Pd on boron-doped hollow carbon spheres – PdO particle size and support effects

Journal of Catalysis 305 (2013) 36–45

Contents lists available at SciVerse ScienceDirect

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

Pd on boron-doped hollow carbon spheres – PdO particle size and support effects Vilas Ravat a, Isaac Nongwe a,b, Reinout Meijboom b, George Bepete a, Neil J. Coville a,⇑ a b

DST/NRF Centre of Excellence in Strong Materials and Molecular Sciences Institute, School of Chemistry, University of the Witwatersrand, Johannesburg WITS 2050, South Africa Department of Chemistry, University of Johannesburg, Auckland Park, Johannesburg 2006, South Africa

a r t i c l e

i n f o

Article history: Received 1 February 2013 Revised 3 April 2013 Accepted 5 April 2013 Available online 28 May 2013 Keywords: Carbon supports Pd catalysts Boron-doped carbons Aerobic oxidation Alcohols

a b s t r a c t Boron-doped hollow carbon spheres (B-HCSs) were synthesized using a CVD injection method (BCl3 in toluene; Ar; 900 °C; 4 h) using Stöber silica spheres as template. A Pd complex was loaded onto the BHCSs using deposition and impregnation methods, and the materials were studied by XRD, SEM, TEM, Raman spectroscopy, TGA, ICP-OES, XPS, and 11B solid-state NMR spectroscopy. The boron-doped carbon-supported Pd catalysts (Pd/B-HCS) were compared with a Pd-loaded boron-doped carbon nanotube (Pd/B-CNT) catalyst (both have a hollow interior) in the solvent-free oxidation of alcohols using oxygen as an oxidant at 125 °C under base-free conditions. The Pd particle size was varied (2.5–12 nm) by changing the Pd/B-HCS calcination temperature (200–550 °C), and this affected the activity but not the selectivity of the benzyl alcohol to benzaldehyde reaction. The data revealed the key role of the Pd particle size on the reaction that was influenced by the B–Pd–C interaction. The reaction rate depended on the mean size of Pd particles and showed a maximum when catalysts were calcined at 300 °C, revealing that the aerobic oxidation of benzyl alcohol catalyzed by the supported PdO (dPd > 2.5 nm) nanoparticles was not a structure sensitive reaction. Ó 2013 Elsevier Inc. All rights reserved.

1. Introduction The inertness of carbon in the form of nanotubes (CNTs) and other shaped carbons such as spheres (CSs) enables them to resist low-temperature oxidation for long periods of time. The chemical and physical properties of the new carbons can also be modified by both doping and functionalization processes, and these effects can be expected to influence the oxidative behavior of the carbons. Studies have shown that the stability of carbon against oxidation can be enhanced by passivation with B2O3 (or P2O5) by a high-temperature treatment [1,2]. When the boron doping was increased, the oxidation temperature of the carbon nanotubes increased [3]. Also, the thermal stability of boron-doped hollow carbon spheres (B-HCSs) was found to be higher than that of undoped HCSs [4]. Recent studies have also shown that Pd loaded onto a boron-doped CNT (B-CNT) support gave palladium atoms which were adsorbed on the B-CNT surface above the axial boron-carbon bonds in the B-CNTs [5]. Thus, boron addition to modify the behavior of supported metals in the catalytic reactions is expected. Boron inclusion enhances the resistance to oxidation of graphitic systems [6–12], but the sintering of metal particles can be typically minimized by increasing the interaction between the metal and the support by doping with boron [13,14]. Boron doping is ⇑ Corresponding author. Fax: +27 11717 6749. E-mail address: [email protected] (N.J. Coville). 0021-9517/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jcat.2013.04.018

also known to accelerate the homogeneous continuous graphitization process of carbon without formation of separate distinct carbon components. The boron is also known to enter substitutionally [15–17] into a carbon layer. It has also been found that boron can accommodate a wide variety of bonding configurations in the graphitic carbon structures. Boron doping of carbons should lead to carbons that are well suited for use as supports in oxidation reactions. Indeed, we recently exploited this approach and found that boron doping of hollow carbon spheres (B-HCSs) gave supports that could be used in the Pd-catalyzed oxidation of benzyl alcohol [4]. From studies in the literature, the ease of carbon oxidation appears to follow the sequence: activated carbons > N-MWCNTs (N doped multiwalled CNTs) > MWCNTs > SWCNTs > B-MWCNTs (boron-doped multiwalled CNTs) [18–21]. To further exploit this approach of boron doping, we have investigated the impact of B doping by comparing B doping on two different carbons supports (HCSs and CNTs) both with hollow interiors. To vary the size of the Pd particles on the carbons, different catalyst preparation methods and activation methods have been used. We have used the benzyl alcohol oxidation reaction as the probe reaction to correlate the activity of the new Pd catalysts with their properties. In particular, the influence of the method of Pd complex deposition on different supports as well as the Pd loading and calcination temperature of the catalyst on the benzyl alcohol conversion and benzaldehyde selectivity has been investigated. The

