Journal of Colloid and Interface Science 407 (2013) 816

Contents lists available at SciVerse ScienceDirect

Journal of Colloid and Interface Science

www.elsevier.com/locate/jcis

Quantitative nucleation and growth kinetics of gold nanoparticles

via model-assisted dynamic spectroscopic approach
Yao Zhou, Huixuan Wang, Wenshuang Lin, Liqin Lin, Yixian Gao, Feng Yang, Mingming Du,
Weiping Fang, Jiale Huang, Daohua Sun, Qingbiao Li
Department of Chemical and Biochemical Engineering, College of Chemistry and Chemical Engineering, National Engineering Laboratory for Green Chemical Productions of
Alcohols, Ethers, and Esters, Xiamen University, Xiamen 361005, PR China
Key Laboratory for Chemical Biology of Fujian Province, Xiamen University, Xiamen 361005, PR China

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

Article history: Lacking of quantitative experimental data and/or kinetic models that could mathematically depict the
Received 19 January 2013 redox chemistry and the crystallization issue, bottom-to-up formation kinetics of gold nanoparticles
Accepted 12 June 2013 (GNPs) remains a challenge. We measured the dynamic regime of GNPs synthesized by L-ascorbic acid
Available online 27 June 2013
(representing a chemical approach) and/or foliar aqueous extract (a biogenic approach) via in situ spec-
troscopic characterization and established a redoxcrystallization model which allows quantitative and
Keywords: separate parameterization of the nucleation and growth processes. The main results were simplied as
Gold nanoparticles
the following aspects: (I) an efcient approach, i.e., the dynamic in situ spectroscopic characterization
Kinetics
Model
assisted with the redoxcrystallization model, was established for quantitative analysis of the overall for-
Crystallization mation kinetics of GNPs in solution; (II) formation of GNPs by the chemical and the biogenic approaches
Redox experienced a slow nucleation stage followed by a growth stage which behaved as a mixed-order reac-
tion, and different from the chemical approach, the biogenic method involved heterogeneous nucleation;
(III) also, biosynthesis of aky GNPs was a kinetic-controlled process favored by relatively slow redox
chemistry; and (IV) though GNPs formation consists of two aspects, namely the redox chemistry and
the crystallization issue, the latter was the rate-determining event that controls the dynamic regime of
the whole physicochemical process.
2013 Elsevier Inc. All rights reserved.

1. Introduction density, size and polydispersity of GNPs were estimated to deduce

Noble metal nanoparticles have many important applications
and have long been synthesized by various chemical methods
[1,2]. In Recent years biosynthesis of noble metal nanoparticles
using various species of plants and microorganisms has arisen as
another green economical synthetic approach [310]. Despite
numerous empirical trials [1113], purposive manipulation of the
redox chemistry and the crystallization from a kinetic and thermo-
dynamic level is still necessary [1419] to realize rational design of
noble metal nanoparticles with desirable shapes or sizes. Nonethe-
less, it was remarked that understanding of crystallization, espe-
cially the nucleation process, has been very much limited mainly
due to lack of quantitative experimental data [17,20]. For forma-
tion kinetics of the intensely investigated gold nanoparticles
(GNPs) [21], ever since 1951 [22], temporal evolution of number
Corresponding author at: Department of Chemical and Biochemical Engineer-
ing, College of Chemistry and Chemical Engineering, National Engineering Labora-
tory for Green Chemical Productions of Alcohols, Ethers, and Esters, Xiamen
University, Xiamen 361005, PR China.
E-mail address: kelqb@xmu.edu.cn (Q. Li).
the formation mechanism based upon TEM/SAXS characterization
[14,15,19,2327]. However, the dataset of the overall nucleation
and growth rate was still not accessible. In situ UVvis spectros-
copy could make a compensation [17,28,29]. Occasionally the char-
acterized SPR of GNPs was observed to evolve in a sigmoidal
pattern over the reaction course [14,30]. The Avrami model for so-
lid-state or polymer crystallization [31] was employed to t such
sigmoidal curves [30,32]; yet physical meanings of the model
parameters were still controversial [33]. In general such dynamic
spectroscopic data were restricted to qualitative analysis. Besides,
currently quantitative kinetic study on biosynthesis of GNPs by
aqueous extract of biomass is not available and is a more challeng-
ing issue compared with the chemical approach since the active
bio-ingredients responsible for reducing the gold precursors and/
or stabilizing the GNPs have not been identied. Herein, the forma-
tion kinetics of GNPs was investigated by dynamic in situ UVvis
spectroscopy and a redoxcrystallization model was established
for quantitative analysis of thus obtained kinetic data. Based upon
such an approach, the overall nucleation and growth kinetics for
formation of spherical and aky GNPs via both chemical and bio-
synthetic approaches were for the rst time evaluated and

