Gold–palladium colloids as catalysts for hydrogen peroxide synthesis, degradation and methane oxidation: effect of the PVP stabiliser

The reactivity of AuPd nanoparticle catalysts prepared by sol immobilisation is often explained by a structure activity relationship based solely on particle size or composition. In this contribution, we compare colloidal AuPd nanoparticles stabilised with polyvinylpyrrolidone (PVP) with the same AuPd nanoparticles supported on TiO2 for the direct synthesis of hydrogen peroxide and methane oxidation to methanol. We show that while the particles have similar rates of H2O2 synthesis, supporting the particles can affect the rates of H2O2 decomposition and hence the effectiveness of the catalyst for reactions which rely on H2O2 as an initiator or oxidant. We demonstrate that the absence of PVP results in high rates of H2O2 decomposition in methane oxidation experiments but this can be minimised by the addition of PVP to the reactor. These results also show that for AuPd alloys, both polymer stabiliser and support effects need to be taken into account when describing the activity of the nanoparticles and the active sites should in fact be thought of as a metal–support–polymer interface with many degrees of freedom.


Introduction
The preparation of supported precious metal catalysts by colloidal nanoparticle synthesis methods can result in highly active materials for a range of important chemical processes. 1,2 The advantage of solution phase synthesis of metal nanoparticles is the high degree of control over the particle size, shape and composition that can be achieved before immobilisation onto high surface area supports. 3 Typically, colloidal methods involve fast chemical reduction of metal precursors and steric stabilisation of the growing nanoparticles by polymer additives. 4 In many cases these catalysts are used without high temperature oxidative heat treatments or washing protocols specifically designed to remove the strongly bound polymer. [5][6][7] Therefore, in reality the reported performance of the catalyst material not only depends on the nature of the metal particle (size, shape, oxidation state) but also the nature of the composite particlepolymer surface.
Reaction rates, product selectivity and enantioselectivity can all be altered by the interaction of the polymer ligand or organic modifier with the metal surface. 8 For instance, the introduction of chiral modifiers such as cinchonidine derivatives or napthyl-ethylamine to Pt catalysts have been shown to impart increased enantioselectivity in activated ketone hydrogenation. [9][10][11] Polymer additives can also be used to tune the oxidation states of metal surfaces through binding of electron donating or withdrawing groups which in turn effects reactant binding strengths. 12 Steric effects can also be observed in many cases with changes in reaction selectivity observed in the presence of polymer additives. 13 Despite these potential advantageous degrees of control available by polymer addition, detrimental effects such as site blocking by strong adsorption can reduce activityespecially in gas phase reactions. [14][15][16] This requires careful removal of polymer additives prior to reaction without significantly changing the nanostructure of the catalyst material which has been carefully controlled in the colloidal synthesis of the nanoparticles.
Polyvinylpyrrolidone (PVP) is a common polymer additive used in the preparation of precious metal catalysts for reactions such as glycerol oxidation using Au and Pd based catalysts supported on TiO 2 . [17][18][19][20] We recently demonstrated that while effective catalysts could be prepared in the absence of additives, the presence of PVP and polyvinylalcohol (PVA) has significant effects on product selectivity which is often not considered in structure activity relationships. 21,22 Similar catalyst structures based on Au-Pd particles have been shown to be highly active for reactions such as the direct synthesis of hydrogen peroxide (H 2 O 2 ) from molecular hydrogen and oxygen which would represent an attractive alternative to the current indirect anthraquinone process for on-site H 2 O 2 production. [23][24][25][26][27] Catalysts with varying alloy compositions and nanostructure have been extensively studied for this reaction and have been shown to be highly active for H 2 O 2 synthesis. 27 Titania supported Au-Pd catalysts prepared by colloidal methods with small particle sizes tend to have significant activity for H 2 O 2 degradation by over hydrogenation and decomposition. 28 To date, the challenge of minimising the subsequent degradation of H 2 O 2 (via hydrogenation and decomposition pathways) has only been achieved with a few specific catalyst systems in the absence of acid and halide additives. [29][30][31] Unsupported colloidal nanoparticles can be active for a number of reactions including glucose oxidation, 32 aromatic alcohol oxidation 33 and CO 2 reduction. 34 36 Recently we demonstrated that using colloidal AuPd nanoparticles stabilised with PVP, it was possible to achieve selective oxidation of methane to methanol at 25-50°C with O 2 incorporation using H 2 O 2 as an initiator. 37 A significant result in this study showed that the background H 2 O 2 decomposition rate was minimal for unsupported colloidal AuPd-PVP particles at 50°C, allowing the radical reaction process to propagate rather than terminate; however the H 2 O 2 decomposition rate increased significantly when then the same particles were supported on TiO 2 making this an ineffective catalyst. 37 This result clearly demonstrated that colloidal particles stabilised by polymers can have significantly different behaviour in reactions involving H 2 O 2 than analogous 'bare' particles supported on TiO 2 and while colloidal catalyst systems could have limitations in terms of long term stability it is possible to use them to propose structure-activity relations. In this study, we aim to elucidate if the PVP stabilisers present on AuPd particles effect the H 2 O 2 synthesis and methane oxidation reaction networks and investigate if the reactivity in these systems should be considered as not only a result of interactions with the surface atoms on the metal nanoparticle, but rather with an entity more akin to a metal core-polymer shell composite.

