Journal of Colloid and Interface Science 492 (2017) 191198

Contents lists available at ScienceDirect

Journal of Colloid and Interface Science

journal homepage: www.elsevier.com/locate/jcis

Regular Article

How does the size of gold nanoparticles depend on citrate to gold ratio
in Turkevich synthesis? Final answer to a debated question
Li Shi a,b, Eric Buhler a,b, Franois Bou b, Florent Carn a,
a
Laboratoire Matire et Systmes Complexes (MSC), UMR CNRS 7057, Btiment Condorcet, Universit Paris Diderot, 75013 Paris, France
b
Laboratoire Lon Brillouin (LLB), UMR 12 CEA-CNRS, CEA Saclay, 91191 Gif-sur-Yvette, France

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

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

Article history: The dependence between the size of gold nanoparticle (AuNP) and the citrate to gold molar ratio (X) is
Received 25 August 2016 still a matter of debate 65 years after the seminal work by Turkevich et al. for high X values. We assume
Revised 20 October 2016 that this dispersion of results is due to the variation of certain parameters that are often not mentioned in
Accepted 24 October 2016
the protocols, and to the use of a single characterization technique (dynamic light scattering (DLS) or
Available online 25 October 2016
transmission electron microscopy (TEM)). To adress definitely the question of this dependence, we have
synthesized AuNPs with very precise protocols ensuring that the only parameters to be modified are X
Keywords:
and the sequence of reagent addition. We have then studied, for the first time, the dependence of the size
Nanoparticle
Gold
with X quantitatively with a multimodal approach (UVVisible spectroscopy, DLS and TEM) for 2 syn-
Turkevich thetic routes differing only by the sequence of reagent addition. We show unambiguously that AuNPs
Multimodal characterization size decay monotonically, with X whatever the order of reagent addition. It allows us to exclude the
Dynamic light scattering occurrence of a measurable discontinuity for a peculiar value of X that prompted some authors to postu-
Transmission electron microscopy late the existence of two different reaction pathways when the citrate to gold molar ratio is around this
value. In contrast, our result is in line with one reaction pathway, likely a seed-mediated growth mech-
anism, which should leads to monotonic size decrease. Also, we note that our result agrees with the sole
available theoretical prediction (Kumar et al., 2007) on the whole range of X. Despite this apparent agree-
ment, we point some contradictions between recent experimental results and basal hypothesis of this
model.
2016 Elsevier Inc. All rights reserved.

Corresponding author.
E-mail address: florent.carn@univ-paris-diderot.fr (F. Carn).

http://dx.doi.org/10.1016/j.jcis.2016.10.065
0021-9797/ 2016 Elsevier Inc. All rights reserved.
192 L. Shi et al. / Journal of Colloid and Interface Science 492 (2017) 191198

1. Introduction (a) 100 Ref. [14] Conditions of synthesis

Ref. [15] T = 98 2 C
Gold nanoparticles (AuNPs) are the most widely used metallic Ref. [16]
[HAuCl4] = 0.25 0.05 mM
nanoparticles and drive a variety of applications in nanomedicine, Ref. [21]
sensing, optoelectronics and catalysis [15]. The fine control of Ref. [23]
AuNPs individual characteristics is a prerequisite to exploit the Ref. [26]

R (nm)
macroscopic properties, that emerge from collective effects [6]. 10
Although the syntheses in non-aqueous solvents were preferred
so far for the synthesis of high quality building blocks, great efforts
have been done in the last decade to optimize synthesis pathways
directly in water [7]. The prominent member of this group of aque-
'Kumar' model, Ref. [27]
ous synthesis is the Turkevich protocol introduced in 1951 [8].
1
This method enables to obtain monodisperse citrate-stabilized 0 5 10 15 20
AuNPs by simply changing the relative concentration of citrate
X
molecules that are quickly injected in a boiling HAuCl4 aqueous (b) 100