0021-9797/$ - see front matter 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jcis.2013.06.027
 Y. Zhou et al. / Journal of Colloid and Interface Science 407 (2013) 816
compared quantitatively, and insights were provided into the role
of the redox chemistry and the crystallization issue upon the dy-
namic regime of the whole physicochemical process.
2. Experimental section
2.1. Materials and apparatus
Polyvinyl Pyrrolidone K-30 (PVP, AR) and HAuCl44H2O (AR)
were from Sinopharm Chemical Reagent Co., Ltd., and L-ascorbic
acid (AA) was from Sangon Biotech (Shanghai) Co., Ltd. UVvis
Spectrophotometer (UV-1800) was from Shimadzu Scientic
Instruments and Spectrophotometer Molecular Device (SPECTRA
MAX190) was from Molecular Devices, Inc. in USA.
2.2. Preparation of foliar aqueous extract
The powder of Cinnamomum Camphor (C. Camphor) leaves was
prepared according to procedures described in our previous work.
The foliar lixivium was prepared by mixing and shaking 1 g foliar
powder with 100 mL deionized water for 3 h (150 rpm, 30 C).
Afterward, the mixture was decanted using qualitative lter paper
to obtain the ltrate. The concentration of thus prepared ltrate
was denoted as 10 mg/mL, and it was diluted when necessary.
Also, for one-pot concomitant biosynthesis of spherical and aky
GNPs, the mixture of 1 g foliar powder of Cinnamomum Japonicum
(C. Japonicum) with 50 mL deionized water was boiled for 5 min;
subsequently the mixture was decanted using the qualitative lter
paper to obtain the foliar broth.
2.3. Synthesis of GNPs and measurement of formation kinetic curves
C. Camphora leaves lixivium (3.0 mL, 110 mg/mL) or mixture of
AA and PVP (6 mg/mL) was transferred into a thoroughly-cleaned
cuvette inserted in the UVvis Spectrophotometer. Thereafter, a
portion of aqueous HAuCl4 (15 lL, 0.04856 M) was rapidly pipetted
into the cuvette, and the mixture was immediately blended for
about 510 s. The absorbance at around the SPR peak wavelength
was then recorded against the blank sample at every 2 or 5 s.
The operational time (t0), i.e., the time starting from the addition
of HAuCl4 to the moment when the apparatus began to detect
the absorbance, was recorded. Thus obtained dataset were the
Abs-time (Abs-t) proles. Unless otherwise specied, the above
operations were conducted at room temperature which was
around 17.0 C.
To investigate the effects of temperature, 100 lL lixivium
(6 mg/mL) or AA solution (AA 0.8 mM, PVP 5 mg/mL) was pipetted
into a selected well of the 96 orice plate. Then the plate was pre-
heated to the set temperature ranging from room temperature to
45 C in the Spectrophotometer Molecular Device. Afterward,
100 lL preheated aqueous HAuCl4 (0.48 mM) was rapidly pipetted
into and mixed with the lixivium in the well. The absorbance of the
mixture at 540 nm was recorded against the blank sample at every
5 s by the apparatus to obtain the Abs-t proles. The operational
time t0 was also recorded. Unless otherwise specied, in all of
the above reactions the reductants were in excess with respect to
the gold precursor.
2.4. TEM observation of spherical and aky GNPs
For concomitant biosynthesis of spherical and aky GNPs, 60 lL
aqueous HAuCl4 (0.04856 M) was mixed with 3 mL broth of C.
japonicum leaves in a cuvette. The Abs-t plots at both 540 nm
(for spherical GNPs) and 1020 nm (for aky GNPs) were obtained,
based upon which TEM samples were taken at pre-established
9
time during the reaction. At each time, a carbon-coated copper grid
was soaked in the reaction solution for around 20 s, which then
was dried immediately using a vacuum drier. TEM observations
were performed on a Technai F-30 Transmission Electron Micro-
scope. Size of the resulting GNPs was estimated on the basis of
TEM images. Size of the aky GNPs was denoted by the major axis
length.
3. Results and discussion
Synthetic processes using moderate or weak reductants
[14,19,23] are advantageous for kinetic study compared with boro-
hydride-based synthesis which completed within milliseconds
[15]. For comparison, here spherical GNPs were synthesized by
the widely used L-ascorbic acid [34,35] (AA case) or the lixivium
of C. Camphor leaves [36] (bio case). The Abs-t proles obtained
by recording the absorbance at 526 nm for the AA case and at
540 nm for the bio case by UVvis spectroscopy were found to
be sigmoidal in shape. As exemplied in Fig. 1a, the sigmoidal
curve has an induction stage (I), a growth stage (II) and a saturation
stage (III). The crystallization fraction (x) of a sample could be de-
noted as the ratio between its absorbance at time t (At) and at the
end (Amax) [30]. Such sigmoidal kinetic curves were for the rst
time analyzed by conventional rate laws. As exemplied in
Fig. 1bd, the Abs-t proles were converted to responding xt pro-
les according to the linear rate equations of zeroth, rst and sec-
ond-order reactions (see SI). As seen in Fig. 1b, within ca. 3448 s, a
signicant linear relationship was observed between x and t, sug-
gesting that the overall crystallization rate was zeroth-order with
respect to the reactants (namely gold atoms in this case). There-
fore, at this stage GNPs surface was saturated by gold adatoms,
i.e., excessive gold monomers competed for relatively limited ac-
tive sites. It also indicated that at this stage all of the Au(III) species
had either been reduced into its atomic state or into its lower
charge states. With the crystallization proceeding forward, it chan-
ged to rst-order within ca. 4654 s (Fig. 1c) and subsequently to
second-order within 5256 s (Fig. 1d), which indicated that con-
centration of gold monomers declined and its inuence became
evident on the crystallization rate. Such a mixed-order-reaction
pattern was observed in the growth stage of GNPs for both AA
and bio cases (Fig. S1) and it quite resembled the famous Michae-
lisMenten kinetics in enzyme catalysis [37].
It is widely believed that the kinetics of bottom-to-up formation
of GNPs should take both of the redox chemistry and the crystalli-
zation issue into consideration. The relatively long period of zer-
oth-order reaction observed in Fig. 1b indicated existence of free
gold atoms (M 0l ) in solution. That is, the reduction of the gold pre-
cursors was relatively faster than the crystallization process, i.e.,
the latter process which dominated over the former becomes the
rate-determining step during GNPs formation. Therefore as de-
picted in formula (F-1), we assume reduction of gold precursor
(Mn) by reductants (R) as a reversible reaction in dynamic chemical
equilibrium with an equilibrium constant Kc. Such a assumption
would lead to Eq. (1). Simultaneously, crystallization would occur
whereby free soluble gold atoms in liquid phase (M 0l ) were trans-
formed into solid state (M0s ). Formula (F-2) represents the typical
phase transformation behaviours of M 0l into M 0s via nucleation;
growth of nuclei via interaction between the surface active site
(s*) and M 0l was shown in formula (F-3). k01 and k02 were overall
nucleation and crystal growth rate constants, respectively.
Kc
Mn m $ M 0l F- 1
k01
M0l ! M 0s F- 2
 k02
M01 s ! M0s F-3
10 Y. Zhou et al. / Journal of Colloid and Interface Science 407 (2013) 816

