Liquid phase hydrogenation of methyl-N-Boc-pyrrole-2-carboxylate over tailored Ru nanoparticles

Liquid phase hydrogenation of methyl-N-Boc-pyrrole-2-carboxylate over tailored Ru nanoparticles

Journal of Catalysis 289 (2012) 249–258 Contents lists available at SciVerse ScienceDirect Journal of Catalysis journal homepage: www.elsevier.com/l...

526KB Sizes 0 Downloads 22 Views

Journal of Catalysis 289 (2012) 249–258

Contents lists available at SciVerse ScienceDirect

Journal of Catalysis journal homepage: www.elsevier.com/locate/jcat

Liquid phase hydrogenation of methyl-N-Boc-pyrrole-2-carboxylate over tailored Ru nanoparticles Norbert Steinfeldt ⇑, Michael Sebek, Klaus Jähnisch Leibniz-Institut für Katalyse e.V. an der Universität Rostock, Albert-Einstein-Str. 29a, 18059 Rostock, Germany

a r t i c l e

i n f o

Article history: Received 14 October 2011 Revised 6 January 2012 Accepted 23 February 2012 Available online 27 March 2012 Keywords: Ru nanoparticles Polyol process SAXS PVP Particle size Hydrogenation Pyrroles

a b s t r a c t Ru nanoparticles were prepared in ethylene glycol at different pH, temperature and RuCl3nH2O concentration as well as with and without poly(N-vinyl-2-pyrrolidone) (PVP) as protective agent. They were characterized in the synthesis solution by small-angle X-ray scattering (SAXS) using Guinier analysis and indirect Fourier transformation (IFT) technique and by TEM, XRD, XPS and ATR-IR. From SAXS data evaluation, it was derived that size of the formed Ru nanoparticles depends mainly on the pH of the synthesis solution and only to a minor extent on Ru precursor concentration and reduction temperature. The impact of PVP on particle size was also only very low. The Ru nanoparticles were active for hydrogenation of methyl-N-Boc-pyrrole-2-carboxylate in ethanol solution at 25 °C and p(H2) = 5 bar. Particle size was found to be an important parameter determining hydrogenation activity. A maximum in the activity (TOF = 1000 h 1) was obtained over Ru nanoparticles with a mean particle diameter of about 1.6 nm. However, differently prepared nanoparticles of nearly identical particle size diverge strongly in their catalytic activity. These differences might be due to different surface sites and a blocking of active metal surface sites by surface intermediates formed during nanoparticle formation. Both surface structure and the extent of such blocking strongly depend on the synthesis conditions. If the nanoparticles were protected by PVP, their hydrogenation activity clearly drops in comparison to similarly sized nanoparticles without PVP protection. Ó 2012 Elsevier Inc. All rights reserved.

1. Introduction Metallic nanoparticles are considered as promising catalytic material because their electronic and catalytic properties are strongly influenced by their size and shape [1]. By controlling these parameters, catalysts with high activity and selectivity can be designed [2]. The polyol method [3] allows to prepare transition metal nanoparticles with tunable particle size and narrow size distribution. In this procedure, the polyol acts both as solvent and reducing agent for the metal precursor. Stabilization of the nanoparticles takes place by electrostatic or steric effects or combinations of both [4] and can be provided e.g. by oxidation products of the polyol [5], by addition of acetates ions [6] or more protective stabilizers like PVP (poly(N-vinyl-2-pyrrolidone)), which is one of the most frequently employed stabilizer in nanoparticle synthesis [4]. It is assumed that polymers like PVP protect the nanoparticles from aggregating by steric stabilization and by ligation of nanocluster surface atoms [7]. Among the noble metals, Ru is the cheapest and, therefore, of particular interest for catalytic applications. Ruthenium is active

⇑ Corresponding author. Fax: +49 (0)381 1281 50319. E-mail address: [email protected] (N. Steinfeldt). 0021-9517/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jcat.2012.02.017

in different types of reactions e.g. ammonia synthesis [8] and oxidation [9], NO [10] and N2O [11] reduction and especially in hydrogenation of aromatics [12]. Previously, Ru nanoparticles with tunable particle size and narrow size distribution have been prepared by the polyol process using different Ru precursors (e.g. Ru(acac)3, RuCl3) and polyols like ethylene glycol, propanediol and butanediol often with PVP as surface-capping agent [13–18]. Most of these works were dealing only with nanoparticle preparation and characterization. For catalytic applications, such tailored nanoparticles were deposited on metal oxide supports [15,16,18]. PVP–protected Ru nanoparticles prepared in aqueous solution were also successful applied for the hydrogenation of arenes, olefins and carbonyl compounds with turnover frequencies (TOF) larger than 1000 h 1 [19]. In order to achieve these high activities, elevated temperature (80 °C) and high H2 pressure (4 MPa) have to be applied. The Ru–PVP system was also applied for the selective hydrogenation of o-chlorobenzene to o-chloroaniline using similar conditions [20]. Particle size and size distribution of the Ru nanoparticles were generally derived from TEM analysis. Small-angle X-ray scattering (SAXS) is another powerful technique to provide information about colloidal structures within liquid phases without any pretreatment. It was applied to follow the formation and growth of metallic nanoparticles [21,22] and for determination of the particle size

250

N. Steinfeldt et al. / Journal of Catalysis 289 (2012) 249–258

O

N O

Boc

Ru-NP catalyst 5 bar H2, EtOH, rt

1

O

N O

Boc 2

Scheme 1. Hydrogenation of Methyl-N-(tert-butoxycarbonyl)-pyrrole-2-carboxylate 1.

