Simple size-controlled synthesis of Au nanoparticles and their size-dependent catalytic activity

The controlled preparation of Au nanoparticles (NPs) in the size range of 6 to 22 nm is explored in this study. The Au NPs were prepared by the reduction of tetrachloroauric acid using maltose in the presence of nonionic surfactant Tween 80 at various concentrations to control the size of the resulting Au NPs. With increasing concentration of Tween 80 a decrease in the size of produced Au NPs was observed, along with a significant decrease in their size distribution. The size-dependent catalytic activity of the synthesized Au NPs was tested in the reduction of 4-nitrophenol with sodium borohydride, resulting in increasing catalytic activity with decreasing size of the prepared nanoparticles. Eley-Rideal catalytic mechanism emerges as the more probable, in contrary to the Langmuir-Hinshelwood mechanism reported for other noble metal nanocatalysts.

approach confirmed by experimental tests, that geometric and electronic differences between gold clusters comprising a few atoms and gold nanoparticles of 1 nm or larger determine their activity and selectivity. Typically, small planar clusters are highly active for reactions involving activation of the C-C multiple bonds in alkenes and alkynes by means of Lewis acid-base interactions, and on the other hand 3D gold nanoparticles have better catalytic performance in redox reactions involving bond dissociation by oxidative addition and new bond formation by reductive elimination 35 .
According to the mentioned dependence of catalytic activity of different sized Au NPs, it is necessary to find robust and reliable procedure for the size-selective preparation of Au NPs, moreover with narrow size distribution. Au NPs are prepared commonly by the several basic methods, from which Turkevich's method producing 15 nm sized nanoparticles using citrate as reductant 36 or Martin's method producing 3-5 nm sized nanoparticles using NaBH 4 as reductant 37 are mostly used in practice. Many other procedures producing single sized spherical Au NPs, e.g. 12 nm 38 , 13 nm 39 or 21 nm 40 were published previously, but the most of these methods do not make possible size managed preparation of Au NPs. Brinas et al. introduced size-controllable preparation method of Au NPs with diameter from 2 up to 6 nm by varying the pH of the reaction system 41 , which procedure was recently modified to produce nanoparticles over larger size range 42 . Ohyama et al. prepared sulfur-based ligand protected Au NPs in the same size range as Brinas. The size of Au NPs was in this case tuned by adjusting the thiol/HAuCl 4 ratio 43 . Song et al. published the synthesis of Au NPs with a diameter between 1 and 6 nm, where the size control was achieved by changing the fraction of CHCl 3 in mixture with benzene 44 . Jun et al. used milli-and micro-fluidic flow system for the preparation of Au NPs with sizes ranging from 7 to 25 nm. Ascorbic acid was used as reducing agent in this procedure and particle size was tuned by modifying the ascorbic acid vs. gold salt concentration ratio and the flow rate of the solutions 45 . Akamatsu et al. synthesized nanocomposite microgels with size controlled from 10 up to 30 nm using dimethylamineborane as reducing agent. In this case, the size of prepared Au NPs was controlled by varying the reduction rate 46 . Two step seeded growth method which does not involve application of any surface modifier (except of citrate) during synthetic procedure enabling preparation of size managed gold particles in the size range from 15 up to 300 nm was reported recently 47 . However, all the above mentioned procedures for preparation of Au NPs are rather complicated and do not enable the preparation of larger amounts of nanoparticles necessary for application of nanoparticles in catalysis beyond the laboratory scale.
This work presents a simple and reproducible procedure for one-step preparation of Au NPs that provides a straightforward control of the size of the produced nanoparticles in a size range from 6 up to 22 nm, where the final size of the Au NPs is controlled by varying the concentration of nonionic surfactant Tween 80 (polyethylene glycol sorbitan monooleate) in the reaction mixture. This method of size control allows the realization of a size dependence study on catalytic activity of Au NPs and evaluation of reaction mechanism of the used model reaction. The synthesis procedure is carried out in open vessels, which points to the possibility of large scales essential for application in real practice.