V. Ravat et al. / Journal of Catalysis 305 (2013) 36–45

influence of the reaction conditions (viz. reaction time and temperature) on the catalytic performance has also been studied. 2. Experimental 2.1. Starting materials TEOS (98%, Aldrich), NH3 (25% Fluka), isopropanol (Merck 99%) and deionized water were used as reagents for the synthesis of silica spheres. HF (40% Associated Chemical) was used for silica removal, and a 1.0 M BCl3 solution in toluene (Aldrich) was used as carbon and boron source, while HNO3 (55%, Merck) was used for activation of the HCSs. B-CNT was prepared using acetylacetonates (Sigma–Aldrich, 99%) of the chosen metals dissolved in ethanol and mixed with magnesia powder to give a Fe/Al/MgO (1:1:12) catalyst [22]. Pyridine (99%, Saarchem), toluene (99% Merck), and Pd(CH3CO2)2 (Next Chemical, SA) were used for the catalyst preparation and benzyl alcohol (99%, Fluka) was used as a model substrate. 2.2. Preparation of silica spheres with diameter (1000–1200 nm) Monodisperse silica spheres (d = 1000–1200 nm) were synthesized following a modified Stöber procedure [23]. In this reaction, 22.4 ml of TEOS, 15.4 ml of NH3, and 18 ml of deionized water were added into 130 ml of isopropanol after 1 h of aging under mild stirring at 40 °C. A further 22.4 ml of TEOS was then added to the above reaction mixture which was stirred at room temperature overnight, resulting in the formation of a white silica colloidal suspension. The silica particles were centrifuged, separated, and washed with ethanol and distilled water four times and dried at 80 °C for 12 h. They were finally sonicated for 15 min in a water and ethanol mixture to give well-separated monodisperse silica spheres. 2.3. Preparation of boron-doped hollow carbon spheres and nanotubes Deposition of carbon onto the silica spheres was carried out using a CVD injection method. A 1 M BCl3 in toluene solution was used as a carbon and boron precursor to make the B-HCSs. Briefly, the silica spheres were placed in a quartz tube and were heated to 900 °C for 4 h at a heating rate of 10 °C min1 under an Ar flow (100 ml min1) at atmospheric pressure during which time the silica spheres shrank (750–800 nm). The BCl3 containing solutions were placed in a 20 ml syringe and injected into the heated tube by means of a SAGE syringe pump (0.067 ml min1 injection rate) over 4 h. The solutions were injected into the tube reactor (20 cm  2 cm) via a specially designed quartz tube cooled by water [24]. When the solution injection was complete, the electrical furnace temperature was allowed to cool down to room temperature under an Ar flow. Black carbon/silica sphere composites were obtained. Removal of the silica using 20 wt% HF solution (15 ml HF was added to the 1.5 g carbon/silica composite and the mixture aged for 48 h) yielded hollow carbon spheres (B-HCSs). Before use as a support, the carbon supports were functionalized with concentrated HNO3 solution at 90 °C for 8 h. The B-CNTs were synthesized by the catalytic chemical vapor deposition (CCVD) method. The catalyst used in this study was Fe/Al/MgO [22] with a molar ratio of 1:1:12. A quartz boat containing 300 mg catalyst powder was inserted into the center of a quartz tube reactor. The reactor was heated at 900 °C in 5% H2/Ar at a flow rate of 240 ml/min. At the required temperature, the 1 M BCl3 in toluene solution was introduced into the reactor by means of a 20 ml syringe driven by a syringe pump at 1 ml/min1 for 20 min. 2.4. Catalyst preparation To a mixture of Pd(CH3CO2)2 (53 mg) and toluene (100 ml) in a 200-ml two-necked flask at 80 °C was added pyridine (50 mg)

37

during which time the brown suspension turned yellowish white to give a slurry containing [Pd(py)2(ac)2] (py = pyridine; ac = acetate) [25,26]. Then, the B-HCS support (1 g) was added to the Pd slurry, and the mixture was stirred vigorously for 1 h at 80 °C to give a 2.5% Pd-loaded material. The obtained mixture was cooled to 0 °C, followed by filtration and washing with diethyl ether (20 ml) three times. The resulting solid was dried in vacuo at room temperature. This catalyst was called Pd/B-HCS. The catalyst was calcined for 3 h at various temperatures 80, 120, 180, 300, 400, 450, 500, 550, 600, and 650 °C and the calcined materials called Pd/B-HCS80, Pd/B-HCS120, Pd/B-HCS180, Pd/B-HCS300, Pd/BHCS400, Pd/B-HCS450, Pd/B-HCS500, Pd/B-HCS550, Pd/B-HCS600, and Pd/B-HCS650, respectively. Two other catalysts were prepared by the same procedure as described above but using silica spheres and B-CNTs as supports. The material generated from silica spheres and B-CNTs calcined at 300 °C were called Pd/SiS and Pd/B-CNT. Also, 1, 2, and 3 wt% palladium-loaded catalysts were prepared by the same method and called Pd/B-HCS1, Pd/B-HCS2 and Pd/BHCS3, respectively. A catalyst was also made from an aqueous solution of the Pd(CH3CO2)2 added to the B-HCS by incipient wetness impregnation. This sample was dried at 100 °C overnight and finally calcined at 300 °C for 3 h in air and named Pd/B-HCSImp. 2.5. Catalyst characterization The chemical analysis of the catalysts was performed by using the ICP-OES technique to measure the Pd concentration (Spectro Genesis instrument). Approximately 50 mg of catalyst was used. The carbon was oxidized at 700 °C, and the samples dissolved in 5 ml HF, 4 ml HNO3, and a few drops of HClO4. This procedure was repeated twice after the acid evaporation. Finally, the residue was dissolved in aqua regia and heated until dryness, and an aqueous solution of HNO3 (1%v/v) was then added and the solution transferred to a 50 ml volumetric flask for analysis. A JEM 100s, an FEI Tecnai G2 Spirit and an FEI Tecnai F20 X-Twin at 200 kV FEG with an Oxford EDS system were used for transmission electron microscopy (TEM) studies. All samples were ultrasonically suspended in methanol and a drop of the suspension was transferred to a copper grid and allowed to dry before TEM analysis. The phase composition of the support and catalysts were determined by means of XRD analysis on a Bruker D2 phaser in Bragg Brentano geometry with a Lynxeye detector using Cu Ka radiation at 30 kV and 10 mA. The scan range was 10° < 2h < 90° in 0.040 steps, using a standard speed with an equivalent counting time of 1 s per step. The diffraction peaks were then compared with those of standard compounds reported in the Diffracplus evaluation package using the EVA (V11.0, rev.0, 2005) software package. A Perkin Elmer TG/DTA Thermo gravimetric analyzer was used to measure weight changes of samples heated in air or nitrogen at a constant heating rate of 10 °C/min. The sample mass used was varied between 0.005 and 0.01 g. Scanning electron microscopy (SEM) images were recorded using a Philips XL-30 instrument coupled to an energy dispersion unit using an EDX Link Analytical QX-20000 an accelerating voltage of 5 kV. The samples were mounted on a copper stub with conductive carbon sticky tape. A thin (ca. 5 nm) coating of gold was deposited onto the samples to reduce the effects of charging. Raman spectra of the spent catalysts were obtained on a T64000 Raman spectrometer (Jobin Yvon triple spectrometer) under ambient conditions. A 514.5 nm Ar laser was used as the exciting source with a power density of 1 mW cm2 on the sample surface and a power of 2 mW. The measurements were referenced to Si at 521 cm1 with 16 data acquisitions in 180 s. A low laser power was used because a higher laser power burned the samples. Boron nuclear magnetic resonance (11B-NMR) spectroscopy was utilized to investigate the chemical