and size distribution [23]. Furthermore, it was demonstrated that for nanoparticles in the sub-10-nm size range, SAXS analysis gives comparable particle characteristics as the ones obtained by TEM or XRD [24]. SAXS was also applied to study the spatial distribution of metallic nanoparticles in colloidal dispersions containing polymeric stabilizers like PVP [25,26]. Pyrrolidine derivates are valuable pharmaceutical intermediates and can be synthesized by hydrogenation of pyrroles derivates using supported noble metals [27–29]. Until now, relationships between the structure of the active element and the catalytic performance were not established. Therefore, in the present work, tailored colloidal dispersed Ru nanoparticles synthesized in ethylene glycol at different conditions (pH, c(Ru), T) were employed for hydrogenation of methyl-N-(tert-butoxycarbonyl)-pyrrole-2-carboxylate (butoxycarbonyl = Boc) in liquid phase at mild reaction conditions (see Scheme 1). SAXS data analysis is used to elucidate the influence of the synthesis conditions on the particle structure and to identify parameters that might influence the catalytic performance of the Ru nanoparticles in the hydrogenation reaction. Additionally, the influence of a pretreatment step on the catalytic activity of the nanoparticles was studied. The obtained results could be useful for the development of tailored catalytic material with improved catalytic performance. 2. Experimental section 2.1. Synthesis of ruthenium nanoparticles In a typical experiment, RuCl3nH2O was dissolved in 10 mL ethylene glycol at room temperature with or without PVP (see Supplementary Material). The pH was adjusted by dropwise addition of NaOH solved in ethylene glycol (0.5 M) under stirring. Finally, the solution was filled up to 20 mL with an appropriate amount of ethylene glycol. Next, the solution was attached in a preheated oil bath for 3 h under vigorous stirring (600 rpm) in a weak argon flow. At the end, the dark brown solution was cooled down to room temperature. For the seed-mediated growth of ruthenium particles, 2 mL of the colloidal solution prepared at pH 8.3 and 150 °C (0.04 mmol Ru) were mixed with a solution of RuCl3nH2O (0.04 mmol Ru) in 1.75 mL ethylene glycol. The pH was adjusted to 8.3 with NaOH in ethylene glycol (0.5 M), and 0.25 mL ethylene glycol was added. The solution was stirred under an argon flow for 3 h (600 rpm) at 150 °C and then cooled down to room temperature. 2.2. Pretreatment of the ruthenium nanoparticles A 0.5 mL of colloidal solution was mixed with 1.5 mL acetone. After addition of 6 mL toluene, the mixture was stirred until a brown precipitate occurred. Toluene was added to the acetone– nanoparticle mixture to avoid irreversible particle aggregation in the samples, which did not contain PVP as stabilizing agent. The clear and colorless organic solvent was decanted, and the procedure was repeated two times until a high viscous precipitate remained. The viscous precipitate that contains the nanoparticles and a rest of ethylene glycol was finally redispersed in 10 mL

ethanol. For the hydrogenation experiments, an aliquot amount of Ru was given to the reaction mixture. 2.3. General procedure for the hydrogenation experiment The hydrogenation was performed in a 100 mL autoclave (Büchi miniclave) equipped with a magnetic stirrer. The substrate 1 (112 mg, 0.5 mmol) and ruthenium nanoparticles (1  10 3 mmol Ru) solved in ethylene glycol (synthesis solution) or ethanol (after pretreatment) were added to a mixture of 20 mL ethanol charged with 56 mg n-undecane as analytical standard. The autoclave was purged five times with hydrogen, and afterwards, the reaction mixture was stirred (1100 rpm) at ambient temperature and 5 bar hydrogen. To determine the reaction progress, samples (0.25 mL) were taken from the reaction mixture at regular intervals and diluted with 0.75 mL ethanol. 2.4. Small-angle X-ray scattering (SAXS) SAXS measurements were carried out with a Kratky-type instrument (SAXSess, Anton Paar, Austria) operated at 40 kV and 50 mA in slit collimation using a two-dimensional CCD detector (T = 40 °C). The 2D scattering pattern was converted into a onedimensional scattering curve as a function of the magnitude of the scattering vector q = (4p/k) sin (h/2) with SAXSQuant Software (Anton Paar). A Göbel mirror was used to convert the divergent polychromatic X-ray beam into a collimated line-shaped beam of Cu Ka radiation (k = 0.154 nm). Slit collimation of the primary beam was applied in order to increase the flux and to improve the signal quality. The liquid sample cell consisted of a quartz capillary (internal diameter: 1 mm) stacked in a metal body with two windows for the X-ray beam. Using this sample cell, an identical volume of the solution was always irradiated. Scattering profiles of Ru nanoparticles were obtained by subtraction of the detector current and the scattering profile of the solvent from the scattering profiles of the Ru nanoparticles containing solution. 2.5. ATR-IR measurements ATR-IR spectra were acquired using a Bio-Rad FTIR spectrometer (FTS 3000MX) and a horizontal ATR accessory (Spectra Tech). To separate the nanoparticle from the ethylene glycol bulk, the colloidal solution was pretreated with a similar procedure as described in Section 2.2. After this pretreatment, the nanoparticles were again solved in ethanol and deposited on the ZnSe crystal surface (5  1 cm) as a thin film. After evaporation of the ethanol, the nanoparticle deposition was repeated several times until a thin brown film was visible on the crystal surface. The data collection consisted of 128 scans per spectrum with a resolution of 2 cm 1. 3. Results and discussion 3.1. Ru nanoparticles prepared without PVP 3.1.1. Influence of synthesis condition on the particle structure SAXS profiles (log I(q) vs log q, I – intensity) of Ru nanoparticles synthesized at three different pH values but at constant temperature of 150 °C and Ru concentration of 20 mM are presented in Fig. 1. A single scattering curve contains general information about the particle size, shape and the surface per volume ratio of the nanostructure [30]. To extract such parameters from the SAXS profile, different methods of SAXS data analysis can be applied e.g. Guinier analysis [31] and indirect Fourier transformation (IFT) technique [32,33]. A thorough discussion of these methods is provided in the Supplementary Material. The scattering profile of

N. Steinfeldt et al. / Journal of Catalysis 289 (2012) 249–258

Fig. 1. Scattering profiles of Ru nanoparticles synthesized at different pH (T = 150 °C, c[Ru] = 20 mM, solid lines – fits using the IFT method).

