The direct synthesis of hydrogen peroxide from H2 and O2 using Pd–Ga and Pd–In catalysts

The direct synthesis of hydrogen peroxide is investigated using PdGa/TiO2 and PdIn/TiO2 catalysts prepared by an acid-washed sol-immobilisation procedure, which allows for good control of particle size. The introduction of both Ga and In into a supported Pd catalyst leads to significantly reduced rates of H2O2 degradation in comparison to the monometallic counterpart. Detailed characterisation reveals that this enhancement is a result of selectively tuning the ratio of Pd oxidation states.


Introduction
Hydrogen peroxide (H 2 O 2 ) is a powerful, green oxidant that finds use primarily in the textile/paper and pulp industry and in the chemical synthesis sector. 1 This is in part attributed to the growing demand for propylene oxide which is used primarily in the production of polyurethane plastics, produced via the integrated HPPO process. In addition, H 2 O 2 finds application in the treatment of waste streams, where it is superseding chlorine containing oxidants, primarily due to increasing environmental legislation. 1 In recent years, global H 2 O 2 production has exceeded 3.3 million tons per annum 2 and is predicted to rise annually at a rate of approximately 4% with demand forecast to reach 5.2 million tons per annum by 2020. 3 This increase in demand in part can be related to the growing application of H 2 O 2 in the chemical synthesis sector; in particular the rise in demand for H 2 O 2 can be related to its use in the production of both propylene oxide, via the integrated HPPO process, and cyclohexanone oxime, which are key intermediates in the production of polyurethane and nylon-6 respectively. Other significant applications of H 2 O 2 are found in, but not limited to, alkene epoxidation, 4,5 organo-sulphur oxidation 6,7 and ketone oxidation. 8,9 Currently, the production of H 2 O 2 on an industrial scale is met via the well-established anthraquinone oxidation (AO) process. Although highly efficient the underlying chemistry has changed little since first developed by BASF, utilising an anthraquinone carrier molecule, H 2 and O 2 with the former reduced over a Pd-based catalyst producing a diol which is oxidised to produce H 2 O 2 and regenerate the anthraquinone. Initially relatively low concentrations of H 2 O 2 (0.8-1.9 wt%) are produced, with this raised through numerous extraction and purification and distillation steps to yield H 2 O 2 solutions in excess of 70 wt% prior to shipping to the end user. As such the AO process is only economically viable on a large scale. 10 It should be noted that the typical on-site application of H 2 O 2 requires concentrations in the range of 1-10 wt% with dilution prior to use required. Furthermore, the low stability of H 2 O 2 , even at relatively mild temperatures, requires the use of acidic stabilizing agents to prevent its decomposition to water. Although effective in enhancing H 2 O 2 stability the use of such stabilizing agents adds additional costs associated with decreased reactor lifetime and their removal from product streams. 1,10 The on-site production of H 2 O 2 via the direct combination of H 2 and O 2 would offer an attractive alternative to the AO process, allowing for production of H 2 O 2 at appropriate concentrations and alleviating the additional costs associated with the shipping and storage of H 2 O 2 . The direct synthesis of H 2 O 2 represents a greener, more atom efficient process for H 2 O 2 production that can potentially be adopted at point of use. The use of Pd-based catalysts has received the greatest academic focus, with the first patent filed by Henkel and Weber in 1914. 11 However, Pd-based catalysts often suffer from poor selectivity and require the use of acidic or halide stabilizing agents to limit the production of H 2 O via decomposition and hydrogenation pathways. Building on the work of Landon et al. 12 and Haruta and co-workers, 13 who simultaneously demonstrated the activity of Au-based catalysts for the direct synthesis of H 2 O 2 , numerous studies have since focused on AuPd nanoparticles, which have been demonstrated to offer excellent selectivities towards H 2 O 2 in the absence of acidic or halide stabilizing agents. Indeed, Edwards et al. 14 have previously demonstrated that through acidic pre-treatment of a carbon support prior to immobilization of Au and Pd it is possible to reach H 2 O 2 selectivities in excess of 95%, with a similar enhancement subsequently reported when utilizing oxide supports. 15,16 More recently, Freakley et al. 17 have reported that it is possible to replace Au with cheaper, more abundant base-metals such as Sn, Ni and Co, while maintaining excellent selectivity towards H 2 O 2 . Further research has since been placed on the modification of Pd with a range of secondary metals including Sn, 18 Ag, 19 Zn, 20 Ni, 21 Sb 22 and Te 23 in an attempt to enhance catalytic selectivity towards H 2 O 2 .
The use of sol-immobilisation procedures to prepare bimetallic catalysts has been extensively studied in the literature and is known to offer excellent control of elemental composition and mean particle size in comparison to more widely used catalyst preparation techniques such as wet-impregnation. 24,25 The formation of bimetallic alloyed nanoparticles is well known to offer distinct differences in catalytic performance with electronic and chemical properties that are typically very different from those of the monometallic analogues and often display enhanced selectivity, activity, and stability. 26 In particular, the choice of ligand is well known to play a key role in influencing catalytic activity and selectivity. 27,28 Therefore, obtaining "clean" particles, while also controlling particle size and composition has become a major challenge in the case of supported catalysts prepared via colloidal methods. 29 Instead of removing ligands to cause sinter, here we prepared nanoparticles with same method, same type and amount of ligands to limit the effect on catalytic properties.
In this work, we investigate the catalytic activity of colloidal PdGa and PdIn nanoparticles immobilized on a TiO 2 support for the direct synthesis of H 2 O 2 from molecular H 2 and O 2 .