Fig. 1. (a) Typical sigmoidal Abs-t prole and its responding x-t proles obtained by treating the Abs-t prole with (b) zeroth-order, (c) rst-order and (d) second-order rate
laws for GNPs synthesized in the AA case.

Assuming the three formulas are pseudo-elementary reactions, parameters, i.e., k1 the apparent overall nucleation rate constant
then we have Eq. (2) which denotes the crystallization rate. and k2 the apparent overall growth rate constant, The unit of k1
and k2 is s1. For convenience, thereafter Eq. (6) was referred to
M01  K c  R  Mn  1 as redoxcrystallization model according to its derivative mecha-
nism. Specically, for tting of the Abs-t proles by the RC model,
dMos  Eq. (6) was further transformed into Eq. (7).
k01 M 01  k02 s M01  2
dt
Approximately, [s*], the concentration of the active sites, is pos- k1 k2
x1 6
itively proportional to the total surface area of the GNPs in the k2 k1 sk1 k2 t
solution. Thus, Eq. (3) could be derived whereby [s*] was denoted
 
as a linear function of [M 0s ], with a constant e and MAu denotes k1 k2
the molar mass of Au, qAu the density of gold, rt the equivalent At Amax 1  7
k1 s k1 k2 t k2
average radius of GNPs.
Interestingly, RC model has identical mathematical expression
3MAu 0 as the FinkeWatzky model which was established upon an auto-
s  M  eM0s  3
qAurt s catalytic mechanism [38] (remarks on their essential differences
Meanwhile, i the mass conservation law leads to Eq. (4) where see SI). The RC model performs very well in tting those sigmoi-
[Mn]0 was the initial concentration of the gold precursor. dal Abs-t proles (vide infra), which justied the validity of its
mechanism. Such a result indicated that though thus formation
Mn 0 M n  M0l  M 0s  4 of GNPs is a physicochemical process which consists of two as-
pects, i.e., the redox chemistry and the crystallization issue, its dy-
The sum of Eqs. (1)(4) gives the crystallization fraction x in Eq.
namic regime highly resembles the behaviours of a crystallization
(5) with a K c R=K c R 1.
process which is a pure physical phase-transformation
M 0s k02 eM n 0 k01 phenomenon.
x n 1 n n 5 As demonstrated in Fig. 2a and b, Eq. (7), namely the RC model,
M 0 k02 eM 0 k01 eak01 k02 eM 0 t
demonstrated excellent performance in tting the sigmoidal Abs-t
Let k1 = ak01 and k2 = ae[Mn]0k02, and then Eq. (5) could be plots for GNPs prepared by different concentrations of AA or foliar
transformed into Eq. (6) through which the crystallization fraction lixivium, with square of correlation coefcient (R2) varying from at
x was denoted as a function of the reaction time t, with two kinetic least 0.9950 to 0.9999 depending on the data density. The resulting
 Y. Zhou et al. / Journal of Colloid and Interface Science 407 (2013) 816 11