Results and Discussion
Synthesis of various-sized Au NPs. Application of variable concentrations of nonionic surfactant Tween 80 in the reaction mixture led into significant changes in the size distribution of the resulting gold nanoparticle dispersions. Figure 1 shows the size distribution (obtained by DLS) of synthesized Au NPs as a function of Tween 80 concentration.
According to DLS measurements, increasing concentration of the Tween 80 in the reaction mixture in the range from 0.1 up to 10 mmol/L results in formation of Au NPs with decreasing average diameter from around 80 nm to 10 nm. The observed size differences between samples prepared with various concentration of Tween 80 were qualitatively confirmed by collecting UV-Vis spectra (Fig. 2). Except the non-modified system, the UV-Vis absorption maximum of Au NPs' surface plasmon is shifted to the lower wavelengths with increasing concentration of Tween 80. This phenomenon reflects the changes of optical properties of Au NPs caused by decreasing particle diameter. Namely, by assuming spherical gold nanoparticles, with decreasing size of the particles the spectrum undergoes a blue shift, while the extinction coefficient drops as well. This explains the shift of the absorption maximum located at 538 nm for 80 nm nanoparticles to 515 nm for the smallest 10 nm nanoparticles, as well as the drop in absorption with decreasing of the size of the nanoparticles 48,49 . There is an apparent discrepancy with the position of the absorption maximum in the case of non-modified system, which could arise from the multimodal particle size distribution with broader size-and shape particle distributions than found for Tween-treated samples (see Table 1). Also some contribution of Tween 80 to the change of the dielectric constant of surrounding media can cause slight red shift of surface plasmon absorption maxima of Tween modified nanoparticles. Although this effect will be subtle (the diffractive index of Tween 80 is about 1.473 in comparison with value for water 1.330) the observed shift between surface plasmon maxima for unmodified and Tween modified may be influenced also by this phenomenon 50,51 . However, based on the available data, an unambiguous determination of the cause of this anomaly cannot be made. The size and morphology of the prepared Au NPs was further studied by transmission electron microscopy (TEM) and typical images are reported on Fig. 3. As it can be seen, in the case of the Au NP dispersions prepared with 0, 0.01 and 0.1 mmol/L of Tween 80 aggregation of the prepared Au NPs takes place, while particle size distribution decreases with the increasing concentration of Tween 80 as also shown by DLS. Moreover, in the case of experiment conducted without Tween 80 a multimodal distribution of the emerged gold colloid is evident with a dominating fraction of very small (<4 nm) particles (Fig. 3). Consequently, this system exhibits maximum of surface plasmon absorbance at a shorter wavelength compared to dispersions of nanoparticles prepared at low concentrations of Tween 80 (0.01 and 0.1 mmol/L) which do not contain such small nanoparticles.
Similar observation of size dependency of the surface plasmon resonance wavelength were observed for other metal nanoparticles, e.g. Ag nanoparticles in He matrix 52 .
Based on the histograms presented in Fig. 3, the mean diameter (d ) and corrected sample standard deviations (s) were calculated as follows: The results obtained from these calculations are collected in Table 1 along with the Au(III) to Au(0) reduction half-times determined from kinetic curves (Fig. 4). We note that in the case of synthesis without added Tween 80 it was not possible to calculate the corrected standard deviation and the polydispersity index due to a bi/trimodal distribution of the resulting Au NPs. The shortest half-time observed for this sample can be understood due to the absence of the surfactant which may block the active sites of the surface of the formed nuclei and therefore the reaction can proceed without any obstacles. Besides the size control of the Au NPs by the presence of the Tween 80 a significant reduction of polydispersity was observed with decreasing Au NP size: the polydispersity index of Au NPs decreased from 0.073 to 0.026 with increasing concentration of Tween 80 from 0.01 mmol/L to 10 mmol/L.