38

V. Ravat et al. / Journal of Catalysis 305 (2013) 36–45

environment of B in the B-HCS samples. All 11B MAS NMR spectra were recorded with a BRUKER probe head with a zirconium rotor of 4 mm diameter in a Bruker AVANCETM 500 MHz instrument. Boric acid was chosen as reference at 0 ppm for the 11B-NMR spectra. The elemental composition of Pd/B-HCS400 catalyst was determined by X-ray photoelectron spectroscopy (XPS) using a PHI5600 spectrometer equipped with a monochromatic Al Ka source (1486.6 eV).

2.6. Catalytic oxidation reaction The liquid-phase oxidation of benzyl alcohol over the supported Pd catalysts was carried out in a magnetically stirred two neck flask (25 ml), fitted with a thermometer (for measuring the reaction temperature) and a reflux condenser. The reaction mixture containing benzyl alcohol (30 mmol) and catalyst (75 mg) was placed in the flask, while oxygen from an O2 balloon was bubbled into the reaction mixture heated at 125 °C with vigorous stirring for 5 h. At the end of the reaction, the mixture was filtered, the solid catalyst was washed with acetone, and the reaction products and unconverted reactants from the filtrate were analyzed by gas chromatography with an FID, using a capillary column (Phenomenex Zobron, 30 m  .53 mm I.D) and N2 as the carrier gas. All alcohol oxidation products were analyzed using the same GC and capillary column.

3. Results and discussion 3.1. Support characterization Figs. 1 and 2 show SEM and TEM images of the silica and B-HCS supports, respectively. The silica spheres used were Stöber spheres (d = 1000–1200 nm; 1100 nm average) made using a classical procedure (Fig. 2a) [27,28]. The spheres were placed in a CVD oven heated to 900 °C for 4 h during which time the spheres shrank (d = 750–800 nm). BCl3 in toluene was then introduced into the reactor and flowed over the spheres in a CVD process (900 °C/ 4 h/Ar). The material produced was shown by SEM analysis to consist of accreted spheres (d = 800 nm) that generated sphere clusters (Fig. 1a). Removal of the silica in the spheres was achieved by using a solution of HF (20 wt%, 15 ml HF) added to 1.5 g of the carbon/silica composite. The mixture was aged for 48 h and yielded hollow carbon spheres (HCSs) which have rough surfaces (Fig. 1b). Finally, the B-HCSs were treated with nitric acid to give the functionalized B-HCSs and Pd/B-HCS catalyst (Fig. 1c and d). The SEM images do not provide information as to the removal of the silica; this is provided by TEM studies (see below). The purification processes did not destroy the macroscopic structure of the B-HCSs. TEM studies also showed that the silica spheres had a smooth surface morphology (Fig. 2a). Addition of BCl3 in toluene via a CVD injection process (4 h) gave monodispersed core/shell spheres with an intact rough external surface of boron/carbon with an

Fig. 1. SEM images of (a) B-HCSs obtained after CVD of carbon at 900 °C for 4 h, (b) B-HCS after silica removal using HF solution, (c) B-HCSs obtained after activation in concentrated HNO3, and (d) Pd/B-HCS catalyst calcined at 250 °C in air.