Ct 2-

t 3-
solution. In spite of its advantages, this approach suffers from lack

AuCl4- / HCt2-

/C
of reproducibility and of predictability due to a poor theoretical

/H

-
2
H) -

H)
understanding of the mechanism of formation. This is due to the

(O
(O
fact that citrate and gold ions exist both in different chemical

AuCl-4/ H3Ct

2
Cl
3
Cl

Au
forms, which can be reversibly converted into each other. This is

also due to the multiplicity of roles played by citrate molecules
(i.e. reductive agent, stabilizing agent and pH mediator), which
results in multiple intricate steps that are hard to probe separately.
However, one can attempt the following description. First, AuIII is
slowly reduced to AuI thanks to citrate decarboxylation into dicar-
boxy acetone (DCA) which leads to electron transfer. Differential
functional theory simulations have shown that the most favorable
reaction path is obtained at low pH where [AuCl4]
is the dominant
structure of aureate complex due to the presence of highly labile
Cl
ions around AuIII [10,19]. We wish to underline that pH is con-
trolled by citrate to gold molar ratio (X) at fixed gold concentration.
As formed AuI atoms can be assumed to form multi-molecular
complexes with DCA [8] but experimental proofs are lacking
[11,28]. Au0 atoms could be formed in bulk when the concentration
of AuI species increases locally to a high enough level ([AuCl2]
-
 10 nM [9]) to trigger homogeneous disproportionation [12]. Fur-
ther disproportionation may lead to larger aggregates of Au0 atoms
and when the aggregate size reaches a critical radius of order
1.5 nm, a stable nucleus of gold atoms should be formed [13,15].
Disproportionation can also occur at the particle surface, leading
to nucleus growth and also to the regeneration of some AuIII [20].
The reactant molar ratio, X, is the eldest and probably the most
frequent lever for controlling the size and the polydispersity of
AuNPs. Forty years ago, Frens mentioned a steep decrease of
AuNPs size by a factor 6 when X was varied from 0.8 to 2 at
[Au] = 0.25 mM. This result has been confirmed by several studies
[8,2127] together with a concomitant decrease of polydispersity.
It is also recognized that this evolution of size is restricted to low
values of X. Indeed, above a certain value, noted X in Fig. 1c, which
vary with the absolute gold concentration [23], the different results
are conflicting. This is shown in Fig. 1a which summarizes the
results obtained for the most popular conditions of synthesis:
[Au] = 0.25 0.05 mM and T = 98 2 C. One can discern two fam-
ilies of behaviors:
(i) On one hand, several studies [14,16,23] showed a discontin-
uous evolution of the size with a sharp minimum at X  3.5
[14,23]. Ji et al. [14] observed that this minimum corre-
sponds to a pH at which the dominant structures of aureate
complexes and citrate ions modify (Fig. 1b). These authors
propose that the observed discontinuity may be related to
a modification of the mechanism of formation (Fig. 1c). A
two-steps process involving nucleation followed by a slow
diffusion-controlled growth should dominate above X,
while a three-steps process involving fast nucleation and
Au
R (nm)
10
AuCl-4/ H2Ct-
1
1 2 3 4 5
3 4 5 6 7
(c) pH
Fig. 1. Evolution of the averaged radius of citrated AuNPs as a function of (a) the
molar ratio X and (b) the pH for similar conditions of synthesis given in Inset of
Fig. 1a and summarized in Table 1 of S.I. All radii are number averaged values
obtained by TEM except for [23] which corresponds to intensity averaged
hydrodynamic radii. The line corresponds to Kumar model prediction [27]. The
different colors of patterns and the numbers in Fig. 1b characterize the regions of
the speciation diagrams [19]. (c) Scheme representing the different results and the
proposed reaction pathways. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)
random binding, followed by an intra-particle ripening,
should dominate below X. This last mechanism may involve
transient nanowire-shaped clusters that have been
detected by TEM [17,22], but never by SAXS or in situ AFM
[15,18], suggesting that they result from the reduction of
AuI and AuIII species during drying [18].
(ii) On the other hand, several studies [22,26] showed that the
size decreases continuously with X on the whole range of
X, with lowest radius ranging between 2.5 nm [26] and
10 nm [15] at X = 20. A so-called four-steps seed-
mediated growth mechanism, which is supposed to be con-
gruent with a monotonic decrease of the size, was proposed
on the basis of coupled in situ XANES and SAXS during AuNPs
formation, for X > X [15]. Such proposal is based on the
observation that gold equilibrium is shifted from [AuCl4]