We assume that the dependence of size and polydispersity of the prepared colloids on the Tween 80 concentration reflects the mode of interaction of Tween 80 with gold species present in the solution, with dominant ion-dipole interaction and strong coordination interaction of gold with hydroxyl groups and oxygens of Tween 80 53,54 . Higher concentrations of Tween 80 enable interaction with a greater fraction of growing gold species which are formed during the reduction process. The surface of these species is partially or fully blocked by adsorbed surfactant molecules. Due to formation of this shielding barrier the catalytic reduction of Au(III) ions on the surface of existing nuclei is slowed down by adsorbed surfactant layer and much slower homogeneous mechanism of formation of new nuclei get the same significance in this case. Also, coalescence of the emerging nuclei and primarily formed very small nanoparticles which is an important step in the growth of the nanoparticles to large crystallites 55 was suppressed by this surface layer of Tween 80. As a result, the process can lead to the final formation of a greater number of Au NPs with smaller diameter. Blocking the surface by Tween 80 and the slowdown of particle growth is also reflected in the growing delay in the onset of the changes in the UV-Vis spectra with the reaction time (kinetic curves) recorded at the wavelength characteristic for the final particle size at a given concentration of Tween 80 (Fig. 4).
The evolution of the kinetic curves indicate that the formation of Au NPs likely proceeds in several phases 55 . In the first phase, through two interconnected steps, a formation of small Au nuclei (clusters of several atoms of Au) takes place through the reduction of the precursor and these nuclei quickly aggregate to form very small nanoparticles. This first phase of the reaction is reflected in a slow increase (nearly negligible in the non-modified system) of the absorption maxima signal measured at λ max , thus the optimum wavelength for the final nanoparticle size. In the second phase of the reaction, the primarily formed nanoparticles are sufficiently large for autocatalytic growth, and therefore they quickly grow almost to their final size, which is reflected by the rapid increase of the absorption signal. In the last reaction phase a slow recrystallization of formed nanoparticles takes place depending on their size. This reaction phase is reflected by very slow and nearly negligible growth of the kinetic curve. According to Fig. 4, in the cases of non-modified system and system modified with the lowest concentration of Tween 80 (0.01 mmol/L), the surface of the forming new nuclei is likely not saturated with the molecules of the surfactant and therefore a fast growth to the final size Au NPs takes place. On the other hand, in the cases of the two highest concentrations of the surfactant (1 mmol/L and 10 mmol/L), the surface of the new nuclei is expected to be highly covered/saturated with Tween 80 which slows down subsequent particle growth in the next reaction steps. The kinetic curve of the system modified by 0.1 mmol/L Tween 80 seems to represent a transition between the above described two extremes.
The difference in the reaction rate between first and consecutive phases becomes smaller with increasing concentration of Tween 80, which is reflected by the changes in slope of the recorded kinetic curves. As principal measurable characteristic of changes in kinetics of the studied reaction could be used the reaction half time which is commonly determined as point of inflexion of the kinetic curve. The obtained dependence of the half time on the concentration of Tween 80 in the reaction system (Table 1, Supplementary Fig. S1) clearly demonstrates the essential influence of Tween 80 on the reaction course. Substantial deceleration of reduction process by the existence of inhibiting layer of used surfactant on the surface emerging Au NPs is then reflected in decrease of their average size. Formation of new nuclei is preferable in the case when surface of existing particles is blocked by the layer of adsorbed molecules of surfactant. This phenomenon results in the observed decrease of final size of the prepared nanoparticles.
Catalytic activity of various-sized Au NPs. The catalytic activity of the prepared Au NPs was studied using a model reaction based on the reduction of 4-nitrophenol to 4-aminophenol by sodium borohydride. The advantage of this reaction is that its reaction course can be monitored using UV-Vis spectroscopy ( Fig. 5 and Supplementary Fig. S2-S5).