V. Ravat et al. / Journal of Catalysis 305 (2013) 36–45

39

Fig. 2. TEM images of (a) Stöber spheres of 1000–1200 nm, 1100 nm diameter, (b) B-HCSs obtained after removal of silica by treatment with HF solution, and (c and d) B-HCSs after activation of HCSs in concentrated HNO3 solution (inset in Fig. 2b shows the TEM image of the B-HCS after silica removal at higher resolution).

external diameter of about 800 nm (not shown). The B-HCS sample was treated with HF solution which removed the silica and gave a hollow structure with an average shell thickness of about 100 nm (Fig. 2b). The inner diameter of the empty B-HCSs was estimated to be around 600 nm, smaller than that of the diameter of the boron-doped carbon spheres (800 nm). The decrease in the size of BHCSs is due to silica sphere shrinkage at 900 °C. The shell thickness depends on the silica sphere diameter, the carbon precursor, reaction temperature, and reaction time [29]. After activation in HNO3 (Fig. 2c and d), the shell thickness and external morphology of the B-HCS surface remained unchanged. A TEM image of the Pd/BCNTs is shown in the Supplementary section (Fig S1) at low resolution. The chemical environment of the boron in the B-HCSs was studied by MAS (magic angle spinning) NMR spectroscopy (Fig. 3). In the 11B MAS NMR spectrum, a dipolar interaction occurs through a homonuclear B–B interaction, whereas under the MAS condition, the heteroatom 13C has a very small nuclear spin and a negligible interaction with B. Though there is a possibility for a second-order quadrupole interaction due to 11B (I = 3/2), the MAS eliminates this secondary quadrupole interaction. It hence does not contribute to the line shape [30]. Two B chemical environments with broad peaks (70–80 ppm) were observed (Fig. 3). The data suggest that the boron atoms are bonded to carbon (or oxygen) atoms in two environments, and there is no possible quadrupole interaction due to the presence of a B–B bond. The 11B NMR spectra of the

Fig. 3. 11B MAS NMR spectra of (a) B-HCS before silica etching, (b) B-HCS after silica etching, (c) B-HCS after functionalization, and (d) B-HCS calcined at 650 °C samples.

as-synthesized materials before silica etching gave one major peak (d = 5 ppm) along with one small shoulder (d = 70–80 ppm). The removal of silica did not affect the spectrum (Fig. 3b). When the BHCSs were functionalized with a strong oxidizing agent (concentrated HNO3), a small upfield shift (d = 9 ppm) was observed for

40

V. Ravat et al. / Journal of Catalysis 305 (2013) 36–45

Fig. 4. Powder XRD patterns of (a) Pd(py)2(ac)2, (b) Pd/B-HCS120, (c) Pd/B-HCS180, (d) Pd/B-HCS300, (e) Pd/B-HCSI, (f) Pd/B-B-CNT (g) Pd/B-CNT400, and (h) B-HCS650 samples [(- O -) B2O3, (- d -) PdO, and (- . -) Pd].

Fig. 5. TG/DTA profiles of (a) Pd(py)2(ac)2, (b) Pd/B-HCS80, (c) Pd/B-HCS120, (d) Pd/B-CNT, (e) B-HCS, and (f) HCS samples.

the B-HCS. The calcination of B-HCS at 650 °C generated B2O3 (confirmed by XRD and Raman data) which gave an upfield shift (d = 16.1 ppm). This shift to higher values is consistent with the presence of the high electron withdrawing oxygen (in B–O) when it replaces the carbon in the structure [30].

An X-ray diffraction pattern showed that the B-HCSs were similar to the HCSs and B-CNTs as shown in Fig. 4 and Fig. S2 (Supporting information). TG/DTA analysis was carried out in an air atmosphere in the temperature range 20–970 °C on the B-HCS and B-CNT materials

V. Ravat et al. / Journal of Catalysis 305 (2013) 36–45

as shown in Fig. 5e and S2 (supporting information). In B-CNT, an initial weight loss was observed in the range between 400 and 500 °C and a broad weight loss between 600 and 850 °C. These weight losses were well supported by two endothermic DTA peaks at 412 °C and 600 °C. The HCS support had a maximum rate of oxidation at 750 °C, and as expected, the addition of B in the HCSs (and CNTs) resulted in the oxidation of carbon shifting to a higher temperature (Fig. 5e and f). The Raman spectra of the B-HCSs and B-CNTs after activation with HNO3 (Fig. 6a, b, and f) showed peaks at 1358 cm1 (D band) due to the disorder-induced phonon mode (breathing mode, A1g band) and at 1594 cm1 (G band) assigned to the Raman allowed phonon mode (E2g band). In summary, the B-HCS and B-CNT materials show classical properties indicative of the presence of boron. Both materials have hollow interiors and are easy to synthesize. The B-HCSs have the advantage over the B-CNTs in that they are not made with a metal catalyst. This means that they do not contain catalyst impurities that are impossible to remove from the interior of the tubes. Addition of Pd to the supports was then monitored. The data revealed that there were two types of Pd materials formed on the supports – those formed below and above a calcination temperature of 300 °C.

3.2. Characterization of the Pd complexes at low calcination temperatures The XRD pattern of the Pd complex [Pd(py)2(ac)2] is shown in (Fig. 4a). After [Pd(py)2(ac)2] was added to the B-HSCs, XRD patterns were recorded. The XRD pattern of the Pd/B-HCS120 material shows reflections indexed to peaks associated with [Pd(py)2(ac)2] deposited on the B-HCS surface (Fig. 4b). The XRD pattern of Pd/ B-HCS180 (Fig. 4c) gave diffraction peaks associated with carbon as well as a Pd complex different from that of [Pd(py)2(ac)2]. This conversion was also confirmed by TGA data (Fig. S3b). This indicates partial decomposition of the complex at 180 °C. By 300 °C,