the Ru nanoparticles formed at the pH value of 8.3 differs from those synthesized at the pH 12 and 10 as becomes evident from a stronger decrease in the scattering intensity at lower q-values. Scattering profiles of Ru nanoparticles formed at an initial pH of 8.3 and different temperature or Ru concentrations are shown in Fig. 2. As can be seen, both parameters had approximately no influence on the shape of scattering profiles. Fig. 3 shows Guinier plots from the scattering profiles of Ru nanoparticles prepared at different pH. Similar plots were also obtained for Ru nanoparticles synthesized at pH = 8.3, but at different temperatures and Ru concentrations. Guinier analysis can be used to estimate the particle size. Results of Guinier analysis are summarized in Table 1. The linear correlation of the scattering intensity over a wide q range in the Guinier plots (0.1 < q < 3 nm 1) indicates that the Ru nanoparticles have a narrow size distribution. The strong increase in scattering intensity at low q-values for the particles formed at the pH of 12 hints that at this high pH value beside the main population of small nanoparticles, also a minor fraction of Ru nanoparticles with larger particle size might be formed. The scattering profiles of Figs. 1 and 2 were also transformed into real space applying indirect Fourier transformation (IFT) method using the program GIFT [34]. The result of this free-form transformation is the pair distance distribution function (PDDF). The PDDF represents a histogram of the distances inside the particle weighted with the electron density difference to the solvent and contains information about the maximum particle size, shape and internal structure of the particle in real space [35]. Fig. 4 shows the PDDF of the Ru nanoparticles synthesized at different synthesis conditions. The shape of the PDDF is very symmetrical indicating that the particle shape is not far away from that of a sphere [30]. Some PDDF’s show a second maximum with very low intensity at higher r-values, which might be due to the presence of a minor fraction of larger Ru structures. The most probable particle radius

251

Fig. 3. Guinier plots for Ru nanoparticles synthesized at different pH (T = 150 °C, c(Ru) = 20 mM).

that is represented by the maximum of the PDDF [36] again most strongly depends on the pH value. For the pH value of 8.3, it is 0.81 nm in comparison to 0.64 nm for the pH of 10 and 0.67 nm for the pH of 12. Guinier analysis supports this effect and provides very similar mean radii of Ru nanoparticles of 0.87, 0.69 and 0.72 nm for the pH value of 8.3, 10 and 12, respectively. The slightly higher values obtained from Guinier analysis might be due to a stronger weighting of the larger particles in this method [37]. Particle size of Ru nanoparticles obtained by SAXS data analysis is in accordance with the result obtained by TEM (TEM images are presented in the Supplementary Material). As can be seen from the values in Table 1, differences in particle size between both techniques were only in the sub-nanometre scale. TEM results also confirm the narrow size distribution of the Ru nanoparticles derived earlier from Guinier plots. If the nanoparticles prepared at pH = 8.3 were used as seeds for formation of larger particles, then at the same pH, both the particle size and the size distribution increase. However, increase in size distribution (0.27 to 0.46 nm) is clearly larger than the increase in particle diameter (1.69 to 1.99 nm). The presented results reveal that Ru nanoparticle size is mostly affected by the pH value of the synthesis solution and only to a minor extent by the temperature and the Ru concentration. The synthesis solution pH as a key factor that influences nanoparticle size in ethylene glycol was already proposed for formation of PtRu nanoparticles [5]. The role of the pH was explained by the formation of glycolate via oxidation of ethylene glycol, which acts as stabilizer for the formed nanoparticles. The pH range where the size of the nanoparticles could be varied was given between 6 and 2 [5]. Here, in agreement with literature [38], the smallest particles were obtained at pH 10.3 and 12 in strong alkaline solution where particle formation proceeds at nearly constant pH and high

Fig. 2. Scattering profiles of Ru nanoparticles synthesized at pH 8.3 at different (a) temperature (c[Ru] = 20 mM) and (b) Ru concentration (T = 150 °C); Solid lines are fits using the IFT method; in (a), the profiles were shifted by a factor of 2 for clarity.

252

N. Steinfeldt et al. / Journal of Catalysis 289 (2012) 249–258

Table 1 Results of SAXS data evaluation (Guinier radius, PDDF) and TEM for Ru nanoparticles formed in ethylene glycol at different synthesis conditions. c(Ru) (mol L

1

)

20 After pretreatment 20 20 20 20 5 50 20 + 20b a b

T (°C)

pH

150

8.3

150 150 115 190 150 150 150

10.3 12 8.3 8.3 8.3 8.3 8.3

RG (nm) 4

0.89 ± 6.59  10 0.89 ± 4.59  10 4 0.69 ± 3.2  10 4 0.72 ± 3.3  10 4 0.89 ± 1.6  10 3 0.86 ± 1.6  10 3 0.89 ± 4.0  10 3 0.89 ± 1.6  10 3 1.09 ± 3.8  10 4

Max of PDDF (nm)

TEMa (nm)

0.81 0.81 0.64 0.67 0.84 0.81 0.80 0.80 1.00

1.69 ± 0.27 1.78 ± 0.27 1.25 ± 0.20

1.99 ± 0.46

Particle diameter. Ru nanaparticles prepared at pH = 8.3, T = 150 °C and c(Ru) = 20 mM were used as seeds.

Fig. 4. PDDF of Ru nanoparticles synthesized at different (a) pH, (b) temperature and (c) Ru concentration obtained by IFT technique (see also Table 1).

glycolate concentration. At these conditions, nanoparticle formation might be dominated by fast nucleation reactions. The increasing particle size at pH 8.3 indicates a larger influence of growth reactions on particle formation compared to pH 10.3 and 12. The low influence of temperature on the particle size of the Ru nanoparticles might be explained by the assumption that for the applied pH (8.3), reduction of the metal precursor and nucleation already take place at temperatures lower than 115 °C. This hypothesis is supported by observation of the temperature range where the color of the synthesis solution was changed during heating up. Color change already begins at 90 °C and finished at 115 °C with a dark brown color that indicates nanoparticle formation [9]. Assuming that the main part of the metal precursor was consumed during a short nucleation period between 90 and 115 °C, the final temperature will influence the particle size only in low extent. The Ru nanoparticles prepared at pH 12 and 8.3 at different temperatures were also subjected to XRD (see Supplementary Material). For all samples studied, the XRD pattern shows a broad peak with a maximum at about 42° (2Theta) [17]. This maximum is between the maximum peak of metallic Ru in face-centred cubic (ICDD-PDF-Nr [088-2333]) and hexagonal close-packed modification (ICDD-PDFNr [070-0274]). None of the different diffractograms showed XRD peaks that can be attributed to RuO2. These results are in agreement with literature data reporting that crystalline Ru nanoparticles in metallic state are prepared with the ethylene glycol method [15,17]. From the HRTEM of a particle prepared at pH = 8.3 and 150 °C (c[Ru] = 20 mM), a lattice fringes distance of 0.233 nm was determined, which fit those of the (1 0 0) Ru crystal plane (ICDD PDF-Nr [070-0274]). This indicates that the Ru nanoparticles are not only in metallic state, but also single crystalline. This assumption is further supported by comparable values of the mean crystallite size derived from the XRD line width and the mean particle size determined by SAXS. To obtain information about the nanoparticle surface state, Ru nanoparticles prepared at pH 8.3 (T = 150 °C, c[Ru] = 20 mM) and at pH 12 (T = 150 °C, c[Ru] = 20 mM) were deposited on alumina after a similar pretreatment procedure used