Heterogeneous catalytic reactions can be in general described by two kinetic models, namely by Langmuir-Hinshelwood mechanism or Eley-Rideal mechanism, which differ in the adsorption processes of the reactants on surface of catalyst. In the case of the Langmuir-Hinshelwood mechanism, the reaction between two molecules takes place on the catalyst's surface after adsorption of both reactants onto the active area of catalyst's surface 56 . In the case of the Eley-Rideal mechanism only one of the reactant molecules is adsorbed on the surface prior to its reaction with the second molecule which collides with the complex without coadsorbing first 57 . The catalytic reaction according to Langmuir-Hinshelwood model consists of several steps. The first one includes NaBH 4 reaction on the nanoparticle surface and forming surface-hydrogen species 24,30 . Second step is connected with 4-nitrophenol diffusion from solution and adsorption on the metal surface. Subsequently 4-nitrophenol molecule is reduced by surface-hydrogen species produced by BH 4 − hydrolysis and final reaction step include desorption of reduction products from the surface and their diffusion away from the surface of the catalyst. If diffusion, adsorption and desorption processes are reversible and fast, than the reduction of 4-nitrophenol molecule becomes the rate-determining step 58 . Hence the reaction can be described by Equation 4 (Langmuir-Hinshelwood mechanism), where the catalytic reduction takes place on the surface of gold nanoparticle 57 where k is the rate constant of the reaction, S is the surface of the catalyst, K A and K B represents adsorption coefficients and C A and C B are the concentrations of reactants in a reaction mixture ("A" index belongs to 4-nitrophenol, "B" index belongs to NaBH 4 ). Because of very low concentration of 4-nitrophenol compared to sodium borohydride (K A C A ≪ K B C B ), it is possible to simplify the kinetic equation (4) to the following form Substitution of all constants except C A into an overall constant k app(L-H) leads to the well-known first order kinetic equation 56 In the case of the Eley-Rideal mechanism, only one of two reactants is adsorbed onto metal surface which then reacts with the second reactants coming from solution without adsorbing on the surface of the catalyst. Applying the same approach as used above for the Langmuir-Hinshelwood model, the reaction according to Eley-Rideal mechanism can be described by equation (7) Substitution of all constants except C A into an overall constant k app (E-R) and using presumption K A C A «1 (very low concentration of reactant A) leads again to the first order kinetic equation As a result, in both considered cases, the reaction course of 4-nitrophenol reduction can be evaluated as the relative absorbance change (A t /A 0 , were A t and A 0 denotes absorbance at time t and at beginning respectively) at 400 nm (absorption maximum of 4-nitrophenol) with time. The apparent rate constant was determined from the slope of the linear correlation of ln(A t /A 0 ) with time according to Wunder et al. 60 .
The obtained kinetic curves (Fig. 6), show the rate of 4-nitrophenol reduction increasing with decreasing Au particle size (see Supplementary Table S1). However, in the case of the two smallest Au NPs (prepared by two highest concentrations of Tween 80, 1 mmol/L and 10 mmol/L), the situation is somewhat complicated. After the initial fast part of the reaction, a slowdown of reaction rate was observed. This effect is most probably caused by saturation of active surface by reaction product 4-aminophenol and its slow desorption. The inhibition of the reaction by the reaction product is a common complication in heterogeneous catalysis, and in the case of studied reaction can be ascribed to the greater affinity of 4-aminophenol to gold surface compared to 4-nitrophenol. The difference in adsorption activity can be assigned to the different mode of adsorption of these two molecules onto gold surface. In the case of 4-aminophenol, the functional group -NH 2 has character of Lewis base while gold is a typical Lewis acid. The acid-base interaction is therefore responsible for the strong adsorption of 4-aminophenol on gold surface. On the other hand, the 4-nitrophenol lies in flat, parallel orientation on the gold surface, which is typical for van der Waals interaction mediated by the conjugated system of aromatic ring's π-electrons 61 . The acid-base interaction is generally very strong (energy of this interaction is in the range typical for chemical bond) compared to the weak van der Waals interaction and therefore 4-aminophenol is adsorbed more strongly on the gold surface than 