41

complete decomposition of all Pd complexes had occurred to give PdO (see Fig. 4d). The TG/DTA profiles of [Pd(py)2(ac)2] (Fig. 5a) showed that the decomposition of the Pd complex (to PdO) took place through two endothermic reactions. The [Pd(py)2(ac)2] complex did not show any significant loss in weight up to 150 °C. A total weight loss of 75% was then observed. An initial weight loss of 50% occurred in the range 150–195 °C, followed by a weight loss of 25% in the range 195–230 °C. The weight loss corresponded to the loss of the ligands as the complex was converted to PdO. The same decomposition patterns were observed for Pd/B-HCS80 and Pd/B-HCS120 catalysts in the range of 150–230 °C, and the data indicated that not all the [Pd(py)2(ac)2] had decomposed by 120 °C, consistent with the XRD data. 3.3. Characterization of the Pd complexes at high calcination temperatures TEM images of the Pd/B-HCS, Pd/B-HCSImp, Pd/SiS and Pd/BCNT catalysts, after a high calcination temperature, are shown in Fig. 7a–d. The images reveal the hollow structure of the B-HCS and B-CNT supports. The Pd average particle sizes were 2.4, 8.0, 8.9, and 4.0 nm for Pd/B-HCS, Pd/B-HCSImp, Pd/SiS, and Pd/BCNT, respectively. The size of the palladium particles varies with the supports, and the size distribution is narrow for the Pd/BHCS and Pd/B-CNT catalysts when the deposition method was used. The results indicate that both B-HCSs and B-CNTs act in a similar fashion to give the Pd-supported materials, indicating a similar Pd–support interaction. These diameter values are consistent with the results from the XRD analysis, indicating a crystalline nature for these nanoparticles. The impregnation method to give Pd/B-HCSImp and the deposition method to give Pd/SiS gave larger Pd particle sizes with a broader size distribution. An X-ray diffraction study of the Pd/B-HCS, Pd/B-CNT, and BHCS samples calcined in air at different temperature was performed to establish the presence of boron in the sample, the thermal stability of the carbon, and the nature of the Pd-active species.

Fig. 6. Raman spectra of (a) B-HCS after silica etching, (b) B-HCS after HNO3 activation, (c) Pd/B-HCS600, (d) Pd/B-HCS650, (e) Pd/HCS600, and (f) B-CNT samples.

42

V. Ravat et al. / Journal of Catalysis 305 (2013) 36–45

Fig. 7. TEM images of (a) Pd/B-HCS300, (b) Pd/B-HCSImp, (c) Pd/SiS, and (d) Pd/B-CNT catalysts samples (20 nm of Pd/SiS and Pd/B-CNT images also shown to indicate Pd particle size 20 nm scale bar).

The XRD pattern of Pd/B-HCS300 indicated that the Pd particles were very small and well dispersed on the B-HCSs; only the carbon peaks could be seen (Fig. 4d). The XRD patterns revealed that PdO (JCPDS file No-00-043-1024) was observed at all temperatures above 300 °C. XPS data were recorded on the Pd/B-HCS400 sample and revealed only the presence of PdO and PdOx (no Pd) species (see Supp. Figs. S4 and S5). Interestingly, Pd metal (JCPDS file No-01-088-2335) was also observed at T > 400 °C (see Supp. Fig. S5). This suggests that the carbon support can reduce the Pd at high temperatures, even in an oxygen environment. At T > 400 °C, it has been reported that C reacts with the oxygen attached to Pd and releases it as CO2 at high temperature [4]. The reduction of Pd in oxygen has been reported previously, in this instance in the presence of CO as reducing agent and a higher oxygen concentration [31]. A similar result was also reported in the literature by Chen [32]. At a calcination temperature of 650 °C, the diffraction peaks due to B2O3 (JCPDS file No00-013-0570) were clearly observed in the B-HCS sample (Fig. 4h), indicating that any B-C bonds had been converted to B– O bonds and that B2O3 had formed. No indication of other new phases or complexes, such as boron carbides, was observed. The XRD profile of B-HCS revealed that carbon was still present in the material after O2 treatment at 650 °C.

The Pd/B-CNT samples were also analyzed by XRD. At high temperatures, PdO was detected in the catalysts [Pd/B-CNT (300 °C, 4.2 nm) and Pd/B-CNT (400 °C, 6.6 nm)] (Fig. 4f and g). As found for the Pd/B-HCS samples, Pd metal was detected at T > 400 °C. The Pd/B-HCSImp sample (Fig. 4f) showed a peak in the XRD due to PdO (8 nm) indicating that the catalyst prepared by the impregnation method showed a lower dispersion as compared to the catalyst prepared by the deposition method. Doping of B in HCSs and CNTs is expected to increase both the dispersion of Pd as well as the thermal stability of the support. TGA data were recorded on Pd/B-CNT (Fig. 5d). The TGA recorded on Pd/B-CNT (i.e., calcined at 300 °C) showed a new oxidation peak at ca 425 °C as well as the carbon oxidation peak which now occurred in a broad temperature range (450–800 °C). Addition of the Pd results in some carbon catalyst oxidation, but this occurs at high temperatures (Fig. 5b–d). A Raman study of Pd/B-HCS samples calcined in air at different temperatures was performed to establish the presence of boron in the samples. Indeed, the Raman spectra recorded on Pd/B-HCS600 and Pd/B-HCS650 show a sharp peak at 878 cm1 associated with B2O3. The peak observed at 646 cm1 is due to PdO (Fig. 6d). The spectra also show that the carbon is oxidized at the higher calcination temperature. A comparison with Pd/HCS600 (that contains no