for catalytic experiments and analyzed by XPS using the Ru 3p3/2 binding energy. The XPS spectra (see Supplementary Material) show a symmetrical peak at around 461 eV for both samples that indicate that the nanoparticles prepared at pH 12 and pH 8.3 are in metallic state even after pretreatment under ambient conditions [14]. 3.1.2. Catalytic results of Ru nanoparticles without PVP stabilization Because of the absence of steric stabilizers in the synthesis solution, the Ru nanoparticles are stabilized electrostatically by surface adsorption of organic molecules like glycolates. The pretreatment step was applied to reduce both the amount of ethylene glycol and adsorbed organics as far as possible without the nanoparticles aggregate before using them for hydrogenation reaction. With this step, also the Na+ and Cl ions concentration will be minimized. SAXS data and TEM analysis (Table 1) support the assumption that such pretreatment step has only a minor influence on the primary structure of the Ru nanoparticle. The synthesized Ru nanoparticles were applied for hydrogenation of the methyl-Boc-pyrrol-carboxylate 1 in liquid phase at ambient conditions. In preliminary experiments, the reaction solution was investigated by NMR in order to find out whether beside the expected product methyl-N-(tert-butoxycarbonyl)-pyrrolidine-2-carboxylate 2 also other products were formed. Both with NMR and GC–MS analysis, no further products were detected, which indicated that 2 was the only product of the hydrogenation reaction. Results of hydrogenation without and with pretreatment are presented in Fig. 5 for nanoparticles synthesized at three different pH values. Conversion over nanoparticles prepared at pH 8.3 is much higher than conversion over nanoparticles prepared at pH 10 and 12 independent of whether the particles were pretreated or not. While pretreatment of nanoparticles prepared at pH of 8.3 did not influence their catalytic activity, it improved clearly the activity of nanoparticles prepared at pH 10 and 12. Without pretreatment, conversion of 1 over these particles after 3 h was lower than 10%. With pretreatment, conversion increased

N. Steinfeldt et al. / Journal of Catalysis 289 (2012) 249–258

253

from 5 to 1 bar. Otherwise, an increasing H2 pressure from 5 to 10 bar gives the same catalytic activity as obtained for the experiment at 5 bar. That means that for the applied conditions, H2 mass transfer limitations are not the reason for the observed differences in the catalytic activity between the Ru nanoparticles prepared at different synthesis conditions. Concentration of hydrogen in the reaction solution can be approximated using Henry’s law constant for absorption of hydrogen in ethanol [39] (H = 29,380 Pa m3/mol). Applying the relation H = p/C, the hydrogen concentration in the liquid phase increases from 3.4  10 3 mol/L at 1 bar to 1.7  10 2 mol/L at 5 bar. At 5 bar, the H2 concentration in the liquid phase is in the same range as the concentration of methyl-N-Boc-pyrrole-2-carboxylate. Fig. 5. Conversion of methyl-N-Boc-pyrrole-2-carboxylate 1 versus reaction time over Ru nanoparticles prepared at different pH without (open symbols) and with pretreatment (closed symbols); (c(1) = 2.5  10 2 mol/L, c(Ru) = 5  10 5 mol/L, T = 25 °C, pH2 = 5 bar, the line are drawn for clarity).

in this time interval to 30% for particles prepared at pH 10 and to 40% for particles prepared at pH 12. The conversion over the nanoparticles prepared at pH 8.3 by using seeds is only slightly higher than conversion over particles prepared at pH = 12. Fig. 6 compares the temporal evolution of conversion over Ru nanoparticles prepared at constant pH (8.3), but different temperatures (a, c(Ru) = 20 mM) and Ru concentrations (b, T = 150 °C) again with and without pretreatment. It can be seen that both parameters (temperature and Ru concentration) affect the catalytic performance of the Ru nanoparticles in the hydrogenation reaction. Differences in conversion of 1 between nanoparticles formed at 150 °C and 190 °C (X after 2.5 h > 90%) are low compared to the conversion over particles formed at 115 °C (X after 2.5 h < 70%). Relatively, high hydrogenation activity was also obtained for nanoparticles formed with Ru concentration of 20 and 50 mM in the precursor solution (Fig. 6b). For catalytic active Ru nanoparticles, pretreatment again had only a relatively small influence on conversion of 1 as already observed for the nanoparticles prepared at different pH. The largest influence of pretreatment was obtained for the nanoparticles prepared at low Ru concentration (5 mM). Here, conversion obtained after 3 h increases from 20% (without pretreatment) to 70%. Generally, there seems to be a tendency that the influence of pretreatment is higher the lower the catalytic activity without pretreatment. In order to exclude that the observed differences in conversion of 1 are dominated by mass transfer of hydrogen, the experiments with the most active Ru nanoparticles were repeated using different H2 pressures. Results of these experiments are shown in Fig. 7. Conversion of 1 decreased strongly if the H2 pressure was reduced

3.2. Ru nanoparticles protected by PVP 3.2.1. Influence of PVP on particle structure To study the influence of PVP on particle structure and catalytic activity, Ru nanoparticles with different PVP to Ru ratio were synthesized at two different pH. The used Ru concentration (20 mM) and temperature (150 °C) corresponds to the synthesis method, which revealed Ru nanoparticles with the highest hydrogenation activity at absence of PVP. Fig. 8 shows scattering profiles of Ru nanoparticles prepared at pH = 8.3 employing different Ru to PVP ratios and the corresponding PDDF obtained by the IFT method. Scattering patterns of Ru nanoparticle in the central part of the scattering curve (1 < q < 3 nm 1) are similar for all samples. In contrast to the sample without PVP, scattering curves from Ru nanoparticles protected with PVP show some increase in scattering intensity at low q-values (Fig. 8a), which is most pronounced for the sample with the lowest molar PVP to Ru ratio (0.5:1). The PDDF of the PVP-protected particles differ considerably from that prepared without PVP. The PDDF is symmetrical only for r-values lower than 2 nm and shows for larger r a longer tailing with some oscillations. This tailing increased with decreasing PVP to Ru ratio. The appearance of the long tailing is connected with the occurrence of different slope in scattering intensity between the innermost (q < 0.5 nm 1) and the central part of the scattering curve (see Table 2 and also Guinier plots in the Supplementary material). A large-size tailing of the PDDF without oscillations was earlier observed for polymer protected PbS nanoparticles and explained with the assumption that one polymer coats several nanoparticles [40]. For the PVP to Ru ratio of 0.5:1, the size of the larger Ru/PVP structures is about 15 nm. However, the number of polymer chains that contain more than one Ru nanoparticles will always be low. The symmetrical shape of the sharp PDDF peak at r < 2 nm might indicate a nearly spherical shape of the particles in agreement with previous results [16]. The maximum position of the PDDF, which