Fig. 2. Fitting of the Abs-t proles by RC model for gold nanohydrosol synthesized by (a) different concentrations of C. Camphora leaves lixivium and (b) by different
concentrations of AA solution; and the resulting values of (c) the nucleation rate constant k1 and (d) the growth rate constant k2. Dots are experimental data, the red line
calculated, similarly hereinafter. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

apparent overall nucleation rate constant k1 and the crystal growth
constant k2 were depicted in Fig. 2c and d, respectively. As ob-
served, at relatively lower concentration of foliar lixivium/AA, both
k1 and k2, which are apparent rate constants, increase with increas-
ing initial concentrations of foliar lixivium or AA. And as shown in
Fig. 2c, the nucleation rate constant k1 for the bio case (around
103) were at least around two orders of magnitude larger than
that of the AA case (around 105). However, contrarily, the growth
rate constant k2 for AA was larger than those of its counterpart
(Fig. 2d). Such a surprisingly large nucleation rate constant for
the bio case implied that the nucleation process during biosynthe-
sis of GNPs by foliar lixivium might involve heterogeneous nucle-
ation. For biosynthesis of GNPs, the foliar lixivium was usually
prepared by decanting the lixivium through the qualitative lter
paper. Therefore it is quite possible that some substances or even
large biocompounds existing in the lixivium might serve as heter-
ogeneous nucleus that provides active sites for nucleation which
thus could dramatically reduce the nucleation energy barrier and
thus boost the nucleation rate. As well, k1 for the bio case increases
almost mono-directionally via increasing initial concentration of
foliar lixivium, which might be because that higher initial concen-
tration of initial foliar lixivium would lead to denser heterogeneous
nucleus in the solution and thus larger extent of heterogeneous
nucleation in the reaction. Such results suggested that signicant
importance should be attached to the way how the foliar aqueous
extract was prepared or pretreated, since it determined the extent
of the heterogeneous nucleation.
In previous researches GNPs growth activation energy was eval-
uated by tting mean diameters of GNPs versus reaction time
[15,18,23,27], but even the SAXS-based researches failed to exam-
ine the nucleation dynamics [15,23]. As shown in Fig. 3a and b, sig-
moidal Abs-t proles for both the AA and bio cases obtained from
25 C to 45 C were well tted by RC model, with most R2 larger
than 0.999. Accordingly, values of k1 and k2 were obtained as a
function of temperature, which were well tted by Arrhenius
equation (Fig. 3c and d). Thus obtained apparent overall activation
energy for nucleation (DE(k1)) and growth (DE(k2)) for the AA case
was 40.5 1.2 and 20.2 1.7 kJ/mol, respectively. That is, the en-
ergy barrier for homogeneous nucleation is larger than the crystal
growth process, which thus allows kinetic control over their rela-
tive rate since larger activation energy indicates higher tempera-
ture sensitivity. Size and size distribution of GNPs is essentially
determined by relative rate of nucleation and growth [39]. Smaller
GNPs with narrower size distribution would be expected under
higher temperature. Such a prediction could nd support in exist-
ing researches, e.g., size of GNPs decreases mono-directionally
with rising temperature when 9-borabicyclo[3.3.1] nonane was
used as reductant [19]. It is worth mentioning that existence of
PVP as a ligand in the AA case could slow down the reaction be-
tween AA and HAuCl4. Therefore, without the addition of PVP,
the energy barrier for formation of GNPSs by L-ascorbic acid is sup-
posed to be even smaller than those acquired in the present work.
DE(k1) for the bio case was 35.8 1.3 kJ/mol, smaller than that of
the AA case, and DE(k2) for the bio case was 44.4 1.2 kJ/mol.
DE(k2) for the bio case was much larger than the AA case, implying
that the reducing potential of the plant extract was on average
lower than L-ascorbic acid. And a relatively low DE(k1) for the bio