43

V. Ravat et al. / Journal of Catalysis 305 (2013) 36–45

B) indicates how the B retards the carbon oxidation reaction (Fig. 6f). In summary, the Pd complex supported on B-HCSs (or B-CNTs) started decomposing at a calcination temperature of 180 °C. At higher temperatures, this complex decomposed and was converted to PdO. At temperatures >400 °C, the phase due to Pd was observed on the carbon support. 3.4. Catalytic activity study The Pd-catalyzed benzyl alcohol to benzaldehyde reaction was chosen as the test reaction to evaluate the effect of the boron-containing carbon supports. 3.4.1. Study of catalysts calcined <200 °C Reaction was performed on catalysts calcined at T between 120 °C and 200 °C (Table 1, entry 3). Excellent results (100% conversion; selectivity >99%; 5 h) were obtained at all temperatures. A mixture of Pd complexes in the Pd2+ oxidation state (including PdO and [Pd(py)2(ac)2]) is expected to co-exist on the B-HCS surface in this catalyst calcination regime. The activities of these catalysts decreased when used in further (recycled) reactions (Table 1). After repeated use of Pd/B-HCS120 (6 cycles), elution of Pd was detected and this led to reduced activity even though the selectivity (99%) was retained. The hot filtrate test also indicated some activity associated with the eluted Pd sample (Table 1). While the catalyst performance at the lower calcination temperature (120 °C) is good, elution of the complex will limit the potential of these catalysts. A possible mechanism for the reaction has been reported by others [33]. 3.4.2. Study of catalysts calcined >200 °C 3.4.2.1. Effect of reaction temperature. Fig. 8 presents the influence of reaction temperature on the catalytic activity of Pd/B-HCS under

standard reaction conditions (see Section 2). It can be seen from the figure that at lower temperatures, the conversion of alcohol is low but the selectivity is good. However, as the reaction temperature is raised, an improvement in activity is noticed while the selectivity of benzaldehyde decreases.

3.4.2.2. Effect of catalyst amount. The reaction was also carried out by taking different amounts of catalyst (Fig. 9). It can be inferred from Fig. 9 that the benzyl alcohol conversion increases with increase in catalyst amount. However, diffusion constraints occur for a mass >0.075 g catalyst. In later studies, the temperature (125 °C), time (5 h), and 0.75 g of catalyst were used to ensure that diffusion effects were not responsible for the catalyst behavior.

3.4.3. Effect of Pd content on the activity of the Pd/B-HCS catalyst Four different Pd loadings (1, 2, 2.5, and 3 wt%) were used to study the effect of catalyst loading on the oxidation of benzyl alcohol. The results in Fig. 10 show that increasing the Pd loading from 1 to 2.5 wt% in Pd/B-HCS catalyst, the benzyl alcohol oxidation activity increased markedly (same selectivity). For a Pd loading of 3 wt%, diffusion effects were again observed.

3.4.4. Effect of catalyst preparation method The results in Table 1 show an influence of the method of Pd catalyst preparation on the catalyst performance in the reaction. The catalyst Pd/B-HCSImp studied under optimized conditions gave 80% conversion of benzyl alcohol with 99% selectivity to benzaldehyde. The catalyst, Pd/B-HCS300, gave complete conversion (100%) of benzyl alcohol into benzaldehyde (selectivity >99%). The excellent results were also observed for Pd/B-CNTs. The higher activity of Pd/B-HCS (and Pd/B-CNT) is related to the smaller particle size of Pd (TEM and XRD results) in these catalysts.

Table 1 Surface area, Pd content, particle size, conversion, selectivity, and activity of different Pd catalysts in the benzyl alcohol to benzaldehyde oxidation reaction. Sample

Pd cont.%

1 2

B-HCS Pd/B-HCS (2.4 nm) 6th recycle Hot filtrate test Pd/B-HCS120 6th recycle Hot filtrate test Pd/B-HCS @200 (2.3 nm) @ 400 °C (3.7 nm) @ 450 °C (4.5 nm) @ 500 °C (5.5 nm) @ 550 °C (6.7 nm) Pd/B-HCS300 1 wt% loading 2 wt% loading 3 wt% loading Pd/B-HCSImp (8 nm) SiS Pd/SiS (8.9 nm) @ 400 °C (10.5) @ 500 °C (11.7) B-CNT Pd/B-CNT (4.2 nm) 6th recycle Hot filtrate test @ 400 °C (5.4 nm) @ 450 °C (6.0 nm) @ 500 °C (6.8 nm)

– 2.45 2.17 (12%)

30 22 25

2.42 1.65 (32%)

19 19

2.45 2.48 2.53 2.53 2.53 0.97 1.92 2.89 2.25 – 2.43

3

4

5

6 7 8

9 10

a

SBET (m2/g)

Entry

Conver. (%)

Select. (%)

<2 100 90 4 100 61 14

89 99 99 98 99 99 98

22 25 22 20 19

100 94 81 75 62

99 98 99 98 98

4.05 3.79 3.21 2.96 2.55

2.45 2.45 2.45

136 125 120

61 80 100 80 <2 75 64 45 <2 98 91 4 87 70 62

99 99 99 99 85 96 95 94 82 99 99 99 99 99 99

5.49 4.16 3.46 3.33

2.45 2.19 (10%)

28 27 15 23 5 3 4 5 150 140 135

2.43

1

Activity ðmol g1 ðPdÞ h

Þ

4.05 4.20 4.13 3.71

3.15 2.63 1.89 4.08 4.16 3.57 2.85 2.51

Benzyl alcohol = 30 mmol; catalyst weight = 75 mg; T = 125 °C; oxidant = O2; time = 5 h., Pd content after 5th recycle and Pd content calculated by ICP-OES method. Pd/BHCS, Pd/B-CNT, and Pd/SiS calcined at 300 °C.