Fig. 6. Conversion of methyl-N-Boc-pyrrole-2-carboxylate 1 versus reaction time over Ru nanoparticles prepared at pH = 8.3 and (a) – different temperature (c[Ru] = 20 mM) and (b) – Ru concentration (T = 150 °C); without (open symbols) and with pretreatment (closed symbols); (c(Ru) = 5  10 5 mol/L, c(1) = 2.5  10 2 mol/L, T = 25 °C, pH2 = 5 bar, the lines are drawn in for clarity).

254

N. Steinfeldt et al. / Journal of Catalysis 289 (2012) 249–258

Fig. 7. Influence of H2 pressure on the conversion of methyl-N-Boc-pyrrole-2carboxylate over Ru nanoparticles prepared at pH = 8.3, 150 °C with c(Ru) = 20 mM (c(1) = 2.5  10 2 mol/L, c(Ru) = 5  10 5 mol/L, T = 25 °C).

corresponds to the most probable particle radius is nearly identical (0.75–0.84 nm) for the samples (see Table 2). Guinier analysis from the central part of the scattering curve revealed mean particle radii between 0.82 and 0.85 nm (Table 2). These radii are approximately the same as found for the sample without PVP (0.81 nm, Table 1) using identical synthesis conditions. This indicates that the size of the primary formed nanoparticles is only influenced marginally by the presence of PVP. TEM results of the sample with the PVP to Ru ratio of 0.5:1 confirm the particle size derived from SAXS data analysis. The difference in particle size obtained by SAXS data analysis and from the corresponding TEM image was again in sub-nm range (Table 2). Fig. 9 presents scattering curves and the corresponding PDDF for Ru nanoparticles prepared at pH = 4.3 with two different PVP to Ru ratios. At such low pH, protective stabilizers like PVP have to be present in order to avoid a high degree of irreversible particle agglomeration and precipitation. The obtained PDDF suggests that Ru nanoparticles with nearly spherical shape and narrow size distribution were formed when preparing them with the PVP to Ru ratio of 5:1. The most probable particle radius is 1.22 nm, and the

maximum particle radius is about 5 nm. For the molar PVP to Ru ratio of 0.5:1, the form of PDDF is, however, quite asymmetric, and a large particle tailing can be observed. This might be due to a broader size distribution or as already observed for the particles prepared at pH = 8.3 due to the presence of larger Ru/PVP structures that might consists of one polymer chain that coats several particles with the latter being more probably. The most probable particle radius was shifted to 1.44 nm. Guinier radius from the central part of the scattering curve is again in correspondence with the maximum of the PDDF. There is a general tendency that the value of the Guinier radius determined at low q (<0.4 nm 1) increases if the molar PVP to Ru ratio decreases (Table 2). The low influence of the PVP concentration on the particle size at pH = 8.3 might be explained by a high reduction and nucleation rates and a weak interaction between nanoparticle surface atoms and polymer functional groups. Increasing acidity of the synthesis solution (pH = 4.3) should lead to lower reduction and/or nucleation rates. Therefore, the amount of Ru that was used for nucleation will be lower, and the growth processes are involved to a larger extent on the particle formation. As a result, particle size increases. Due to the slower nanoparticle formation at pH = 4.3, interactions between surface metal atoms and polymer functional groups might be stronger and influence the final particle size. Previously, it was reported that with lowering PVP/Ru molar ratio, larger Ru nanoparticles were formed [12]. The presented results obtained in acidic media (pH = 4.3) confirm this observation. Both the pH and PVP to Ru ratio can have an influence on the spatial distribution of the nanoparticles. At low molar PVP to Ru ratios, more than one nanoparticle might be stabilized in neighborhood to each other inside of a single PVP polymer chain. This leads to formation of larger Ru/PVP structures that were detected by SAXS at low qvalues. The probability that more than one Ru nanoparticle interacts with a single polymer chain decreases with increasing PVP to Ru ratio, and therefore, the formation of larger structures is suppressed. This finding is also in agreement with results from literature where Taylor dispersions measurements indicate that one polymer chain can attach to multiple nanoparticles [41]. Otherwize, MALDI-TOF experiments suggest that one nanoparticle might

Fig. 8. (a) Scattering profiles of Ru nanoparticles synthesized with different molar Ru to PVP ratios and (b) the corresponding PDDF (pH = 8.3, T = 150 °C, c(Ru) = 20 mM, solid lines – fits using the IFT technique; curves in (a) were shifted for clarity).

Table 2 Results of SAXS data evaluation (Guinier analysis, PDDF) and TEM for samples containing PVP (T = 150 °C, 20 mM Ru). pH 8.3 8.3 8.3 4.3 4.3 a

PVP:Ru 5:1 2.5:1 0.5:1 5:1 0.5:1

Particle diameter.

RG (nm) (0.1 nm 1.64 ± 4.0  10 2.85 ± 1.0  10 3.85 ± 4.7  10 1.64 ± 4.0  10 3.29 ± 2.2  10

3 2 3 3 3

1

< q < 0.4 nm

1

)

RG (nm) (1 nm

1

0.85 ± 1.1  10 0.85 ± 8.8  10 0.82 ± 7.7  10 1.33 ± 6.6  10 1.54 ± 5.4  10

3 4 4 4 4

< q < 3 nm

1

)

Max of PDDF (nm)

TEMa (nm)

0.75 0.75 0.84 1.21 1.44

1.56 ± 0.28

N. Steinfeldt et al. / Journal of Catalysis 289 (2012) 249–258

255

Fig. 9. (a) Scattering profiles of Ru nanoparticles protected by PVP and synthesized at pH = 4.3 and their fits (solid lines) and (b) the corresponding PDDF (T = 150 °C, c[Ru] = 20 mM).

be stabilized by one, two or three polymer chains depending on its size [42]. The XRD pattern of the PVP-protected Ru nanoparticles show also a single broad peak at about 42°, which indicates that the particles are crystalline and in metallic state (see Supplementary Material). The pattern are identical with those XRD pattern obtained from Ru nanoparticles prepared without PVP. HRTEM image of a single Ru nanoparticle prepared at pH 8.3 and a PVP/Ru ratio of 0.5:1 reveals a lattice fringes distance of 0.226 nm, which fit those of the (1 1 1) Ru crystal plane (ICDD PDF-Nr [088-2333]).