case further proved that biosynthesis of GNPs by foliar lixivium
12 Y. Zhou et al. / Journal of Colloid and Interface Science 407 (2013) 816
Fig. 3. Fitting of Abs-t proles by RC model for spherical GNPs synthesized under different temperatures for (a) the bio case and (b) the AA case, values of (c) k1 and (d) k2
versus temperature for the bio and AA cases.
might involve heterogeneous nucleation which could dramatically
reduce the nucleation energy barrier [40]. More experimental evi-
dences about this speculation would also be contributed by our fu-
ture work.
The above results veried the applicability of the RC model to
tting of the sigmoidal Abs-t dataset. An evident advantage of the
RC model is that the nucleation rate and growth rate are sepa-
rately parameterized. With these two kinetic parameters, i.e., k1
and k2, the mechanistic information conceived in different stages
of the sigmoidal could be revealed via both mathematical devia-
tions and experimental results. As shown in Eq. (8), the exponen-
tial term in the RC model could be linearized when the term,
(k1 + k2)tnu where tnu denotes the length of the induction stage
in the Abs-t prole, is sufciently small (e.g., smaller than 0.3).
Therefore, Eq. (6), namely the RC model, could be simplied into
Eq. (9) where xtnu denotes the crystallization fraction at tnu. Eq. (9)
could be further converted to Eq. (10) which suggests that the
nucleation parameter k1 was inversely proportional to the length
of the induction stage of the Abs-t kinetic curves. Such linear rela-
tionship was further tested using values of k1, k2 obtained by tting
the previously obtained Abs-t proles with the RC model. And tnu
was approximately dened as the reaction time when the crystal-
lization fraction at tnu, xtnu , was equal to 0.05. It was found that val-
ues of k1, k2 and tnu experimentally obtained in the bio case could
satisfy the linearization condition required for establishment of Eq.
(8); accordingly, as shown in Fig. 4c, k1 and 1/tnu illustrated a sig-
nicant linear relationship, with a slope 26.9 which is close to the
constant r in eq 10 (which was calculated as 19.0 when xtnu 0:05)
in terms of magnitude. Nonetheless, for the AA case, we found that
its experimental values of k1, k2 and tnu failed to meet the lineari-
zation condition for establishment of Eq. (8). Consequently, as
demonstrated in Fig. 4 a, the linear relationship between k1 and
1/tnu became indistinctive for the AA case, yet a larger nucleation
rate constant k1 generally associated with smaller tnu, i.e., a shorter
induction stage.
expk1 k2  tnu  1 k1 k2  tnu 8
1
1  xtnu 9
1 k1 tnu
1  xtnu
t1
nu k1 k1 r 10
xtnu
The above mathematical derivation and experimental results
proved that the nucleation parameter k1 closely linked with the
induction stage of the sigmoidal Abs-t curves. Such a result indi-
cated that the induction stage with negligible absorbance was
actually a nucleation stage and production of gold nanoclusters
(GNCs) below 2 nm which have no SPR absorptions [21] was ex-
pected to be the dominating event at this stage. However, due to
lack of quantitative in situ microscopic techniques to characterize
the temporal evolution of density of GNCs within the nucleation
stage, currently we were not able to decide whether the nucleation
at this stage was a continuous process or a burst of nucleation ob-
served in the classical Lamer theory [41].
Also, we proved, both mathematically and experimentally, that
the growth parameter k2 is positively proportional to the maxi-
mum slope of the Abs-t prole in stage II. The rst-order differen-
tiation of Eq. (7) results in Eq. (11) where m is the sum of k1 and k2.
Generally, the previous experimental results revealed that k2 was
at least one order of magnitude larger than k1. Therefore, Eq.
(12), which gives the maximum crystallization rate, would be ob-
tained at the time tmax shown in Eq. (13). Such a relationship
was also conrmed by the experimental results. The maximum
 Y. Zhou et al. / Journal of Colloid and Interface Science 407 (2013) 816 13