Results and discussion
Recent studies by Giorgianni et al. reported that the addition of PVA to Pd catalysts results in enhanced H 2 O 2 yields due to hindered H 2 O 2 back diffusion to the metal surface via hydrophobic interactions in addition to greater observed catalyst stability when PVA is present. 38 In this current study on H 2 O 2 synthesis and methane oxidation, we focus on PVP as an additive because in methane oxidation reactions traces of MeOH have been previously observed as a result of PVA contamination or oxidative degradation under our reaction conditions. A series Au-Pd colloidal nanoparticle solutions were synthesised according to previously described procedures. 37 The colloidal solutions of Au-Pd nanoparticles (Au : Pd = 1 : 1 by moles) were synthesised via NaBH 4 reduction of metal salts and stabilised by PVP. A portion of the AuPd-PVP colloidal nanoparticles were immobilised onto TiO 2 (1 wt% metal loading) and dried at 110°C to provide comparative supported samples. Both the supported and unsupported AuPd-PVP materials were then tested for the direct synthesis of H 2 O 2 and methane oxidation, using experimental conditions which have both been previously described in the literature. 28,37 The colloidal metal nanoparticles and the supported colloidal nanoparticles have been previously characterised by X-ray photoelectron spectroscopy. 37 XPS analysis of the colloidal AuPd-PVP nanoparticle solution in the Au(4f) and Pd(3d)/Au(4d) regions, after drop casting onto a solvent cleaned silicon wafer, showed that both elements were present in predominantly the metallic state (Au(4f) ∼ 83-84 eV, Pd(3d) ∼ 334.5-335.5 eV, the range in energies varying depending on particle size and support interactions) as could be expected from the presence of the strong reducing agent at the high molar ratio used in the preparation. For the unsupported colloid, a Au signal at 85.5 eV and a Pd signal (ca. 388 eV) were also detected and attributed to residual or leached metal chlorides. XPS analysis of the immobilised nanoparticles exhibit both Au and Pd binding energies ca. 1 eV lower compared to the unsupported colloid. Such low Pd binding energies may be attributed to a particle-support interaction making the zero-valent supported Pd more electron rich. The corresponding Au(4f 7/2 ) signal exhibits a binding energy of 83.1 eV and lower than the characteristic 84 eV for bulk Au, although such binding energies have been widely reported, and are typically attributed to small, low coordination atoms, charge transfer from PVP to Au, and charge transfer between Pd and Au, increasing the s-state occupancy of Au indicating alloy formation. 39,40 TEM analysis ( Fig. 1a and b) showed that the mean size of the bimetallic particles in the colloidal solution was 3.0 ± 2.0 nm, which increased slightly to 4.1 ± 1.3 nm when the particles were immobilised onto the support. This is in good agreement with our previous studies on these materials where the apparent increase in size occurs as a result of the AuPd nanoparticle flattening slightly and faceting to form an interface with the TiO 2 support ( Fig. 1c and d). 28
This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

View Article Online
Previous studies have demonstrated that colloidal Pd and AuPd-PVP nanoparticles, in the presence of strong acid and halide additives, can produce H 2 O 2 via the direct reaction of H 2 and O 2 . 41 These un-supported particles were stabilised by PVP and were shown to exhibit appreciable rates for direct H 2 O 2 synthesis suggesting that the presence of a support is not essential for high activity in this reaction. We carried out comparative tests between our colloidal AuPd-PVP materials in the unsupported and TiO 2 supported state ensuring that each reaction had the equivalent moles of metal (0.66 μmol). The results, shown in Table 1, entries 1-3, demonstrate that over a 30 min reaction period the unsupported colloidal Au-Pd-PVP nanoparticles were capable of producing 0.10 wt% H 2 O 2 compared to 0.07 wt% H 2 O 2 when using TiO 2 supported counterparts; their corresponding apparent turnover frequencies (TOFs) based on total moles of metal were 8.0 × 10 2 and 5.2 × 10 2 mol H 2 O 2 mol metal −1 h 1 respectively. The TiO 2 support alone showed no background activity towards H 2 O 2 synthesis or H 2 O 2 degradation (via over-hydrogenation or decomposition). Next, the amount of AuPd-PVP colloidal nanoparticle solution used was varied to identify the kinetic regime in which the reaction was operating. Fig. 2 shows that increasing the amount of Au-Pd colloidal nanoparticle solution added to the reaction results   42 Over a 30 min reaction time, significant H 2 O 2 degradation is likely when using Pd containing nanoparticles in the 2-5 nm size range, with halide and acid stabilisers typically used to minimize degradation pathways, suggesting that the reported TOFs after 30 min of reaction are under-estimates of the actual TOFs, because some fraction of the H 2 O 2 produced will have been destroyed by subsequent over-hydrogenation or decomposition reactions. 43 Fig. 3