to less reactive [AuCl3
x(OH)1+x]
during the fast seed parti-
cle formation (30 s after mixing), so that for X P X, the
kinetics of seed formation and pH neutralisation should be
 L. Shi et al. / Journal of Colloid and Interface Science 492 (2017) 191198
approximately the same irrespective of the molar excess
thus explaining the size independence. For X < X, the molar
excess might still be good enough to reduce AuIII species but
not to shift the gold complex equilibrium. Hence, homoge-
neous and heterogeneous disproportionation can also occur
unselectively during the entire synthesis, increasing the
polydispersity. At very low X, seed particles should grow
up to a size that could be stabilized by the available citrate
molecules. Interestingly, the continuous evolution of size
with X has been theoretically predicted by Kumar et al.
[27] until R  7.5 nm at X = 20, but with a model neglecting
the dependence of the aureate complexes structure and
reactivity with the pH.
In summary, we note that there is no consensus on the evo-
lution of the size with X. We believe that this dispersion of
results is due (i) to the variation of certain parameters that
are often not mentioned in the protocols, and (ii) to the use
of a single characterization technique (TEM or DLS). In this arti-
cle, we want to answer experimentally at the following open
questions:
Is the evolution of AuNPs size with X monotonous or not?
What is the size of AuNPs for X P X? Indeed, all authors
agrees that the size is independent of X and the polydispersity
is low but the size vary on a decade between R = 2.5 and 19 nm.
What is the influence on our results of the sequence of addition
of the reagents?
What is the validity of the only theoretical prediction, as
checked in the light of our results?
We synthesized AuNPs with very precise protocols ensuring
that the only parameters to be modified are X and the sequence
of reagent addition. The as-obtained batches of AuNPs have
allowed us to give clear-cut answers by studying thoroughly
the size dependence as function of X on a large range of X val-
ues (i.e. 1 6 X 6 20) by combining UVVisible spectroscopy,
dynamic light scattering and transmission electron microscopy
for 2 synthetic approaches differing by the sequence of reagent
addition.
2. Experimental
2.1. Materials
Gold (III) chloride trihydrate (HAuCl43H2O, >99.99%), trisodium
citrate dihydrate (Na3C6H5O72H2O, P99%) were purchased from
Sigma-Aldrich and used as received without further purification.
All the content of a gold salt powder batch was used at the first
opening to prepare, using a glass spatula, a mother solution at
10 g/L in milliQ water that was stocked for period not exceeding
3 months in dark area to minimize photo-induced oxidation. The
same batch of trisodium citrate has been used for all synthesis; it
has been stored in desiccators after first opening. All glassware
and Teflon-coated magnetic bars were washed thoroughly with
freshly prepared aqua regia and rinsed with milliQ water after each
synthesis. All solutions were prepared with milliQ water
(R = 18.2 MX).
2.2. Synthesis of citrate stabilized AuNPs
Classical synthesis. We took the standard method initially
proposed by Turkevich et al. [8]. The gold (III) mother solution
was diluted with water to obtain 50 mL of a yellow orange solution
at [AuIII] = 0.25 mM in a 100 mL double-necked round flask. The
193
flask was then immersed in a temperature controlled oil bath at
96 3 C without reflux and was vigorously stirred at 900 rpm.
10 min after reaching the desired temperature, 2.5 mL of a citrate
solution (final X ranging between 1 and 20) preheated during
10 min at the reaction temperature were all added at once with a
thermalized tip. After 15 min under vigorous stirring at 96 C, we
switch off the stirring and the oil bath. The magnetic bar is
removed but not the oil bath until reaching 70 C (i.e. after
20 min). Then, the solution was allowed to cool down to the room
temperature. The time evolution of the oil bath and of the reactive
media temperatures measured during a typical synthesis are plot-
ted in Fig. 3 of S.I. to enable easy reproduction of our synthesis as
pointed in [29].
Inverse synthesis. We used the protocol proposed by Ojea-
Jimnez et al. [16]. The desired amount of Citrate was diluted in
a 250 mL double-necked round flask with water and stirred during
20 min at 25 C. The solution was then heated during 15 min at
100 C (Toil bath = 140 5 C) with reflux and vigorous stirring at
900 rpm. Then, 1 mL of mother precursor solution ([Au]
= 25 mM) preheated at 100 C during 10 min was injected all at
once with a thermalized tip into the citrate solution under the
same vigorous stirring. After 15 min at 100 C, we switched off
the stirring and the oil bath. The magnetic bar was removed but
not the oil bath until the solutions temperature had reached
70 C (i.e. after 40 min). Then, the solution was allowed to cool
down to the room temperature. We show in Fig. 1 of S.I. the evo-
lution of pH and electrical conductivity as a function of X for
AuNPs prepared by this method.
Each synthesis has been reproduced at least 3 times with the
same protocol, set-up and batches of chemicals.
2.3. Methods
Dynamic light scattering (DLS) experiments were carried out
using on a NanoZS apparatus (Malvern) operating at
k = 632.8 nm. The dependence of the characteristic relaxation
times upon the scattering angle (H) was determined with a 3D
DLS set-up (LS Instruments) operating at k = 632.8 nm in conven-
tional autocorrelation mode. The dispersions were never filtered
before measurements.
For dispersions characterized by two diffusive relaxations, the
normalized electric field autocorrelation function g(1)(q, t) were fit-
ted with the sum of two exponentials:
g 1 q; t Afast qe
t=sfast Aslow qe
t=sslow
where sfast and sslow represent the two cooperative characteristic
relaxation times, and Afast and Aslow are their corresponding ampli-
tudes with Afast q Aslow q 1: Electrophoretic mobility (l) was
also measured with a NanoZS. This set-up operates with an electri-
cal field of 25 V cm
1 oscillating successively at 20 Hz and 0.7 Hz to
reduce the electro-osmosis effect due to the surface charge of the
capillary cell. The particle velocity is measured by LASER Doppler
velocimetry.
Ultraviolet visible spectroscopy (UVvis.) was performed on
Cary 50 Scan UVVisible Spectrophotometer of brand Varian. The
diluted AuNPs solutions were poured into a 5 mm thickness
Hellma cell (quartz). The absorbance values were recorded after
baseline correction.
Transmission electron microscopy (TEM) experiments were
performed on a JEOL 2010 instrument operating at 200 kV. Sam-
ples were prepared by casting a single drop of a 1 g/L aqueous dis-
persion onto a standard carbon-coated Formvar films on copper
grids (200 mesh) and drying in air for >30 min. A minimum of
300 NPs per sample was considered for the size statistics except
at X = 1 for which 50 NPs were considered.
194 L. Shi et al. / Journal of Colloid and Interface Science 492 (2017) 191198