Fig. 4. Relationship between k1 and tnu for the AA case (a) and the bio case (c); relationship between k2 and {dAt/Amaxdtx=50%} for the AA case (b) and the bio case (d).

crystallization rate was approximately denoted by the term {dAt/

Amaxdtx=50%} which was calculated at the point of the Abs-t dataset
where the crystallization fraction was around 50%. As observed in
Fig. 4b and d, signicant linear relationship was observed between
k2 and {dAt/Amaxdtx=50%} for both AA and bio cases. The coefcient
was 0.197 in Fig. 4 b for the AA case and 0.248 in Fig. 4d for the
bio case, very close to the theoretical value 0.250 in Eq. (13). Such
results veried that the stage II of the sigmoidal Abs-t proles, i.e.,
thereafter at t > tnu, was a growth stage whereby crystal growth
dominated over the nucleation behaviors.
!2 !2
0:5
dAt mm  k2 m
k2
 p 11
Amax dt m  k2  exp0:5mt 2 k2
exp0:5mt

 
dAt k2
12
Amax dt max 4

1 k2
t max ln
k1 k2 k1
Hence, the sigmoidal dynamic spectroscopic dataset in Fig. 1a
itself could enable evaluation of the overall nucleation and crystal
growth rate. And the separation of nucleation and crystal growth,
which severely affects the size distribution of GNPs [17], could
be visually characterized by the shift of stage I to stage II in the
Abs-t prole. In addition, our preliminary results revealed that size
of GNPs grew larger with increasing values of k2/k1 (Fig. 5) and size
distributions of GNPs narrowed down with shorter induction stage
or larger k1 (not shown here).
Compared with various chemical methods [42], biological
materials seem amazingly effective in manufacturing anisotropic
GNPs with two dimensions. Rapid high-yield of gold nanoprisms
was realized at room temperature using a long list of biomasses
[29,4347]. Various biocompounds were speculated as shape-
directing agents leading to formation of aky GNPs [29,4446].
Herein, aky GNPs were also synthesized facilely by broth of C.
Japonicum leaves, as shown in Fig. 6(1). Thus formation kinetics
of the aky GNPs was also understood based upon the dynamic
spectroscopic characterization together with the RC model. The
13 Fig. 5. Sizes and size distributions of biosynthesized GNPs (see SI for its specic
synthetic conditions) versus values of k2/k1.
Abs-t proles at 540 nm and 1020 nm were respectively followed,
both of which were sigmoidal in shape (Fig. 6(2)). The nucleation
stage at 540 nm lasted ca. 750 s whereas that at 1020 nm lasted
ca. 1150 s. The total reaction time in Fig. 6 was much longer than
that in Figs. 2 and 3 where only spherical GNPs were produced,
indicating formation of aky GNPs was a much slower process.
We have also synthesized aky GNPs under a series of other condi-
tions (Table S1), and the responding Abs-t proles for both spher-
ical and aky GNPs were acquired, all of which resembled what
depicted in Fig. 6(2) (see Fig. S2). As exemplied in Fig. 6(2), RC
model performed very well in tting those Abs-t proles, giving
responding values of k1 and k2 (Table 1). For all the circumstances,
values of k1 for spherical GNPs were larger than those for aky
GNPs synthesized in the same batch. Moreover, as mathematical
and experimental results proved that the nucleation and growth
process could be separately visualized by the Abs-t proles, purpo-
sive TEM sampling could be enabled to characterize the formation
14 Y. Zhou et al. / Journal of Colloid and Interface Science 407 (2013) 816