44

V. Ravat et al. / Journal of Catalysis 305 (2013) 36–45

3.4.5. Effect of calcination temperature The results in Fig. 11 show a strong influence of calcination temperature on the performance of the catalyst in the benzyl alcohol oxidation reaction for the Pd/B-HCS catalysts. With increasing calcination temperature, both the catalytic activity (benzyl alcohol conversion) and benzaldehyde selectivity remain constant until a calcination temperature at 300 °C. Above this temperature, the conversion started to decrease. The lower performance of the catalyst at higher calcination temperatures relates to the increased Pd particle size. Thus, for Pd/B-HCS, the size increases with calcination temperature [300 °C (2.4 nm), 400 °C (3.7 nm), 450 °C (4.5 nm), 500 °C (5.5 nm), and 550 °C (6.7 nm)]. Similar results were found for Pd/SiS [300 °C (8.9 nm), 400 °C (10.5), and 500 °C (11.7)] and Pd/B-CNT [300 °C (4.2 nm), 400 °C (5.4 nm), 450 °C (6.0 nm), and 500 °C (6.8 nm)].

Fig. 8. Influence of the reaction temperature and time on the conversion and selectivity in the solvent-free oxidation over the Pd/B-HCS300 catalyst.

Fig. 9. Effect of the amount of Pd/B-HCS300 catalyst (calcined at 300 °C) on the conversion and selectivity in the solvent-free oxidation of benzyl alcohol.

Fig. 10. Effect of the palladium loading in the Pd/B-HCS300 catalyst on the conversion in the solvent-free oxidation of benzyl alcohol at optimized conditions (125 °C/O2/5 h/75 mg).

3.4.6. Effect of supports The use of different supports at optimized reaction condition was studied. No conversion was found for the B-HCS, SiS, or BCNT supports (Table 1, entries 1, 7 and 9) that did not contain Pd. The results of the catalytic studies (Table 1) show that Pd on the different supports leads to different conversions but with the same selectivity [Pd/B-HCS (ca. 100%, entry 2), Pd/SiS (ca. 75%,

Fig. 11. Influence of the calcination temperatures of the Pd/B-HCS300 catalyst on the solvent-free oxidation of benzyl alcohol at optimized conditions (125 °C/O2/5 h/ 0.75 g).

Fig. 12. Effect of particle size on the activity of benzyl alcohol oxidation reaction.

V. Ravat et al. / Journal of Catalysis 305 (2013) 36–45

entry 8) and Pd/B-CNT (ca. 100%, entry 10)]. Among the Pd catalysts studied on the different supports, the Pd/B-HCS and Pd/BCNT catalysts showed the best performance (100% conversion; 99% selectivity) in the oxidation of benzyl alcohol to benzaldehyde. The lower activity of the Pd/SiSs support relates to the larger Pd particle size (8.9 nm) as compared to those found on other supports as well as the nature of support. This is clearly seen in Fig. 12. 3.4.7. Effect of Pd particle size The mean size of the Pd particles was controlled by both the support and calcination temperatures. The activity for the oxidation of benzyl alcohol reaction relates to the particle size of Pd (Figs. 11 and 12). By changing the support, Pd nanoparticles with mean sizes ranging from 2.2 to 9 nm were obtained. Table 1 (entries 9 and 4) shows that Pd/B-HCS and Pd/B-CNT catalysts gave remarkably better catalytic performances than the Pd/SiS catalyst.

45

10 nm, the catalyst activity decreased indicating the role of Pd particle size on the reaction. The data also suggested that similar results can be obtained from both hollow carbon spheres and CNTs (both with unfilled cores). However, it is to be noted that the HCSs do not contain metal particle impurities. Thus, any residual Fe catalyst particles from the CNT synthesis played no role in the reaction. While this is true for this reaction, it may not be true in general. Acknowledgments We thank the NRF, the University of the Witwatersrand and the DST/NRF Centre of Excellence in Strong Materials for financial support. The authors would like to acknowledge Dr. Richard Mampa for assisting with the solid-state NMR measurements. Appendix A. Supplementary material

3.5. Catalyst durability (recycling test) and reaction mechanism An important property of heterogeneous catalysts is their durability. Pd/B-HCS120, Pd/B-HCS300, and Pd/B-CNT catalysts were tested for leaching by reusing the catalysts. Each additional reaction run was carried out under identical conditions, and reuse of the recovered catalyst was performed for another 5 h after the catalyst had been removed from the reaction mixture by filtration. After 6 runs, the selectivity did not change but the conversion had declined by about 45% in the case of Pd/B-HCS120 (Fig. 3a), 10% in the case of Pd/B-HCS300 (Fig. 3b) and 9% for Pd/B-CNT catalysts. The reason for this decline was different for the different catalysts. The catalyst Pd/B-HCS120 undergoes complex elution and also decomposition during the oxidation reaction. For the Pd/ B-HCS300 (Fig. 3b) and Pd/B-CNT catalysts after 6 runs, the selectivity did not change but conversion had decreased by about 10%. The reason for the activity loss for Pd/B-HCS and Pd/B-CNT catalysts is only due to leaching of Pd from the catalyst surface. Doping of B in Pd/B-HCS and Pd/B-CNT enhances the interaction between palladium and the support and reduced the leaching effect relative to the undoped carbons. This is supported by ICP analysis of the used catalysts; oxidation of the carbonaceous material revealed leaching of Pd (12% for Pd/B-HCS and 10% for Pd/B-CNT). 4. Conclusions The preparation of graphitizable hollow carbon spheres (HCSs) with B-doped single shells was achieved from Stöber silica spheres as a template and 1 M BCl3 in toluene as B and C sources. A CVD temperature at 900 °C resulted in a higher graphitic nature of the B-HCS. Doping of B into the HCSs enhanced the crystallinity and thermal stability of the B-HCS. An optimized conversion of benzyl alcohol to benzaldehyde was observed when catalysts were calcined at 300 °C for 5 h at 125 °C) with a 2.5 wt% Pd loading. The Pd/B-CNT catalyst also gave similar activity to that obtained for the Pd/B-HCS. The Pd/B-HCS (and Pd/B-CNT) catalysts are highly promising, easily separable and reusable catalysts for the solvent-free selective oxidation reaction. The mean size of Pd particles and supports played a key role in the aerobic oxidation of benzyl alcohol. On changing the mean size of Pd particles from 2.2 to