View Article Online
Due to the measured difference in particle size distribution of the two catalytic systems, the total number of surface atoms available in each reaction was estimated using the particle size distributions obtained by TEM and applying the same model for atom packing to each system. This approximation assumes activity is related to the total number of exposed surface atoms and is not correlated with specific surface sites such as edges or corner atoms. In addition it assumes that the polymer coverage of the metal surface is consistent between both samples. This analysis gave 1.2 × 10 17 surface atoms for the colloidal system assuming spherical particles versus 5.7 × 10 16 for the TiO 2 supported catalyst assuming hemispherical supported particles meaning that the colloidal samples have ∼2.1 times the exposed metal surface. Based on the amount of H 2 O 2 produced after 2 min of reaction, where the contribution from subsequent reactions (hydrogenation and decomposition) is minimal, we find that the moles of H 2 O 2 produced at each exposed surface site is comparable between the colloidal metal nanoparticle solution (1.7 × 10 −21 mol H 2 O 2 metal site −1 h −1 ) and the supported AuPd nanoparticles (1.5 × 10 −21 mol H 2 O 2 per metal site per h). Considering the assumptions made during this comparison this suggests that H 2 O 2 formation occurs on the surface of the Au-Pd nanoparticles at similar rates in both catalytic systems when minimal H 2 O 2 degradation pathways are operating. Due to the large difference in catalyst mass between colloidal and supported nanoparticle catalysts rates normalised to both total moles of metal and moles of metal surface for 30 min reactions are reported in Table 1 assuming no significant change in the catalyst structure takes place.
Experiments were also carried out to investigate the effect that supporting the colloidal AuPd-PVP nanoparticles on TiO 2 had on the degradation of a 4 wt% H 2 O 2 solution, which included contributions from both H 2 O 2 hydrogenation and decomposition pathways. Comparing the unsupported colloidal AuPd-PVP solution to the corresponding TiO 2 supported catalyst, the total degradation after 30 min (Table 1, entries 1-2) was similar for both situations. However, when normalising to the difference in surface sites between the colloidal and supported nanoparticles, it is clear to see that the supported particles have significant higher rates per metal surface site by a factor of ∼2.5. By comparing the rates of H 2 O 2 decomposition under a 25% O 2 /CO 2 atmosphere to remove over-hydrogenation from the possible reaction pathways, it was observed that the colloidal AuPd-PVP solution decomposed only 1% of the H 2 O 2 present compared to the supported particles which decomposed 7% of the initial H 2 O 2 suggesting a significant difference in how the two samples interact with H 2 O 2 .
The decomposition of H 2 O 2 occurs much faster when the Au-Pd particles are supported on TiO 2 , despite the supported particles having larger mean particle size and therefore less exposed surface available (by roughly a factor of 2) meaning an increase in observed decomposition rate of about 14 times for the supported particles as compared to the unsupported colloidal particles. Assuming that the H 2 O 2 hydrogenation and decomposition pathways are independent of each other, this suggests that either it is the act of supporting the particles which generates a metal-metal oxide support interface or the removal of PVP on washing the solid which changes the predominant H 2 O 2 degradation pathway. Our previous studies have shown that supporting Au colloids on crystalline metal oxides such as TiO 2 results in a more highly faceted particle structure than when they are supported on amorphous carbon structures. 14,15,44 We prepared an analogous sample by supporting the same colloidal AuPd nanoparticle solution on an activated carbon. This sample showed higher H 2 O 2 production, but also no H 2 O 2 decomposition activity, supporting the notion that immobilisation of the AuPd colloid onto TiO 2 P25 surface results in increased H 2 O 2 decomposition rates possibly through the formation of highly faceted surfaces as the particle restructures to minimise the energy needed to interact with the underlying support lattice and form an extended intimate support/metal interface. This highlights the possibility that the underlying support material can possibly indirectly effect the decomposition pathways of H 2 O 2 on the surfaces of the alloy nanoparticle attached to it.