Fig. 6. (1) UVvis spectra and (2) sigmoidal Abs-t proles tted by RC model for spherical and aky GNPs synthesized in one batch. The black lines are experimental data
and the red curves are calculated by RC model. af in subgure (2) and (3) indicate TEM sampling points which characterize the evolution of spherical and aky GNPs versus
reaction time.

GNPs appeared in sample (c). This was consistent with the fact that
Table 1
RC model kinetic parameters for spherical and aky GNPs biosynthesized under
tnu for aky GNPs was around 400 s longer than that of the spher-
different conditions. ical GNPs (Fig. 6(2)). And thereafter both size and number density
of aky GNPs increased continuously within sample (c)(f). The
Cases* k1 105 (s) k2 103 (s)
above dynamic spectroscopic and microscopic analysis revealed
Sphere Flake Sphere Flake that nucleation behaviors of spherical and aky GNPs occurred
(a) 3.49 0.74 2.2 2.76 asynchronously. It has been recognized that twinned seed which
(b) 4.19 1.36 1.54 1.92 formed due to coalescence event between two unstable [1 1 1]
(c) 2.56 0.16 0.82 1.21
crystal facets [42,48], or preferential adsorption of shape-directing
(d) 2.21 0.42 1.05 1.13
(e) 3.44 0.68 1.23 1.74 agents would lead to plate-like structures [45]. Based upon such a
(f) 4.65 1.52 2.06 3.24 theory and the asynchronous nucleation behaviours between
(g) 2.52 0.50 2.54 3.05 spherical and aky GNPs, as depicted in Scheme 1, we speculated
(h) 3.08 0.55 3.91 5.11 that nucleation process of aky GNPs might consist of two steps.
*
Specic synthetic conditions for cases (a)(h) were shown in Table S1. Firstly, small GNCs were formed (which were also nucleus for
spherical GNPs), and then such GNCs coalesced or adsorbed spe-
cic capping agents preferentially, leading to formation of twinned
process. As shown in Fig. 6(2) and (3), sample (a) and (b) were ta- seeds, i.e., the nucleus of aky GNPs. Such a speculation well ex-
ken within the nucleation stage of the spherical GNPs and sample plained why a longer nucleation stage or smaller values of k1
(a)(c) within the nucleation stage of the aky GNPs. Consistent was observed for aky GNPs. As depicted in Scheme 1, the GNCs
with the previous speculation, only GNCs around 12 nm were ob- observed in sample (a) were born with two routes ahead of them:
served in sample (a) whereas sparse spherical GNPs appeared in (i) to grow into bigger spherical GNPs by annexation of gold ada-
sample (b). Thereafter, the number density of spherical GNPs grew toms (isotropic growth) or (ii) to form twinned seeds that grew
continuously within sample (c)(f). Compared with spherical into aky GNPs (anisotropic growth). The selectivity for conversion
GNPs, there was a delay in time for appearance of aky GNPs. No of GNCs to twinned seeds and thus aky GNPs was determined by
aky GNPs was observed in sample (a) and (b), and sparse aky the kinetic competition between routes (i) and (ii). In extreme
 Y. Zhou et al. / Journal of Colloid and Interface Science 407 (2013) 816
Scheme 1. Schematic illustration for kinetic-controlled biosynthesis of aky GNPs.
Fig. 7. Size and size distribution of spherical and aky GNPs shown in Fig. 6 versus
reaction time.
circumstances, fast redox reaction led to saturation of GNCs sur-
faces by gold adatoms (as evidenced by the zeroth-order reaction
depicted in Fig. 1b), leaving few active sites for coalescence or pref-
erential adsorption of shape-directing agents. As a result, only
spherical GNPs were produced in previous cases. Therefore, forma-
tion of aky GNPs is a kinetic-controlled process favored by rela-
tively slow redox reaction. That means, any factors that could
affect redox kinetics could have control over the formation of aky
GNPs. Such a prediction could nd support in existing researches
[45,47]. It also indicates that foliar extract is a natural factory in
15
manufacturing aky GNPs since our previous results proved that
compared with AA, the foliar aqueous extract in general have mod-
erate or weak reducing potential towards the gold precursors.
Furthermore, at the growth stage neither decrement in SPR
intensity (Fig. 6(2)) nor coalescence behaviours (Fig. 6(3)) was ob-
served for spherical GNPs, implying that aky GNPs grew via
annexation of adatoms on the twinned nucleus. This is differenti-
ated from the mechanism where aky GNPs grew via consuming
spherical GNPs, e.g., through assembly, fusion and sintering of
spherical GNPs [43]. Besides, values of k2 for aky GNPs were found
larger than those for the spherical GNPs synthesized in the same
batch (Table 1) and the spherical GNPs grew from 24 to 33 nm
whereas the aky GNPs grew from around 56 to 140 nm within
the growth stage (Fig. 7). Such shape-dependent growth rate might
be explained from two aspects: (I) a aky GNP with larger specic
surface area could provide more active sites to interact with ada-
toms than a spherical one; (II) the surface atoms varied in chemical
reactivity due to difference in coordination number in different
crystal facets [49].
4. Conclusion
In summary, we demonstrated that the dynamic spectroscopic
approach assisted with RC model was an efcient and revealing
methodology that enabled visualization and quantitative analysis
of the nucleation and growth processes for GNPs synthesized via
both chemical and biosynthetic approaches. The contributions lied
in the following several aspects: (I) the mechanism and the appli-
cability of the RC model to the formation kinetics of GNPs was
validated, which suggested that the crystallization issue was the
controlling factor that determined the dynamic regime of the
whole physicochemical process, yet the results also suggested that
the redox chemistry could signicantly affect the nucleation and
growth process and thus inuences the shapes and size of GNPs;
(II) meanwhile, for GNPs synthesized by AA or foliar extract which
were both moderate reductants, a slow nucleation stage and a
growth stage which evolve as a mixed-order reaction was sepa-
rately identied, and for biosynthesis of GNPs by foliar aqueous ex-
tract, it involved evident extent heterogeneous nucleation; (III)
formation of aky GNPs was also understood, whose nucleation
process occurred via two steps and thus needed a longer nucle-
ation stage than the spherical GNPs synthesized in the same batch,
and therefore formation of aky GNPs was a kinetic-controlled
process favored by slow redox reduction of gold precursors. Fur-
ther work which verify the applicability and validate the mecha-
nism of the RC model will be contributed to clarify the
fundamental differences and/or similarities between formation of
16 Y. Zhou et al. / Journal of Colloid and Interface Science 407 (2013) 816