Synthesis of unsupported Pd, Ga, In and bimetallic Pd-Ga and Pd-In nanoparticles
Unsupported Pd-Ga and Pd-In nanoparticles, of varying ratio were prepared with the ratio of Pd : M (M = Ga, In) indicated by the nomenclature used, so that the Pd2Ga has a theoretical molar Pd : Ga ratio of 2 : 1.
All syntheses were performed using standard Schlenk techniques. In a typical synthesis, PdĲacac) 2 and MĲacac) 3 (M = Ga, In) were dissolved in OLAM (40 mL) in a four-neck flask. The synthesis parameter within this work can be seen in (Table S1 †). PdĲacac) 3 (0.6 mmol) and MĲacac) 3 (0.3 mmol) were employed to prepare Pd2M. PdĲacac) 3 (0.45 mmol) and MĲacac) 3 (0.45 mmol) were mixed to prepare Pd1M. PdĲacac) 3 (0.3 mmol) and MĲacac) 3 (0.6 mmol) were used to prepare Pd0.5M. The mixture was flushed with argon, heated from room temperature to 60°C over a period of 5 min and stirred for 30 min. After adding 2 mL TOP, the mixture was heated to 200°C (heating rate 8°C min −1 ) for 30 min with stirring at 400 rpm. Next the temperature was further increased to 300°C (heating rate 8°C min −1 ) for 30 min. After this the solution was cooled to room temperature, the bimetallic Pd-M nanoparticles were precipitated by addition of ethanol and thoroughly purified by dissolution in CHCl 3 , precipitation with ethanol, and centrifugation steps.
Pd nanoparticles were prepared using a similar procedure. PdĲacac) 2 (0.90 mmol) was dissolved in OLAM (40 mL) in a four-neck flask. The mixture was flushed with argon, then heated to 60°C quickly and stirred for 30 min. After adding 2 mL TOP, and the mixture was heated to 200°C (heating rate 9°C min −1 ) which was kept for 30 min. After cooling to room temperature, the Pd nanoparticles were purified as outlined above.
Ga nanoparticles were synthesized in a similar procedure as well. GaĲacac) 3 (0.90 mmol) was dissolved in OLAM (40 mL) in a four-neck flask. The mixture was flushed with argon, then heated to 60°C and kept for 30 min. After adding 2 mL TOP, heat the mixture to 330°C at the rate 9°C min −1 and stirred for 50 min. After cooling to room temperature, collect the samples by washing with CHCl 3 and ethanol.
In nanoparticles were obtained with the typical method. InĲacac) 3 (0.90 mmol) was dissolved in OLAM (40 mL) in a four-neck flask. The mixture was flushed with argon, then heated to 60°C and kept for 30 min. After adding 2 mL TOP, the mixture was heated to 200°C (heating rate 8°C min −1 ) which was kept for another 30 min while stirring. Afterwards, the temperature was further increased to 300°C (heating rate 8°C min −1 ) and kept at 300°C for additional 30 min. After cooling to room temperature, the In nanoparticles were purified as outlined above.