3.2.2. Catalytic performance of the PVP-protected Ru nanoparticles Fig. 10 compares the conversion of 1 over Ru nanoparticles formed without PVP with those of nanoparticles formed at different molar PVP to Ru ratios and pH. In these experiments, the Ru nanoparticles were always pretreated with the acetone/toluene mixture. Comparing temporal evolution of activity for the sample without PVP with those of the samples containing PVP, it can be clearly seen that the conversion of 1 decreases considerably if the nanoparticles are protected by PVP independent on their size. Without PVP protection, 100% conversion was obtained after 2 h. In the same time interval, the conversion of the most active PVPprotected Ru nanoparticles prepared at pH = 4.3 with a PVP to Ru ratio of 5:1 was only 30%. Also differences exist in catalytic activity between samples prepared at different PVP to Ru ratios. Ru nanoparticles prepared at pH = 8.3 with the PVP to Ru ratio of 2.5:1 and 0.5:1 show a higher activity than the sample prepared with a ratio of 5:1, while this dependence is reversed when using the nanoparticles prepared at pH = 4.3. Here the activity is higher for nanoparticles prepared with a PVP to Ru ratio of 5:1 compared to that with a ratio of 0.5:1.

3.3. Activity-determining properties In Fig. 11, the initial activity of the Ru nanoparticles prepared at different synthesis parameters is plotted versus the mean particle radius that was determined from SAXS data analysis (maximum of the PDDF). The activity was obtained from the conversion versus time plots using data at low conversion of 1 (<20%). It is generally accepted that the catalytic properties of nanoparticles are strongly influenced by their size [1,3]. In the presented work, nanoparticles with a mean diameter of about 1.6 nm reveal the highest hydrogenation activity. The TOF (number of consumed mol pyrrole per mol ruthenium per hour) of the most active Ru nanoparticles determined from the initial hydrogenation activity amounts to about 1000 h 1. If the particle size decreased, the catalytic activity compared to the most active nanoparticles drops down approximately by a factor of about 4. An increase in particle size also leads to a clear decrease in activity. While the lower activity of the larger particles (mean diameter 2 nm) might be explained by geometrical factors like lower surface area and decreasing number of surface atoms, there is no geometrical explanation for the loss in activity with decreasing particle size (mean diameter 1.4 nm). XRD pattern and XPS spectra of the nanoparticles prepared at pH 12 and 8.3 indicate that the Ru nanoparticles are crystalline and in metallic state. Therefore, it is not assumed that a different degree of nanoparticle surface oxidation is responsible for the observed differences in hydrogenation activity. During nanoparticle synthesis, ethylene glycol is partly oxidized to aldehydes, carboxylic acids and CO2 [5]. Such processes involve the formation of surface intermediates that might be desorbed or removed from the metal surface so long as particle formation is not finished and an electron transfers occurs. At the end of the formation process, some of these intermediates will stay on the

Fig. 10. Conversion of methyl-N-Boc-pyrrole-2-carboxylate versus reaction time over Ru nanoparticles prepared with different molar Ru to PVP ratios (150 °C, c(Ru) = 20 mM at (a) pH = 8.3 and (b) pH = 4.3 (c(1) = 2.5  10 2 mol/L, c(Ru) = 5  10 5 mol/L, T = 25 °C, pH2 = 5 bar).

256

N. Steinfeldt et al. / Journal of Catalysis 289 (2012) 249–258

particle surface. Previously, it was reported that ethylene glycol on Rh surfaces can be decomposed by dehydrogenation and CAC bond breaking reactions [43]. Studies of the electrooxidation pathways of ethylene glycol in alkaline aqueous solution on Pt electrodes have shown that Pt converted ethylene glycol to carbonate through sequences of chemisorbed intermediates [44]. Assuming similar processes in strong alkaline medium on the surface of Ru during particle formation, a larger number of active surface sites might be blocked from relatively strong adsorbed species that can not be removed during the applied pretreatment (Fig. 5). ATR-IR spectra of Ru nanoparticles prepared at different pH and that of pure ethylene glycol are shown in Fig. 12. Besides IR bands, which can be assigned to residual ethylene glycol (two bands around 2900 cm 1), the IR spectra of Ru nanoparticles contains three new bands located at about 2000 cm 1 and one new band located at about 1640 cm 1, which are assigned to adsorbed surface intermediates. Three IR bands around 2000 cm 1 were already observed previously on PVP-protected Ru nanoparticles after CO preadsorption [45] and assigned to linearly bonded CO (2050 cm 1), to bridging carbonyls (1945 cm 1) and to multiple individual CO molecules adsorbed on low-coordinate Ru atoms (1975). The position of these IR bands corresponds to frequencies that are attributed to CO adsorbed on zerovalent Ru adsorption sites [46]. A broad IR band about 1600 cm 1 was previously obtained after adsorption of formic acid on the Ru nanoparticle surfaces and attributed to different adsorbed formate species [45]. The nearly identical positions of the bands in Fig. 12 with that given in literature indicate that during Ru nanoparticle formation,