4-nitrophenol. Electrochemical study of the 4-nitrophenol reduction on the gold electrode was performed to support this difference of the reaction components in the adsorption behaviour. The cyclic voltammogram obtained on gold electrode at pH = 10 and scan rate 500 mV/s ( Supplementary Fig. S6) confirmed the hypothesis about inhibition of the reaction by the strong adsorption of the reduction products. The reduction peak of 4-nitrophenol observed at −0.820 V (vs. saturated AgCl/Ag reference electrode) during the first scan disappeared at second scan which is a clear evidence of blocking of electrode surface by the reduction product 4-aminophenol [62][63][64] . Inhibition by slow desorption of the reaction product 4-aminophenol is probably also responsible for the slowdown of the reaction rate after the quick phase observed for smallest catalytic gold nanoparticles (Fig. 6). The turning point on the kinetic curve was observed at the same conversion of the 4-nitrophenol (at about 65%) for both fastest catalytic experiments for which the inhibition is most likely due to the shortest reaction time. In these two cases the reduction rate is so fast that desorption of reaction product 4-aminophenol turns out to be the rate limiting step in the overall reaction mechanism. Therefore, based on the discussed presumption of change in the reaction rate limiting step, the apparent rate constants taken from the beginning quick part of these kinetic curves were used for evaluation of dependence of reaction rate on the total surface of catalyst. Due to the difference between Langmuir-Hinshelwood and Eley-Rideal mechanisms, reflected in the different dependence of reaction rate on the catalyst's surface, regression of the dependence of the obtained kinetic data on total surface of gold catalyst was conducted. In the case of Langmuir-Hinshelwood model the correlation should be along the quadratic function (Eq. 4) while in the case of Eley-Rideal mechanism the correlation has a linear form (Eq. 7). In the studied cases, kinetic data (apparent rate constants) depend on the total surface of gold catalyst linearly (coefficient of determination 0.9088) as predicted by the Eley-Rideal mechanism. However, if the same data are interpolated by the quadratic function as is true in the case of Langmuir-Hinshelwood model the correlation is significantly worse, as the correlation coefficient is only 0.8319 (Fig. 7). As reported in the literature, for a typical reaction which fulfils this model (e.g. the reduction of methylene blue with NaBH 4 catalysed by Ag NPs), a correlation of the data with quadratic dependency k app = f(S 2 ) is much better (correlation coefficient 0.949) 56 .
The catalytic efficiency of the prepared Au NPs was compared with literature data on the basis of the apparent rate constants normalized to the same concentration of gold in the reaction system (Supplementary Table S2). Attention was paid to studies which were aimed at application of surface modified Au NPs in the same model reaction as was used in this study.
Primarily, the reported results confirm the strong effect of the size of Au NPs on their catalytic activity. The apparent rate constants are steeply growing with decreasing size of nanoparticles due to growth of total surface area at the same concentration of the catalyst in the reaction system. Additionally to this main effect, the catalytic activity is also dependent on the surface modification of the nanoparticles. Modification of the catalyst surface by citrate 40 has nearly the same effect as in the case of Tween 80 used in this study. Some improvement of the catalytic activity of Au NPs were obtained using azacryptand (uncharged molecule with N heteroatom -Lewis base) as surface modifier 40   In summary, the presented study introduces a controlled preparation of Au NPs in the range from 6 up to 22 nm. The size of Au NPs was controlled by varying the concentration of Tween 80, which, controls the size and the polydispersity of resulting Au NPs. Higher concentration of Tween 80 results in interaction with a greater number of growing gold nuclei which are formed during the initial phase of the reduction process. With their surface blocked by Tween 80, the formation of new nuclei is preferred over the further growth of the preformed nuclei, as reflected by slowdown of the reduction rate and in final formation of smaller nanoparticles with increasing concentration of the surfactant used during the synthesis of the Au NPs. Analysis of the kinetic data obtained from catalytic study with the prepared Au NPs reveals that reduction of 4-nitrophenol is going in this case along with the Eley-Rideal mechanism of heterogeneous catalysis.