2.4. Model reported values at 100 C while the other rate constants were
adjusted to obtain the best fitting of the AuNP sizes, determined
In this section we briefly present the main elements of the with TEM by [8,21,24,25] variation with X in the range:
model proposed by Kumar et al. [27]. 0.4 6 X 6 7.5.
Main hypothesis. The model considers that: (i) the critical It is important to underline that these rate constants remain to
nucleus size is about 2 nm, (ii) all the AuNPs including those be measured experimentally so that they cannot be fixed a priori.
formed by aggregation are spherical, (iii) the repulsive interaction The initial gold concentration is fixed at [T]t=0 = 0.3 mM and the
between AuNPs is only due to double layer interaction, (iv) the rate initial citrate concentration [C]t=0 is varied to cover an X range
of AuNPs aggregation could be derived from the expression intro- between 0.4 and 40.
duced by Reerink and Overbeek [30] and (v) AuNPs aggregation is
irreversible.
3. Results
This model does not take into account: (vi) the possible disag-
gregation of intermediate aggregates leading to individual AuNPs,
AuNPs dispersions prepared by the Turkevich (i.e. classical)
(vii) the hydrothermal oxidation of citrate and (viii) the pH evolu-
and inverse methods were first characterized in bulk by
tion as a function of the molar ratio X and the inherent effect on the
electrochemical potential of the dominant aureate complexes in
water, (ix) the steric repulsion due to citrate adsorption on the (a) 0.75 X=1
AuNPs surface but the author balances this by overestimating the X=2
X=3
AuNPs surface potential on purpose and (x) the influence on the X = 3.5
stability of citrate oxidation product (DCA) as pointed out by Ref. X=4
[28]. The value of the AuNPs surface potential is hence determined 0.50 X=6

Abs. (a.u.)
X=8
by an empirical relation giving values always higher than the X = 10
90 mV derived from electrophoretic mobility measurements per- X = 13
X = 20
formed at X = 3 by Chow and Zukoski [13]. We show in Fig. 2 of
the S.I. that our AuNPs have almost the same electrophoretic 0.25
mobility than in the Ref. [13] in the same range of X.
Scheme of reactions. The model considers that the formation of
citrated AuNPs result from the following reactions involving auric 400 600 800
and aurous species (noted T and M respectively); citrate, dicar-
boxyacetone and acetone (noted C, DCA and A respectively). (nm)
(b) 0.75 X=1
(1) Homogeneous reduction auric ions by citrate: X=2
X=3
X = 3.5
k1
T C ! M DCA products X=4
0.50 X=6
Abs. (a.u.)

X=8
(2) Degradation of dicarboxyacetone: X = 10
k2
H2 O DCA ! A products X = 13
X = 20

This reaction corresponds to the hydrothermal oxidation of 0.25

dicarboxyacetone.
(3) Homogeneous reduction by acetone:
k3
4T A ! 4M products 400 600 800
Note that the stoichiometric ratio of this reaction has been (nm)
assigned to obtain a complete conversion of auric chloride at a sto-
ichiometric ratio of 0.4 as observed by Frens [21].
(c) 560 7.5
SPR, Classical method
k4 SPR, Inverse method
(4) Nucleation: 3M 2DCA ! T Nucleus 2DCA
<R>Int., Classical method
(5) Heterogeneous disproportionation:
<R>Int. (nm)

<R>Int., Inverse method

SPR (nm)