Acid pretreatment of TiO 2 support
The treatment of the support prior to metal particle immobilization is based upon our previous investigations into the role of acid washing both oxide 15,16 and carbon supports prior to immobilization of precious metals. 14 In a typical procedure, TiO 2 support (10 g) was stirred in an aqueous solution of H 2 SO 4 (2 wt%, 100 mL) for 3 hours, then filtered and washed with H 2 SO 4 (2 wt%) solution and dried under vacuum, 16 h at 30°C.
This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

Supported catalyst preparation
Metal nanoparticles were immobilized on the pre-treated TiO 2 support from the colloidal solution by adding TiO 2 (0.4 g) to the appropriate amount of metal nanoparticles in chloroform and stirring for 3 h at 800 rpm. For most catalysts, the colourless supernatant indicated the complete immobilisation of metal nanoparticles. In some cases, where some nanoparticles remained in solution (as indicated by a black coloured supernatant), ethanol (3 mL) was added and the suspension stirred for another 3 h at 800 rpm. The catalysts were recovered by centrifugation, followed by washing with chloroform and ethanol. The final catalysts were dried (30°C, 3 hours) under vacuum and ground to a fine powder. Table S2 † summarizes the total metal loading of the Pd-Ga and Pd-In supported catalysts. Supported catalysts of varying Pd : Ga and Pd : In ratio have been prepared with this ratio indicated by the nomenclature used, so that the Pd2Ga/TiO 2 catalyst has a theoretical Pd : Ga molar ratio of 2 : 1.