GNPs and a pure phase transformation process from a kinetic [19] R. Sardar, J.S.J. Shumaker-Parry, J. Am. Chem. Soc. 133 (2011) 81798190.
[20] L. Qu, W.W. Yu, X. Peng, Nano Lett. 4 (2004) 465469.
perspective.
[21] M.C. Daniel, D. Astruc, Chem. Rev. 104 (2004) 293346.
[22] J. Turkevich, P.C. Stevenson, J. Hillier, Discuss. Faraday Soc. 11 (1951) 5575.
Acknowledgments [23] K. Biswas, N. Varghese, C.N.R. Rao, Small 4 (2008) 649655.
[24] J. Polte, R. Erler, A.F. Thnemann, S. Sokolov, T.T. Ahner, K. Rademann, F.
Emmerling, R. Kraehnert, ACS Nano 4 (2010) 10761082.
This work was supported by the national Natural Science Foun- [25] J. Polte, R. Kraehnert, M. Radtke, U. Reinholz, H. Riesemeier, A.F. Thnemann, F.
dation of China (No.21036004). Emmerling, J. Phys. Confer. Ser. 247 (2010) 012051.
[26] S.P. Shields, V.N. Richards, W.E. Buhro, Chem. Mater. 22 (2010) 32123225.
[27] R. Seshadri, G. Subbanna, V. Vijayakrishnan, G. Kulkarni, G. Ananthakrishna, C.
Appendix A. Supplementary material Rao, J. Phys. Chem. 99 (1995) 56395644.
[28] M. Sethi, M.R. Knecht, Langmuir 26 (2010) 98609874.
[29] J. Xie, J.Y. Lee, D.I.C. Wang, Y.P. Ting, Small 3 (2007) 672682.
Supplementary data associated with this article can be found, in
[30] P.N. Njoki, J. Luo, M.M. Kamundi, S. Lim, C.J. Zhong, Langmuir 26 (2010)
the online version, at http://dx.doi.org/10.1016/j.jcis.2013.06.027. 1362213629.
[31] J. Yang, B.J. McCoy, G.J. Madras, J. Chem. Phys. 122 (2005) 064901.
[32] L. Han, M.M. Maye, F.L. Leibowitz, N.K. Ly, C.J.J. Zhong, J. Mater. Chem. 11
References
(2001) 12581264.
[33] E.E. Finney, R.G. Finke, Chem. Mater. 21 (2009) 46924705.
[1] P.K. Jain, X. Huang, I.H. El-Sayed, M.A. El-Sayed, Plasmonics 2 (2007) 107118. [34] M. Yamamoto, Y. Kashiwagi, T. Sakata, H. Mori, M. Nakamoto, Chem. Mater. 17
[2] T. Pradeep, Thin Solid Films 517 (2009) 64416478. (2005) 53915393.
[3] E.C. Njagi, H. Huang, L. Stafford, H. Genuino, H.M. Galindo, J.B. Collins, G.E. [35] J.H. Lee, K. Kamada, N. Enomoto, J.J. Hojo, J. Colloid Interface Sci. 316 (2007)
Hoag, S.L. Suib, Langmuir 27 (2011) 264271. 887892.
[4] M. Sathishkumar, K. Sneha, I.S. Kwak, J. Mao, S. Tripathy, Y.S.J. Yun, J. Hazard. [36] J. Huang, Q. Li, D. Sun, Y. Lu, Y. Su, X. Yang, H. Wang, Y. Wang, W. Shao, N. He, J.
Mater. 171 (2009) 400404. Hong, C. Chen, Nanotechnology 18 (2007) 105104.
[5] V. Vinod, P. Saravanan, B. Sreedhar, D.K. Devi, R. Sashidhar, Colloids Surf., B 83 [37] A. Cornish-Bowden, Fundamentals of Enzyme Kinetics, 3rd ed.; Portland Press:
(2011) 291298. London, 2004; chapter 2.
[6] R. Sathyavathi, M.B. Krishna, S.V. Rao, R. Saritha, D.N. Rao, Adv. Sci. Lett. 3 [38] M. Watzky, R.J. Finke, J. Am. Chem. Soc. 119 (1997) 1038210400.
(2010) 138143. [39] D.J. Shaw, Introduction to Colloid and Surface Chemistry, Butterworth-
[7] R.R. Naik, S.J. Stringer, G. Agarwal, S.E. Jones, M.O. Stone, Nat. Mater. 1 (2002) Heinemann: Oxford, 1992; 11-14.
169172. [40] J.W. Mullin, Crystallization, Butterworth-Heinemann: Oxford, 2001; 192-193.
[8] S. Senapati, A. Ahmad, M.I. Khan, M. Sastry, R. Kumar, Small 1 (2005) 517520. [41] V.K.L. Mer, Ind. Eng. Chem. 44 (1952) 12701277.
[9] J.L. Gardea-Torresdey, E. Gomez, J.R. Peralta-Videa, J.G. Parsons, H. Troiani, M. [42] J.E. Millstone, S.J. Hurst, G.S. Mtraux, J.I. Cutler, C.A. Mirkin, Small 5 (2009)
Jose-Yacaman, Langmuir 19 (2003) 13571361. 646664.
[10] P. Mukherjee, A. Ahmad, D. Mandal, S. Senapati, S.R. Sainkar, M.I. Khan, R. [43] K. Ghule, A.V. Ghule, J.Y. Liu, Y.C.J. Ling, Nanosci. Nanotechnol. 6 (2006) 3746
Parishcha, P.V. Ajaykumar, M. Alam, R. Kumar, et al., Nano Lett. 1 (2001) 515 3751.
519. [44] S.S. Shankar, A. Rai, B. Ankamwar, A. Singh, A. Ahmad, M. Sastry, Nat. Mater. 3
[11] T.K. Sau, A.L. Rogach, Adv. Mater. 22 (2010) 17811804. (2004) 482488.
[12] Y. Sun, Y. Xia, Science 298 (2002) 21762179. [45] S.P. Chandran, M. Chaudhary, R. Pasricha, A. Ahmad, M. Sastry, Biotechnol.
[13] H. Itoh, K. Naka, Y.J. Chujo, J. Am. Chem. Soc. 126 (2004) 30263027. Progr. 22 (2006) 577583.
[14] J. Polte, T.T. Ahner, F. Delissen, S. Sokolov, F. Emmerling, A.F. Thnemann, R.J. [46] B. Ankamwar, M. Chaudhary, M. Sastry, Synth. React. Inorg. M 35 (2005) 19
Kraehnert, J. Am. Chem. Soc. 132 (2010) 12961301. 26.
[15] B. Abcassis, F. Testard, O. Spalla, P. Barboux, Nano Lett. 7 (2007) 17231727. [47] B. Liu, J. Xie, J. Lee, Y. Ting, J.P.J. Chen, J. Phy. Chem. B 109 (2005) 1525615263.
[16] R. Patakfalvi, Z. Virnyi, I. Dkny, Colloid Polym. Sci. 283 (2004) 299305. [48] C. Lofton, W. Sigmund, Adv. Funct. Mater. 15 (2005) 11971208.
[17] E.E. Finney, R.G.J. Finke, J. Colloid Interface Sci. 317 (2008) 351374. [49] J.C. Love, L.A. Estroff, J.K. Kriebel, R.G. Nuzzo, G.M. Whitsides, Chem. Rev. 105
[18] J. Parsons, V. Armendariz, M. Lopez, M. Jose-Yacaman, J.J. Gardea-Torresdey, J. (2005) 11031170.
Nanopart. Res. 12 (2010) 15791588.