Materials and Methods
Chemicals. Tetrachloroauric  Synthesis of various-sized Au NPs. The colloidal dispersions of Au NPs were prepared by the reduction of tetrachloroauric acid with saccharide maltose at the laboratory temperature. The final size of resulting Au NPs was controlled by varying the amount of nonionic surfactant Tween 80 present in the solution. In the case of the solution free of Tween 80, 5 ml of the 5 mmol/L tetrachloroauric solution was diluted with 10 ml of deionized water. In the case of solutions with added Tween 80, 25 mmol/L of Tween 80 solution was first added to the 5 mmol/L tetrachloroauric solution, then the solution was topped up to 15 ml with deionized water. The final concentration of Tween 80 in the reaction system ranged between 0.01 mmol/L and 10 mmol/L. The solutions were vigorously stirred using a magnetic stirrer. The reduction was started by the addition of 10 ml of reducing solution containing maltose and sodium hydroxide both at concentration of 25 mmol/L. The formation of the nanoparticles was accompanied with a color change from light yellow to dark red or violet depending on the resulting Au NPs size. The size distribution of the prepared Au NPs was determined by DLS and UV-Vis spectroscopy after completing of the reaction (30 minutes). The UV-Vis spectra were recorded using samples diluted five times due to high absorbance of the produced dispersions. Complementary characterization of the particles was performed by transmission electron microscopy (TEM). The samples for catalytic testing were left in an open beaker for 24 hours to remove all unreacted sodium borohydride.
Characterization techniques. The size of the prepared Au NPs was determined using dynamic light scattering (DLS, Zetasizer Nano ZS, Malvern Instruments, UK) and verified by transmission electron microscopy (TEM, JEM 2100, Jeol Ltd., Japan). Samples for TEM were prepared simple by sedimentation of Au NPs from a drop of dispersion on the microscopic copper TEM grid with carbon layer. Excess of the dispersion was removed by filtration paper and the grids with the Au NPs were dried free on the air (covered by the Petri dish) at laboratory temperature. The nanoscopic character of the prepared gold particles was further confirmed by UV-Vis spectroscopy (Specord S600, Analytic Jena AG, Germany), which was also used for the determination of the concentration of 4-nitrophenol in the catalytic experiments.
Catalytic activity measurements. The catalytic activity of the synthesized Au NPs was tested using the model reaction of 4-nitrophenol (4-NP) reduction to 4-aminophenol (4-AP) by sodium borohydride at pH 10 (adjusted by 1 mol/L NaOH aqueous solution). The consumption of 4-nitrophenol was monitored by the decrease of its absorption peak (located at 400 nm) with time. For the purposes of the kinetic experiments, 2 ml of the 1.5 mmol/L solution of 4-nitrophenol was mixed with 0.47 ml of distilled water and 0.03 ml of the prepared gold colloids at Au concentration of 1 mmol/L in a quartz cuvette. The solution was being mixed by shaking the cuvette. The reaction was initiated by rapid injection of 0.5 ml of a 10 mmol/L solution of sodium borohydride, and the cuvette was quickly placed into the UV-Vis spectrometer for the collection of spectra as a function of time. The kinetic experiments were repeated three times for confirmation of the reproducibility of the catalytic activity of the tested Au NPs.
Calculation of the surface of the nanocatalyst. For the purpose of catalytic activity evaluation of the total surface area per sample of the prepared Au NPs was calculated based on their particle size determined by TEM and the total amount of gold in the reaction mixture. The morphology of the prepared Au NPs was approximated for this calculation by a sphere with diameter equal to the average diameter of the particles in dispersion determined from the TEM images.