Catalyst characterization
Powder X-ray diffraction (XRD) patterns were recorded on a PANalytical X'Pert Pro X-ray diffractometer employing Bragg-Brentano geometry with Cu Kα radiation and a Ni filter. The range between 5 and 120°was measured lasting 16 h. The reflections were compared to reference data reported in the Joint Committee of Powder Diffraction Standards (JCPDS) data base. Pd and Pd-M (M = Ga, In) nanoparticles were precipitated by ethanol from colloidal chloroform solution with centrifugation (7850 rpm, 10 min) and dried at ambient temperature under vacuum. The resulting material was then deposited onto a XRD sample holder under air.
Total metal loading and Pd : M (M = Ga, In) ratio were determined by inductively coupled plasma optical emission spectroscopy (ICP-OES, Agilent 725 ICP-OES Spectrometer). PdM nanoparticles were dissolved in aqua regia. To measure the metal loading of supported catalyst, HF and aqua regia (volume ratio of HF : aqua regia = 2 : 1) were used to dissolve the catalysts.
X-ray photoelectron spectroscopy (XPS) analyses were made on a Kratos Axis Ultra DLD spectrometer. Samples were mounted using double-sided adhesive tape and binding energies were referenced to the C (1s) binding energy of adventitious carbon contamination taken to be 284.8 eV. Monochromatic AlK α radiation was used for all measurements; an analyser pass energy of 160 eV was used for survey scans while 40 eV was employed for detailed regional scans. The intensities of the Pd (3d) and Ga (2p) features were used to derive the PdĲ0)/PdĲII) and Ga/Pd surface ratios. The intensities of the Pd (3d) and In (3d) features were used to derive the PdĲ0)/PdĲII) and In/Pd surface ratios.
Transmission electron microscopy (TEM) was performed on a JEOL JEM-2100 operating at 200 kV. TEM images of metal nanoparticles and high angle annular dark-field (HAADF) scanning transmission electron microscopy (STEM) and energy dispersive X-ray spectroscopy (EDS analysis) of supported metal nanoparticles were carried out with a FEI Tecnai F20 ST TEM (operating voltage 200 kV) which was equipped with a field emission gun and an EDAX EDS X-ray spectrometer (Si (Li) detecting unit, super ultra-thin window, active area 30 mm 2 , resolution 135 eV at 5.9 keV). For TEM/ EDS analysis, a small droplet of the colloidal nanoparticle solution in chloroform or the catalyst powder, accordingly, were deposited on amorphous carbon-coated, 400 mesh Cu grids and air dried.
X-ray absorption spectroscopy (XAS) measurements at Pd K absorption edge were performed at the CAT-ACT beamline 30 (CAT experimental station, using Si (311) double crystal monochromator) of the KIT synchrotron (Karlsruhe, Germany) and the P64 beamline (using Si(111) channel-cut QEXAFS monochromator) of the PETRA III at DESY (Hamburg, Germany) in transmission mode using ionization chambers as detectors. Catalyst samples were measured ex situ as powders packed in 3 mm o.d. quartz capillaries (0.02 mm wall thickness). The spectra were normalized and the extended X-ray absorption fine structure spectra (EXAFS) background subtracted using the ATHENA program (IFFEFIT). 31 The k 1 -, k 2 -, and k 3 -weighted EXAFS functions were Fourier transformed in the k range of 2.5-12.5 Å −1 and multiplied by a Hanning window with sill size of 1 Å −1 . The structural model was based on a Pd metal core (ICSD collection code 52251) and a PdO shell (ICSD collection code 24692). For the Pd2Ga/TiO 2 sample, addition of a Ga shell from Pd2Ga (ICSD collection code 409939) model structure improved the fit. The structure refinement was performed using ARTEMIS (IFFEFIT) 31 using theoretical backscattering amplitudes and phases calculated by FEFF 6.0. 32 The theoretical data were then adjusted to the experimental spectra by a least square method in R-space between 1.0 and 3.2 Å −1 . First, the amplitude reduction factors (S 0 2 = 0.78 for CAT-ACT and 0.88 for P64) were calculated using the Pd foil reference spectrum and then the coordination numbers, interatomic distances, energy shift (δE 0 ) and mean square deviation of interatomic distances (σ 2 ) were refined. The absolute misfit between theory and experiment was expressed by ρ. 31 Metal leaching during the direct synthesis reaction was quantified using an Agilent 7900 ICP-MS equipped with an I-AS autosampler using a 5-point calibration using certified reference materials from Perkin Elmer and certified internal standard from Agilent. All calibrants were matrix matched.

Direct synthesis of H 2 O 2
Hydrogen peroxide synthesis activity was evaluated using a Parr Instruments stainless steel autoclave, equipped with PTFE liner, with a nominal volume of 100 ml and a maximum working pressure of 14 MPa. To test each catalyst for H 2 O 2 synthesis, the autoclave was charged with catalyst (0.01 g) and solvent (5.6 g MeOH and 2.9 g H 2 O). The charged autoclave was then purged three times with 5% H 2 /CO 2 (0.7