540 5.0
k5
3M ! T particle mass
The rate of this reaction is assumed to be proportional to the
AuNP surface area since the gold surface catalyzes this reaction.
The number of gold particles in a given size range R to R + dR hence 520
appears in the mass balance of aurous and auric species. 2.5
The number of particles in a size range as a function of time is 0 5 10 15 20
obtained by solving the population balance equation using a dis-
cretization technique described by the same author in [31]. The
X
population balance equation is obtained by considering that the Fig. 2. Absorbance (Abs.) versus wave length (k) at 25 C for AuNPs dispersions
apparition and disappearance of AuNPs in a given interval of sizes synthesized by (a) classical and (b) inverse methods for different ratio X. (c)
results from growth process due to surface reaction (3) and Brow- Evolution of the wave length at the surface plasmon resonance (kSPR, full symbol)
nian aggregation while new nuclei are formed in continuous (reac- and AuNPs intensity averaged radius (hRiInt., open symbol), determined with Haiss
relation [32], for classical (red circle) and inverse (black square) methods. The
tions (1), (3) and (4)) until the concentration in DCA is insufficient dashed line is an exponential decay and must serve as guide for the eyes. (For
due to the degradation process (2). The rate constant k1 was fixed interpretation of the references to colour in this figure legend, the reader is referred
to make the process time of the same order as the experimentally to the web version of this article.)
 L. Shi et al. / Journal of Colloid and Interface Science 492 (2017) 191198 195

UVvis. spectroscopy as shown in Fig. 2. Whatever the method of
preparation, one observes a significant narrowing of the surface
plasmon resonance (i.e. SPR) peak accompanied by a blue shift of
SPR maximum from 546 nm (direct synthesis)/555 nm (inverse
synthesis) down to 519520 nm when X increases from 1 to 3.5
(Fig. 2c). Then, the SPR peak does not evolve up to X = 8. Above this
value a slight red shift can be detected up to 522523 nm at X = 20.
This evolution agrees qualitatively with the picture of a transition
from large and polydisperse AuNPs at X < 3.5, to small and quite
monodisperse AuNPs at X P 3.5 until a limit at high X values above
which aggregation may occurs. We tried to derive the intensity
average AuNPs size from these absorbance data using the follow-
ing relation proposed by Haiss et al. [32]:
 
A
1 3A SPR
2:2
hRiInt: e 450
2 goes from 2 to 3.5 and no longer varies beyond. At high X ratio,
where ASPR and A450 are the absorbance at the SPR peak and 450 nm
respectively, while the factors 3 and 2.2 were empirically deter-
mined by the authors. The as-determined radii evolution plotted
in Fig. 2c give rise to a rather inaccurate result due to points notice-
ably scattered in the whole range of X. It appears that the Haiss
relation is too much sensitive to uncorrelated variations of the
absorbance and does not enable to extract in a quantitative manner
the evolution of the AuNPs sizes.
To better quantify the size distributions in bulk we performed
DLS measurements. The normalized electric field autocorrelation
functions, g(1) (173, s), measured one day after the synthesis are
plotted in Fig. 3. For the two synthetic routes, the curves depict
two relaxations for X = 1 and one relaxation for: 1 < X < 10 with
the classical method or 1 < X < 20 with the inverse method. In
these intermediate ranges, the relaxation times decrease when X

Fig. 3. Normalized electric field autocorrelation functions, g(1) (173, s), obtained at 25 C, one day after (a) classical or (b) inverse synthesis with different ratio X indicated
on the plot. (c) Intensity weighted distributions of RH, App. for AuNPs prepared by classical (red disc) and inverse methods (black square). (For interpretation of the references
to colour in this figure legend, the reader is referred to the web version of this article.)
196 L. Shi et al. / Journal of Colloid and Interface Science 492 (2017) 191198

these relaxations are stretched or associated with a slow mode of aggregated AuNPs while Cindiv. and Caggregates are respectively the
low amplitude (i.e. X = 13 or X = 20 for classical and inverse concentrations, in g cm
3, of individual and aggregated AuNPs.
methods respectively). This analysis suggests again the formation Hence, we obtain:
 3
of particles of smaller and smaller size until X = 3 after what the C indiv : A R
C aggregates
 A fast H;aggregates
R
 5  104 for X = 13 in the classical
size does not evolve significantly except at high X values where slow H;indiv :

aggregation may appears. method and C indiv :

