C in Table 1)
were measured by DLS (Fig. 6). This technique shows a narrow size
distribution with an average diameter of 26.7 nm. A significantly
larger average diameter was obtained by DLS compared to TEM,
however is due to measuring the hydrodynamic size of the nano-
particles when using DLS.
Conclusion
In conclusion the cysteine-mediated assembly process of Au NPs
was investigated through UV–Vis spectroscopy measurements. It
was observed that the aggregation rate has direct relation with con-
centration and size of Au NPs, and inverse relation with pH. It is con-
firmed that the effect of pH is more significant in the case of low Au
NPs concentration. To accurate investigation of size and concentra-
tion effects of Au NPs, we used the solutions in the same concentra-
tion with different sizes as well as solutions with the same size in
different concentrations. According to observations, a method to
measure the size of Au NPs was setup based on kinetically monitor-
ing the UV–Vis spectra. Consequently the kinetic size-dependent
aggregation of citrate-capped Au NPs was used for size determina-
tion of Au NPs which range in size from 15 to 55 nm at different con-
centrations. Such dependency had been reported from many
previous studies. However by now no calibration model has been
reported to determine the size of Au NPs based on the aggregation
process. The PLS algorithm was used successfully to construct a lin-
ear calibration model with low error for unknown samples. This
method possesses advantages of high accuracy, which is compara-
ble with TEM and DLS methods. This procedure can be expanded
to the size determination of other nanoparticles that exhibit a
change in optical spectra due to kinetic aggregation.
References
[1] C. Weisbecker, M. Merritt, G. Whitesides, Langmuir 12 (1996) 3763–3764.
[2] S. Peschel, G.A. Schmid, Chem. Int. Ed. Engl. 34 (1995) 1442–1443.
[3] P. Alivisatos, Nat. Biotechnol. 22 (2004) 47–52.
[4] B.D. Chithrani, A. Ghazani, W.C.W. Chan, Nano Lett. 6 (2006) 662–668.
[5] Y.G. Kim, S.K. Oh, R.M. Crooks, Chem. Mater. 16 (2004) 167–172.
[6] S. Sato, S. Wang, K. Kimura, J. Phys. Chem. 111 (2007) 13367–13371.
[7] R.J. Arnold, J.P. Reilly, J. Am. Chem. Soc. 120 (1998) 1528–1532.
[8] Y. Negishi, T. Tsukuda, J. Am. Chem. Soc. 125 (2003) 4046–4047.
[9] Y. Sun, Y. Xia, Analyst 128 (2003) 686–691.
[10] A.D. McFarland, R.P. Van Duyne, Nano Lett. 3 (2003) 1057–1062.
[11] C.A. Mirkin, M.A. Ratner, Annu. Rev. Phys. Chem. 101 (1997) 1593–1604.
[12] D.L. Klein, R. Roth, A.K.L. Lim, A.P. Alivisatos, P.L. McEuen, Nature 389 (1997)
699–701.
[13] L.A. Bumm, J.J. Arnold, M.T. Cygan, T.D. Dunbar, T.P. Burgin, L. Jones, D.L. Allara,
J.M. Tour, P.S. Weiss, Science 271 (1996) 1705–1707.
[14] M.A. Rampi, O.J.A. Schueller, G.M. Whitesides, Appl. Phys. Lett. 72 (1998)
1781–1783.
[15] S.K. Ghosh, S. Kundu, M. Mandal, S. Nath, T. Pal, J. Nanopart. Res. 5 (2003) 577–
587.
[16] G. Mie, Ann. Phys. 25 (1908) 377–445.
[17] R.C. Mucic, J.J. Storhoff, C.A. Mirkin, R.L. Letsinger, J. Am. Chem. Soc. 120 (1998)
12674–12675.
[18] C.A. Mirkin, R.L. Letsinger, R.C. Mucic, J.J. Storhoff, Nature 382 (1996) 607–609.
[19] J.J. Storhoff, A.A. Lazarides, R.C. Mucic, C.A. Mirkin, R.L. Lestinger, G.C. Schatz, J.
Am. Chem. Soc. 122 (2000) 4640–4650.
[20] Z. Chen, P. Sheng, D.A. Weitz, H.M. Lindsay, M.Y. Lin, P. Meakin, Phys. Rev. B 37
(1988) 5232–5235.
594 F. Rabbani et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 115 (2013) 588–594
[21] M.Y. Lin, H.M. Lindsay, D.A. Weitz, R.C. Ball, R. Klein, P. Meakin, Proc. R. Soc.
London Ser. A 423 (1989) 71–87.
[22] C. Leung, C. Wirouchaki, N. Berovic, R.E. Palmer, Adv. Mater. 16 (2004) 223–226.
[23] C.M. Niemeyer, Angew. Chem. Int. Ed. 40 (2001) 4128–4158.
[24] E.J. Yoo, T. Li, H.G. Park, Y.K. Chang, Ultramicroscopy 108 (2008) 1273–1277.
[25] T. Kim, C.H. Lee, S.W. Joo, K. Lee, J. Colloid Interface Sci. 318 (2008) 238–243.
[26] I.-I.S. Lim, D. Mott, M.H. Engelhard, Y. Pan, S. Kamodia, P.N. Njoki, S. Zhou, L.
Wang, C.J. Zhong, Luo, J. Anal. Chem. 81 (2009) 689–698.
[27] I.-I.S. Lim, D. Mott, P.N.N.W. Ip, Y. Pan, S. Zhou, Assembly 24 (2008) 8857–
8863.
[28] M. Moskovits, B. Vlckova, J. Phys. Chem. B 109 (2005) 14755–14758.
[29] N. Kallay, S. Zalac, J. Colloid Interface Sci. 253 (2002) 70–76.
[30] R. Prasher, P.E. Phelan, P. Bhattacharya, Nano Lett. 6 (2006) 1529–1534.
[31] T. Laaksonen, P. Ahonen, C. Johans, K. Kontturi, ChemPhysChem 7 (2006)
2143–2149.
[32] B.G.M. Vandeginste, D.L. Massart, Handbook of Chemometrics and
Qualimetrics: Part B, Elsevier, Amsterdam, 1998.
[33] S. Wold, M. Sjostrom, L. Eriksson, Chemom. Intell. Lab. Syst. 58 (2001) 109–
130.
[34] G. Frens, Nature 241 (1973) 20–22.
[35] I.-I.S. Lim, C.J. Zhong, Gold Bull. 40 (2007) 59–66.
[36] Z.N. Gao, J. Zhang, W.Y. Liu, J. Electroanal. Chem. 580 (2005) 6–16.