Results and discussion
We initially investigated the unsupported monometallic Pd, Ga, In and bimetallic Pd-Ga, Pd-In nanoparticles via XRD (Fig. 1). The XRD diffractograms revealed very broad reflections of low intensity, indicative of small nanoparticles. In particular, a broad, low intensity reflection centred at 33.  View Article Online a very narrow size distribution and spherical shape of the as prepared metal nanoparticles. Upon immobilization of Pd-M nanoparticles onto the acid-washed TiO 2 support, mean particle size of the asprepared alloyed particles, determined by transmission electron microscopy (TEM), remained generally well controlled, with mean particle size in the range of 1-4 nm (Table  S3 † and Fig. 3). However, upon immobilisation a relatively minor decrease in mean particle size is observed. Although, in the case of the Pd-only and PdGa/TiO 2 catalysts this is not considered to be statistically relevant; in the case of the PdIn/ TiO 2 series it is possible that this decrease is a result of restructuring of the nanoparticles upon immobilisation. In an attempt to elucidate the extent of restructuring/segregation, if any, EDX analysis of the supported PdIn metal nanoparticles was carried out (Fig. S2-S8 †). Unfortunately, the low signal intensities of the metals mean it is not possible to clearly and distinctly observe the extent of surface segregation. As such a relationship between nanoparticle restructuring and enhancement in the catalytic performance cannot be ruled out.
While ICP-OES (Tables S1 and S2 †) was carried out in order to confirm total metal loading and Pd : Ga and Pd : In ratios of nanoparticles and the supported catalysts. The structure and composition of the supported catalysts was further investigated by X-ray absorption spectroscopy (XAS). XANES analysis (Fig. 4) show mixed features attributed to the presence of both PdO (high intensity just above the absorption edge at ca. 24 367 eV, the so-called "white line") and metallic

View Article Online
Pd (e.g. a valley at 24 367 eV and a peak at approx. 24 390 eV). This has previously been observed by Centomo et al. 33 Table  S6 †). The modification of Pd via the addition of a range of secondary, non-precious metals has previously been demonstrated to enhance catalytic selectivity, with a reduction in the amount of contiguous Pd ensemble sites and an enhancement in the number of isolated Pd sites often reported as the cause for the improvement in catalytic selectivity towards H 2 O 2 . 35 In keeping with these studies, we now report the  For any catalyst operating in a three-phase system the possibility of leaching of the active phase is of great concern, with the low stability of sol-immobilized catalysts well known. 36 Indeed previous work by Dissanayake and Lunsford has demonstrated the high activity of homogeneous Pd species in the direct synthesis of H 2 O 2 . 37 To this end, post reaction solutions were analysed by ICP-AES (Table S7 †) with negligible levels of all metals observed.
To further investigate the effect of Ga and In addition to supported Pd catalysts analysis by X-ray photoelectron spectroscopy (XPS) was carried out (Table 2 and Fig. S10 †). From our analysis it is clear to see that Pd predominantly exits as a metallic species, which can in part be related to the high rates of H 2 O 2 degradation observed over these catalysts. With the catalytic selectivity of Pd-based catalysts known to be highly dependent on the oxidation state of Pd, with metallic Pd species typically more active towards both H 2 O 2 synthesis and its subsequent degradation than an analogous PdO catalyst [38][39][40] or those of mixed oxidation state. 41 Upon introduction of either Ga or In a general rise in PdO content is observed, which in turn can be related to the inhibition of H 2 O 2 degradation activity. This enhancement in PdO content is far more pronounced through.
In incorporation and may explain the lower activity of the Pd-In catalysts towards both H 2 O 2 synthesis and its subsequent degradation.

Conclusions
In conclusion, we have demonstrated an effective means of producing supported bimetallic Pd-Ga and Pd-In catalysts, with effective control for the particle size. The materials resulting from these preparations were found to offer enhanced selectivity towards H 2 O 2 compared to the analogous Pd catalyst. With the introduction of small quantities of Ga in particular shown to markedly inhibit H 2 O 2 degradation and enhance catalytic selectivity, through modification of Pd oxidation states.
We consider that these catalysts represent a promising basis for further exploration for the direct synthesis of H 2 O 2 .

Conflicts of interest
There are no conflicts to declare.