C aggregates
 5  105 for X = 20 in the inverse method.
Accordingly, the intensity weighted distributions of apparent The number weighted size distributions have been then deter-
hydrodynamic radius (RH, App.) obtained with the Contin algorithm mined for the two sets of synthesis by means of TEM (Fig. 4).
(Fig. 3c) reveal a transition from large and polydisperse AuNPs at
X 6 3, to small and quite monodisperse AuNPs at X P 3.5, with: 4. Discussion

hRH, app.i3.5 6 X 6 20 = 9.9 1.2 nm for the classical method, and Fig. 5 summarizes the results obtained during this study. It
hRH, app.i3.5 6 X 6 20 = 10.6 1.5 nm for the inverse method. appears that the intensity averaged radii determined by DLS and
TEM decrease in a monotonous manner when X increases for the
Interestingly, at high X values, the presence of a second popula- two batches of AuNPs. We consider it as a good support for reliabil-
tion of large objects with RH, app > 100 nm is not associated with a ity of the measurements. The size evolution is sharp at low X and
modification of the small objects mean size. In spite of their big not significant above X = 3.5 with hRiint. = 9.5 1.5 nm. This
size (hence large individual scattering) the large objects detected prompts us to a more detailed focus on the slight increase of kSPR
at X = 13 and X = 20 for classical and inverse methods respec- (Fig. 5) for X P 3.5, and to consider that it should not be assigned
tively are always minority in% of intensity meanwhile they repre- to an increase of AuNP size. It appears that the intensity averaged
sent a low number fraction compared to the one of individual NPs. radii determined by DLS and TEM decrease in a monotonous man-
This can be quantified by considering the amplitudes Afast and Aslow ner when X increases for the two batches of AuNPs. We consider it
associated respectively to the fast and the slow mode (see E.S.). as a good support for reliability of the measurements. The size evo-
Indeed, defining Iindiv. and Iaggregates as the time-averaged intensi- lution is sharp at low X and not significant above X = 3.5 with
ties associated with the fluctuations of individual and aggregated hRiint. = 9.5 1.5 nm. This prompts us to a more detailed focus on
AuNPs concentrations respectively, and neglecting the virial the slight increase of kSPR (Fig. 5) for X P 3.5, and to consider that
effects, we obtain: it should not be assigned to an increase of AuNP size.
 3 The comparison between our measurements and previous,
Afast Iindiv : K  C indiv : M w;indiv : C indiv : RH;indiv :
   experimental [1416,21,23,26] and theoretical [27], data for com-
Aslow Iaggregates K  C aggregates M w;aggregates C aggregates RH;aggregates
parable conditions (i.e. [HAuCl4] = 0.25 0.05 mM and
T = 98 2 C) is shown in Fig. 6.
where K is the scattering constant, Mw,indiv. and Mw,aggregates are
One observes that above X = 3 our results only agree, within
respectively the molecular mass, in g mole
1, of individual and
experimental errors, with the two points extracted from [15]. Note
that the monotonic character of the size decay agrees with [26]
15
X=1 15 X=1 even if the measured size for X P 3 is clearly lower. This size dis-
crepancy for a priori similar conditions of synthesis is hard to ana-
0 0 lyze with available experimental details. One can only underline
15
X=2 15
X=2
that parameters often ignored in the experimental procedure, like
0 0 the effect of water latent heat, could act on the final size [29]. Also,
0 30 60 90 0 30 60 90 our results quantitatively agree with the theoretical prediction
proposed by Kumar et al. on the whole X range in contrast with
30 X=3 30 X=3 all other experimental results available so far. Now, let us consider
0 0 how our results enlighten the ongoing discussions about the

30 X=4 30 X=4
0 0
% of number

X=6 X=6 Classical Method

< R >int. (nm)

30 30 50 <RTEM>Intensity
SPR (nm)

0 0 <RH, App.>Intensity 540

30 X=8 30 X=8 25 SPR
0 0 520
30 X = 10 30 X = 10 0
Inverse Method
< R >int. (nm)

0 0 50
X = 13 X = 13
SPR (nm)