ethylene glycol might decompose to different oxygen containing surface intermediates that stay adsorbed on the nanoparticle surface. Furthermore, the ATR-IR results show that the applied pretreatment can not remove the adsorbed surface intermediates completely from the nanoparticle surface. The differences in the IR band intensities might be also an indication that the number of the different structured surface sites depends on the applied pH. Therefore, differences in catalytic activity between the differently sized nanoparticles might be attributed to (a) differences in strength of interaction between the adsorbed intermediates and the surface metal atoms leading to a different number of available surface sites and (b) differences in the structure of the available surface site. Volcano plots for the size-depending activity of nanoparticles in the sub-2-nm range were already observed for the aerobic oxidation of cyclohexane over Au nanoparticles and explained with the unique electronic structure of a certain particle size [47]. Also, there exist clear differences in catalytic performance for particles with the same mean particle size. The particles were synthesized at the identical pH and temperature (or Ru concentration), however, with different Ru concentration (or temperature). The differences in their activity demonstrate that particle size is only one important factor that influences the catalytic activity of the Ru nanoparticles in the hydrogenation reaction. Other important factors might be the particle shape [48] and the number of unsaturated surface atoms. SAXS data analysis using IFT technique has shown that the shape of the PDDF is nearly identical for all nanoparticles prepared at pH 8.3 (Fig. 4). Therefore, it is assumed that small differences in particle shape are not the main reason for the strong differences in hydrogenation activity. An indication of why particles with similar size and shape show clear differences in activity can be obtained again from the comparison of catalytic results with and without pretreatment (Figs. 5 and 6). While pretreatment again has no influence on the catalytic performance of the most active Ru nanoparticles, the activity of pristine particles showing low performance can be increased by the pretreatment step. ATR-IR spectra from Ru nanoparticles formed at different Ru concentrations are shown in Fig. 13. In these spectra, intensity differences of the three bands located at about 2000 cm 1 are clearly visible. Therefore, it can be assumed that the applied synthesis conditions (pH, temperature and Ru concentration) strongly influence the Ru nanoparticle surface structure, the structure of the formed intermediates and the strength of interaction between the nanoparticle surface and the specific intermediate. Also, it can not completely be excluded that differences in the degree of nanoparticle surface oxidation might be involved in the observed differences in hydrogenation activity of equally sized nanoparticles. Surface oxidation of Ru nanoparticles was often observed by XPS [14,16]. Presently, it is not clear whether this oxidation takes place during nanoparticle formation or during sample preparation [16].

Fig. 12. ATR-IR spectra of pretreated Ru nanoparticles prepared at different pH (T = 150 °C, c(Ru) = 20 mM).

Fig. 13. ATR-IR spectra of Ru nanoparticles prepared at different Ru concentration after pretreatment (pH = 8.3, T = 150 °C).

Fig. 11. Initial activity of methyl-N-Boc-pyrrole-2-carboxylate hydrogenation versus particle radius (SAXS) for pretreated Ru nanoparticles (c(Ru) = 5  10 5 mol/L); (number denotes different synthesis conditions – 1(115 °C, 20 mM Ru), 2 – (150 °C, 20 mM Ru), 3 – (190 °C, 20 mM Ru), 4 – (150 °C, 5 mM Ru), 5 – (150 °C, 50 mM Ru); pH = 8.3).

N. Steinfeldt et al. / Journal of Catalysis 289 (2012) 249–258

257

decrease of the catalytic activity that is probably caused by interaction between polymer functional groups and the nanoparticle surface.

Acknowledgments This work was supported by the Leibniz-Gemeinschaft. The authors acknowledge Dr. M.-M. Pohl (TEM), Dr. M. Schneider (XRD) and Dr. J. Radnick (XPS) for the corresponding measurements and for the helpful discussions.

Fig. 14. ATR-IR spectra of Ru nanoparticles prepared at presence of PVP (pH = 8.3, T = 150 °C, c(Ru) = 20 mM).

Otherwise, the results of ATR-IR give no indication of the presence of oxidized Ru sites e.g. CO adsorbed on oxidized Ru sites shows IR frequencies higher than 2070 cm 1 [46]. PVP-protected Ru nanoparticles compared with that of the most active Ru nanoparticles show a clearly lower hydrogenation activity. The activity drop should be connected with a decreasing amount of available active surface sites or a stronger interaction between the PVP and the nanoparticle surface. PVP interact with the nanoparticle surface by donor–acceptor interactions using polymer functional groups (C@O, CAN) [7]. However, the number of functional groups of a single polymer chain that interact with the nanoparticle surface and the true conformation of the polymer around the nanoparticle is presently unknown. Additionally, a slower diffusion of substrate and product within the polymer in which the nanoparticles are embedded might also reduce the observed activity. Hydrogenation reactions with PVP-protected nanoparticles are often carried out at clearly higher hydrogen pressures [19]. ATR-IR spectra from PVP-protected Ru nanoparticles (Fig. 14) shows a intense IR band about 1657 cm 1, which is assigned to C@O stretching vibration mode of PVP. Furthermore, IR bands of surface intermediates at about 2000 cm 1 were also detected with low intensity, which are not visible in the IR spectra of pure PVP and ethylene glycol. The low intensity of these bands is attributed to the presence of PVP, which lower the mass of nanoparticles on the ATR crystal. The differences in the IR band intensities around 2000 cm 1 indicate different surface structures also on PVP-protected nanoparticles if the preparation conditions are changed. This might be also a reason for the observed difference in hydrogenation activity between the Ru particles protected by PVP, but different PVP to Ru ratios. 4. Conclusions Small-angle X-ray scattering is a technique that can be used to characterize the structure of catalytic active Ru nanoparticle prepared in ethylene glycol in the sub-10-nm range. Results of SAXS data analysis can be correlated with catalytic data to identify key parameters that are relevant for high catalytic activity of the nanoparticles in liquid phase hydrogenation at mild reaction conditions. The obtained results point out that particle size is only one important factor that influences the catalytic activity of Ru nanoparticle in hydrogenation reaction. It is believed that differences in catalytic activity between similarly sized nanoparticles are caused by different surface structure and by blocking of active sites though adsorbed surface intermediates that are formed during nanoparticle preparation. The surface structure of Ru nanoparticles and number and bonding strength of the surface intermediates depend on the applied synthesis parameters like pH, T and Ru salt concentration. The protection of the nanoparticles by PVP leads to a clear

Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.jcat.2012.02.017.