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jcat.2013.04.018. References [1] B. Frank, A. Rinaldi, R. Blume, R. Schlogl, D.S. Su, Chem. Mater. 22 (2010) 4462. [2] C.K. Acharya, W. Li, Z. Liu, G. Kwon, C.H. Turner, A.M. Lane, D. Nikles, T. Klein, M. Weaver, J. Power Sources 192 (2009) 324. [3] A.A. Koo, F. Dillon, E.A. Obraztsova, A. Crossley, N. Grobert, Carbon 48 (2010) 3033. [4] V. Ravat, I. Nongwe, N.J. Coville, ChemCatChem 4 (2012) 1930. [5] G.X. Chen, J.M. Zhang, D.D. Wang, K.W. Xu, Physica B 404 (2009) 4173. [6] H.M. Murty, D.L. Biederman, E.A. Heintz, Fuel 56 (1977) 305. ¯ ya, R. Yamashita, S. O ¯ tani, Fuel 58 (1979) 495. [7] A. O [8] W. Cermigani, T. Paulson, C. Onneby, C.G. Pantano, Carbon 33 (1995) 367. [9] W. Han, Y. Bando, K. Kurashima, T. Sato, Chem. Phys. Lett. 299 (1995) 368. [10] X. Ma, Q. Wang, L.Q. Chen, W. Chermignani, H.H. Schobert, C.G. Pantano, Carbon 35 (1997) 1517. [11] L.R. Radovic, M. Karra, K. Skokova, P.A. Thrower, Carbon 36 (1998) 1841. [12] X. Blase, J.C. Charlier, A.D. Vita, R. Car, Phys. Rev. Lett. 83 (1999) 5078. [13] D.Q. Yang, G.X. Zhang, E. Sacher, M.J. Yacaman, N. Elizondo, J. Phys. Chem. B 110 (2006) 8348. [14] D.Q. Yang, E. Sacher, Chem. Mater. 18 (2006) 1811. [15] H.N. Murty, D.L. Biederman, E.A. Heintz, Fuel 56 (1977) 305. [16] S. Marinkovic, C. Suznjevic, I. Dezarov, Carbon 7 (1969) 185. [17] W.V. Kotlensky, Carbon 5 (1969) 409. [18] P. Serp, M. Corrias, P. Kalck, Appl. Catal. A: Gen. 253 (2003) 337. [19] C.J. Lee, S.C. Lyu, H.W. Kim, J.H. Lee, K.I. Cho, Chem. Phys. Lett. 359 (2002) 115. [20] R.S. Oosthuizen, V.O. Nyamori, Plat. Metals Rev. 55 (2011) 154. [21] E.N. Nxumalo, V.O. Nyamori, N.J. Coville, J. Organomet. Chem. 693 (2008) 2942. [22] A. Bachmatiuk, F. Schaffel, D. Placha, G.S. Martynkova, N. Ioannides, T. Gemming, Fuller., Nanotubes, Carbon Nanostruct. 18 (2010) 37. [23] W. Stöber, A. Fink, E. Bohn, J. Colloid Interface Sci. 26 (1968) 62. [24] X.Y. Liu, B.C. Huang, N.J. Coville, Fuller., Nanotubes, Carbon Nanostruct. 10 (2002) 339. [25] S.V. Kravtsova, I.P. Romm, A.I. Stash, V.K. Belsky, Acta Crystallogr. Sect. C 52 (1996) 2201. [26] T.A. Stephenson, S.M. Morehouse, A.R. Powell, J.P. Heffer, G. Wilkinson, J. Chem. Soc. (1965) 3632. [27] J.H. Kim, B. Fang, M. Kim, J.S. Yu, Catal. Today 146 (2009) 25. [28] F. Su, X.S. Zhao, Y. Wang, L. Wang, J.Y. Lee, J. Mater. Chem. 16 (2006) 4413. [29] B. Ottaviani, A. Derre, Q.A.M. Grivei E Mahmoud, M. Guimon, S. Flandrois, P. Delhae, J. Mater. Chem. 8 (1998) 197. [30] K.J. Prabakar, K.J. Rao, C.N. Rao, Proc. Roy. Soc. Lond. A 429 (1990) 1. [31] P.V. McKinney, J. Am. Chem. Soc. 54 (1932) 4498. [32] W. Chen, X. Pan, M-G. Willinger, D. Su and X Bao, J. Am. Chem. Soc. 128 (2006) 3136–3137. [33] S.S. Stahl, B.A. Steinhoff, Org. Lett. 4 (2002) 4179.