30 30 540
25
0 0
30 X = 20 30 X = 20 520
0
0 0 0 5 10 15 20
0 10 20 30 0 10 20 30
X
R (nm) R (nm)
Fig. 5. Evolution of the kSPR and the intensity average radii determined by DLS and
Fig. 4. Number weighted size distribution determined by TEM observations of TEM, hRiint., as function of X for the two methods of synthesis as indicated in the
AuNPs prepared by the traditional method (left column) and the inverse (right figure. For the two plots, the dashed line is the same exponential decay function (i.e.
column) at different X ratio indicated on the plot. Representative TEM images taken RXInt: 176  e
x=0:8 8:8) and must serve as guide for the eyes. The intensity
with the same magnification are shown as inset for each histogram. The width of averaged radii corresponding to TEM observations have been determined as
the images corresponds to 560 nm. described in [33].
 L. Shi et al. / Journal of Colloid and Interface Science 492 (2017) 191198
100
Ref. [14] Kumar model, no coagulation
Ref. [15] This work, classical method
Ref. [16] This work, 'inverse' method
Ref. [21]
<R>Number (nm)
Ref. [23]
Ref. [26]
10
1 note that our result agrees with the sole available theoretical pre-
0 5 10 15 20
X ment, we point some contradictions between recent experimental
Fig. 6. Evolution of the average radii of AuNPs prepared in similar conditions of
synthesis summarized in Table 1 of S.I., as function of the molar ratio X. All the radii
are number averaged values obtained by TEM except for [23] for which we
represent RH, App. The line corresponds to Kumar model prediction without
aggregation [27].
mechanism of formation. Our study allows us to confirm the
monotonic decay observed in [26] and to exclude the existence
of a measurable discontinuity (i.e. X in Fig. 1c) that motivated to
postulate the existence of two different reaction pathways around
X. The observed monotonic evolution has been predicted by
Kumar et al. [27] and the seed-mediated growth mechanism pro-
posed on the basis of measurements performed at X > X (i.e.
X = 6.7 and 10), is expected to be congruent whatever X with this
theoretical prediction [15]. For the first time we obtain experimen-
tal results in line with these assumptions, since their results at
X = 6.7 and 10 superimpose with our data, which, beyond these
two points, reveal a monotonic decrease of size that superimpose
with the theoretical prediction on the whole range of X. Despite
this observation, we prove neither the validity of the Kumar
model, nor the validity of the seed-mediated growth mechanism
on the whole range of X. However, we could mention that some
basal assumptions of Kumar model appear poorly compatibles
with the seed-mediated growth mechanism and with recent
experimental findings. For instance, the hydrothermal oxidation
of citrate occurring before AuIII addition, the effect of citrate con-
centration on pH and the correlated effect on the electrochemical
potential of the dominant aureate complexes in water are not
taken into account. Nonetheless, the rate constants of the reactions
that are taken into account are free parameters of the fitting pro-
cedure since their values remain to be measured. For these reasons,
we believe that the observed agreement with our experimental
results is possibly fortunate. We hope that our study will serve
as a new reference for the evolution of the AuNPs size with X
and that further theoretical studies will complement the beautiful
seminal work done by Kumar et al. by taking into account the
above mentioned experimental findings of the last decade.
5. Conclusions
The effect of citrate to gold molar ratio (X) on the size of citrated
gold nanoparticles (AuNPs) has been studied experimentally for
the most popular condition of synthesis (i.e. [Au]
= 0.25 0.05 mM and T = 98 2 C) by using, for the first time, a
multimodal approach of characterization (UVvis spectroscopy,
dynamic light scattering and transmission electronic microscopy).
This dependence is still a matter of debate at high X values
[14,16,23,26,27]. We show unambiguously that the size of AuNPs
decay monotonically when the molar ratio X = [Citrate]t=0/[AuIII]t=0
increases between 1 and 20. The same evolution and the same
197
sizes have been obtained whatever the order of reagent addition.
We believe that the precise protocols of synthesis which has been
established has allowed to minimize the standard deviation
between experiments and has ensured that the only parameters
to be modified are X and the sequence of reagent addition. For
intermediate X values, our finding allows us to exclude the exis-
tence of a measurable discontinuity for a peculiar X value [14,23]
that prompted authors [14] to postulate the existence of two dif-
ferent reaction pathways when the citrate to gold molar ratio is
around this value. Our result is in line with the continuous evolu-
tion reported by Majzik et al. [26] and one reaction pathways
which should lead to monotonic size decrease [15,27]. Also, we
diction [27] on the whole range of X. Despite this apparent agree-
results and basal hypothesis of this model. For instance, the varia-
tion of pH with X and the inherent effect on the electrochemical
potential of the dominant aureate complexes in water [10] are
not taken into account. For high X values, we confirm the presence
of a plateau at hRinumber  9 nm in contrast with the very low size
reported by Majzik et al. [26]. We hope that our study will serve as
reference for the evolution of the AuNPs size with X and that fur-
ther theoretical studies could rely on it to complement the beauti-
ful seminal work done by Kumar et al. [27]. This article may also
show to the broad readership in colloid and interface science that
the rationalization of metallic NPs synthesis is an actual subject
which necessitates the methodology and the technical knowledge
developed in this community to develop theoretical prediction
tools and new strategies of synthesis based on a profound mecha-
nistic knowledge rather than on imagination.
Acknowledgements
L.S., E.B. and F.C. are indebted to the French Labex SEAM
(Sciences and Engineering for Advanced Materials and devices)
supported by the Commissariat Gnral lInvestissement. The
authors acknowledge the ImagoSeine core facility of the Institut
Jacques Monod, also associated with IBiSA and France BioImaging
infrastructures.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcis.2016.10.065.
References
[1] S.E. Lohse, C.J. Murphy, J. Am. Chem. Soc. 134 (2012) 1560715620.
[2] O.S. Muddineti, B. Ghosh, S. Biswas, Int. J. Pharm. 484 (2015) 252269.
[3] S. Zeng, D. Baillargeat, H.-P. Ho, K.-T. Yong, Chem. Soc. Rev. 43 (2014) 3426
3452.
[4] L. De Sio, T. Placido, R. Comparelli, M.L. Curri, M. Striccoli, N. Tabiryan, T.J.
Bunning, Prog. Quantum Electron. 41 (2015) 2370.
[5] C. Wang, D. Astruc, Chem. Soc. Rev. 43 (2014) 71887216.
[6] N.J. Halas, S. Lal, W.-S. Chang, S. Link, P. Nordlander, Chem. Rev. 111 (2011)
39133961.
[7] P. Zhao, N. Li, D. Astruc, Coord. Chem. Rev. 257 (2013) 638665.
[8] J. Turkevich, P.C. Stevenson, J. Hillier, Discuss. Faraday Soc. 11 (1951) 5575.
[9] B. Rodrguez-Gonzlez, P. Mulvaney, L.M. Liz-Marzn, Z. Phys. Chem. 221
(2007) 415426.
[10] I. Ojea-Jimnez, J.M. Campanera, J. Phys. Chem. C 116 (2012) 2368223691.
[11] M. Doyen, K. Bartik, G. Bruylants, J. Coll. Int. Sci. 399 (2013) 15.
[12] According to IUPAC Recommendations, (i.e. PAC 1994, 66, 577), a reaction of
disproportionation is a reversible or irreversible transition in which species
with the same oxidation state combine to yield one of higher oxidation state
and one of lower oxidation state (for instance: 3 Au+ ? Au3+ + 2 Au0).
[13] M.K. Chow, C.F. Zukoski, J. Coll. Int. Sci. 165 (1994) 97109.
[14] X. Ji, X. Song, J. Li, Y. Bai, W. Yang, X. Peng, J. Am. Chem. Soc. 129 (2007) 13939
13948.
[15] M. Wuithschick, A. Birnbaum, S. Witte, M. Sztucki, U. Vainio, N. Pinna, K.
Rademann, F. Emmerling, R. Kraehnert, J. Polte, ACS Nano 9 (2015) 70527071.
198 L. Shi et al. / Journal of Colloid and Interface Science 492 (2017) 191198