References [1] J.M. Campalo, D. Luna, R. Luque, J.M. Marinas, A.A. Romero, ChemSusChem 2 (2009) 18–45. [2] Y. Li, G.A. Somorjai, Nano Lett. 10 (2010) 2289–2295. [3] H. Bönnemann, K. S. Nagabhushana, in: B. Corain, G. Schmid, N. Toshima (Eds.), Metal Nanoclusters in Catalysis and Material Science – The Issue of Size Control, first ed., Elsevier B.V., Amsterdam, 2008. [4] L.S. Ott, R.G. Finke, Coord. Chem. Rev. 251 (2007) 1075–1100. [5] C. Bock, C. Paquet, M. Couillard, G.A. Botton, B.R. Mac Dougall, J. Am. Chem. Soc. 126 (2004) 8028–8037. [6] J. Yang, T.C. Deivaraj, H.-P. Too, J.Y. Lee, Langmuir 20 (2004) 4241–4245. [7] Y. Borodko, S.M. Humphrey, T.D. Tilley, H. Frei, G.A. Somorjai, J. Chem. Phys. C 111 (2007) 6288–6295. [8] R Schlögl, in: G. Ertl, H. Knözinger, F. Schüth, J. Wietkamp (Eds.), Handbook of Heterogeneous Catalysis, second ed., Wiley-VCH, Weinheim, 2008, pp. 2501– 2575. [9] A. Miyazaki, L. Balint, K. Aika, Y. Nakano, J. Catal. 204 (2001) 364–371. [10] L. Balint, A. Miyazaki, K. Aika, J. Catal. 207 (2002) 66–75. [11] V.G. Komvokis, G.E. Marnellos, I.A. Vasalos, K.S. App, Catal. B 89 (2009) 627– 634. [12] J.A. Widegren, R.G. Finke, J. Mol. Catal. A 191 (2003) 187–207. [13] G. Viau, R. Brayner, L. Poul, N. Chakroune, E. Lacaze, F. Fievet-Vincent, F. Fievet, Chem. Mater. 15 (2003) 486–494. [14] N. Chakroune, G. Viau, S. Ammar, L. Poul, D. Veautier, M.M. Chehimi, C. Mangeney, F. Villain, F. Fievet, Langmuir 21 (2005) 6788–6796. [15] S.H. Joo, J.Y. Park, J.R. Renzas, D.R. Butcher, W. Huang, G.A. Somorjai, Nano Lett. 10 (2010) 2709–2713. [16] M. Zawadzki, J. Okal, Mater. Res. Bull. 43 (2008) 3111–3121. [17] X. Yan, H. Liu, K.Y. Liew, J. Mater. Chem. 11 (2001) 3387–3391. [18] J. Okal, Catal. Commun. 11 (2010) 508–512. [19] F. Lu, J. Liu, J. Xu, J. Mol. Catal. A 271 (2007) 6–13. [20] M. Liu, W. Yu, H. Liu, J. Mol. Catal. A 138 (1999) 295–303. [21] B. Abecassis, F. Testard, O. Spalla, P. Barboux, Nano Lett. 7 (2007) 1723–1727. [22] H. Tanaka, S. Koizumi, T. Hashimoto, H. Itoh, M. Satoh, K. Naka, Y. Chujo, Macromolecules 40 (2007) 4327–4337. [23] G.-W. Lee, K.S. Jin, J. Kim, J.-S. Bae, J.H. Yeum, M. Ree, W. Oh, Appl. Phys. A A91 (2008) 657–661. [24] H. Borchert, E.V. Shevchenko, A. Robert, I. Mekis, A. Kornowski, G. Grübel, H. Weller, Langmuir 21 (2005) 1931–1936. [25] T. Hashimoto, K. Saijo, M. Harada, N. Toshima, J. Chem. Phys. 109 (1998) 5627– 5638. [26] J.-M. Lin, T.-L. Lin, U.-S. Jeng, Y.-J. Zhong, C.T. Yeh, T.-Y. Chen, J. Appl. Cryst. 40 (2007) 540–543. [27] L. Hegedus, T. Mathe, A. Tungler, Appl. Catal. A 143 (1996) 309–316. [28] H.-P. Kaiser, J.M. Muchowski, J. Org. Chem. 49 (1984) 4203–4209. [29] C. Jiang, A. Frontier, J. Org. Lett. 9 (2007) 4939–4942. [30] O. Glatter, K. Kratky, Small Angle X-ray Scattering, Academic Press, New York, 1982. [31] A. Guinier, G. Fournet, Small-Angle Scattering of X-Rays, Wiley, New York, 1955. [32] O. Glatter, J. Appl. Crystallogr. 10 (1977) 415–421. [33] PCG Software Package, Version 2.02.05, University Graz. [34] G. Fritz, O. Glatter, J. Phys.: Condens. Matter 18 (2006) S2403–S2419. [35] O. Glatter, J. Appl. Crystallogr. 12 (1979) 166–175. [36] M.J. Hollamby, J. Eastoe, A. Chemelli, O. Glatter, S. Rogers, R.K. Heenan, Langmuir 26 (2009) 6989–6994. [37] D.R. Vollet, D.A. Donatti, A.I. Ruiz, J. Non-Cryst. Solids 306 (2002) 11–16. [38] Y. Wang, J. Ren, K. Deng, L. Gui, Y. Tang, Chem. Mater. 12 (2000) 1622–1627. [39] P. Lühring, A. Schumpe, J. Chem. Eng. Data 34 (1989) 250–252. [40] E. Koupanou, S. Ahualli, O. Glatter, A. Delgado, F. Krumeich, E. Leontidis, Langmuir 26 (2010) 16909–16920.

258

N. Steinfeldt et al. / Journal of Catalysis 289 (2012) 249–258

[41] N. Toshima, Y. Shiraishi, T. Teranishi, M. Miyake, T. Tominaga, H. Watanabe, W. Brijoux, H. Bönnemann, G. Schmid, Appl. Organometal. Chem. 15 (2001) 178– 196. [42] J.K. Navin, M.E. Grass, G.A. Somorjai, A.L. Marsh, Anal. Chem. 81 (2009) 6295– 6299. [43] M.M.M. Jansen, B.E. Nieuwenhuys, H. Niemantsverdriet, ChemSusChem 2 (2009) 883–886.

[44] S.-C. Chang, Y. Ho, M.J. Weaver, J. Am. Chem. Soc. 113 (1991) 9506–9513. [45] K. Tedsree, A.T.S. Kong, S.C. Tsang, Angew. Chem. Int. Ed. 48 (2009) 1443–1446. [46] S.Y. Chin, C.T. Williams, M.D. Amiridis, J. Phys. Chem. B 110 (2006) 871–882. [47] Y. Liu, H. Tsunoyama, T. Akita, S. Xie, T. Tsukuda, ACS Catal. 1 (2011) 2–6. [48] R. Narayanan, M.A. El-Sayed, Langmuir 21 (2005) 2017–2033.