[16] I. Ojea-Jimnez, N.G. Basts, V. Puntes, J. Phys. Chem. C 115 (2011) 15752 [24] P.L. Freund, M. Spiro, J. Phys. Chem. 89 (1985) 10741077.
15757. [25] J.-P. Abid, Ph.D. Thesis, Ecole Polytechnique Federale de Lausanne, 2003.
[17] B.-K. Pong, H.I. Elim, J.-X. Chong, W. Ji, B.L. Trout, J.-Y. Lee, J. Phys. Chem. C 111 [26] A. Majzik, R. Patakfalvi, V. Hornok, I. Dekany, Gold Bull. 42 (2009) 113123.
(2007) 62816287. [27] S. Kumar, K.S. Gandhi, R. Kumar, Ind. Eng. Chem. Res. 46 (2007) 31283136.
[18] Y. Mikhlin, M. Likhatski, A. Karacharov, V. Zaikovski, A. Krylov, Phys. Chem. [28] D. Grasseschi, R.A. Ando, H.E. Toma, V.M. Zamarion, RSC Adv. 5 (2015) 5716
Chem. Phys. 11 (2009) 54455454. 5724.
[19] D. Goia, E. Matijevic, Colloids Surf. A 146 (1999) 139152. [29] W. Ding, P. Zhang, Y. Li, H. Xia, D. Wang, X. Tao, ChemPhysChem 16 (2015)
[20] J.-W. Park, J.S. Shumaker-Parry, J. Am. Chem. Soc. 136 (2014) 19071921. 447454.
[21] G. Frens, Nature 241 (1973) 2022. [30] H. Reerink, J.T.G. Overbeek, Discuss. Faraday Soc. 18 (1954) 7484.
[22] J. Kimling, M. Maier, B. Okenve, V. Kotaidis, H. Ballot, A. Plech, J. Phys. Chem. B [31] S. Kumar, D. Ramkrishna, Chem. Eng. Sci. 52 (1997) 46594679.
110 (2006) 1570015707. [32] W. Haiss, N.T.K. Thanh, J. Aveyard, D.G. Fernig, Anal. Chem. 79 (2007) 4215
[23] K. Zabetakis, W.E. Ghann, S. Kumar, M.-C. Daniel, Gold Bull. 45 (2012) 4221.
203211. [33] L. Shi, E. Buhler, F. Bou, F. Carn, Langmuir 31 (2015) 57315737.