Numerous observations that supported catalysts prepared from PVP-stabilised colloids show low activity in gas phase reactions, such as CO oxidation, due to the presence of residual polymer suggest that the polymer is still present to some extent after the preparation of the supported catalyst material. 15,16 Further studies by Han et al. report the stability of the PVP polymer under H 2 O 2 synthesis conditions similar to those used within this work supporting the hypothesis that PVP is stable to the reaction conditions used in this study. 6 To further investigate if the presence of PVP contributes to these effects colloidal Au-Pd nanoparticles were prepared using a variety of PVP : metal ratios between 20 and 0.005 in order to study the influence of PVP concentration and particle size on the reaction, (Table 1 entries 5-7, Fig. 4). The mean particle size of the colloidal particles ranged from 6 ± 1.9 nm for a PVP/metal ratio of 0.05 to 1.9 ± 1.3 nm for a PVP/metal ratio of 20. This decrease in mean size correlates well with the increasing amount of polymer allowing the formation of a higher number of smaller nuclei during the preparation. Fig. 4 shows that as the average particle size of the AuPd nanoparticles decreases, the net amount of H 2 O 2 produced after 30 min of reaction increases from 0.08 wt% for 6 nm particles to 0.12 wt% for 1.9 nm particles. In all cases, the colloidal AuPd-PVP particles showed minimal H 2 O 2 decomposition (Table 1, entry 5-7), even when the solution contained smaller particles (i.e., a greater number of exposed metal surface atoms) producing higher amounts of H 2 O 2 over the 30 min reaction. Fig. 4b reports the normalised rate per surface atom as approximated by the TEM derived particle size distributions for a range of catalysts prepared with various PVP to metal ratios for H 2 O 2 synthesis, degradation and decomposition. No significant change in apparent TOF per surface atom was observed over the mean Astruc and co-workers have demonstrated that it is possible to prepare colloidal Au catalysts in the absence of polymer stabilisers relying solely on electrostatic stabilisation of the metal particles by residual salts in the colloidal solution. 45 We recently showed that it is possible to prepare supported AuPd catalysts in a similar manner with a slightly broader particle size distribution compared to conventional immobilisation of polymer stabilised colloidal solutions. 22 This colloidal nanoparticle solution prepared without polymer addition (denoted SF-stabiliser free) was also tested for direct H 2 O 2 synthesis (Table 1, entry 8). Interestingly they showed activity towards H 2 O 2 production and also increased rates of H 2 O 2 decomposition suggesting that the greater accessibility to the surface in the absence of PVP results in Fig. 4 a) 37,40 It was also demonstrated that the TiO 2 supported AuPd colloidal nanoparticles showed low activity in this particular reaction due to high levels of H 2 O 2 decomposition.
To extend our previous study and elucidate the role of PVP in this reaction we conducted further methane oxidation experiments with H 2 O 2 only as an oxidant using the 1 wt% AuPd/TiO 2 catalyst with extra PVP added to the reaction to identify if this unwanted H 2 O 2 decomposition activity, initiated by supporting the AuPd-PVP particles, could be suppressed. Table 2 shows a summary of the catalytic testing results for a colloidal AuPd catalyst prepared with PVP, the TiO 2 supported AuPd-PVP nanoparticles, and the same supported nanoparticles with the addition of extra PVP (giving a metal to PVP ratio 1.2 by weight in the reaction). The results clearly show the superior efficiency of the unsupported AuPd-PVP colloidal system with respect to H 2 O 2 efficiency in producing oxygenates. On adding extra PVP to the supported catalyst an improvement was noted with respect to H 2 O 2 efficiency and oxygenate production from a H 2 O 2 : product ratio from 570 to 376, however, the activity of unsupported colloidal system still remained over an order of magnitude better. This demonstrated that the support induced H 2 O 2 decomposition could not be fully negated by the addition of further PVP stabiliser.
Further experiments were carried out to investigate if the presence of polymer stabiliser was crucial in achieving the high efficiency of methane oxidation with H 2 O 2 and O 2 using colloidal AuPd-PVP nanoparticles. Fig. 6 shows a stark comparison of the catalytic performance of the colloidal metal nanoparticle system synthesised with and without PVP polymer. Without the PVP polymer being present, H 2 O 2 consumption during the 30 min reaction period at 50°C was over 95%, producing minimal oxygenated products. This compares to only 32% H 2 O 2 consumption when the colloid incorporated the PVP ligand. The lower H 2 O 2 consumption in this latter case resulted in significantly higher oxygenated product formation showing that the polymer is crucial in achieving high reactivity, presumably-by controlling the rate of H 2 O 2 decomposition. We then took the stabiliser free colloidal solution prepared in the absence of PVP and added PVP to the methane oxidation reaction. The catalytic performance was significantly improved and indeed, the activity approached that of the colloidal solution prepared with PVP. The H 2 O 2 consumption was 48% with significant oxygenate production demonstrating that it is in fact the polymer additive significantly effects the rate of H 2 O 2 decomposition on the metal surface and therefore the overall efficiency of methane oxidation. The PVP ligand shell presumably controls the rate of H 2 O 2 diffusion to the metal surface and therefore the production of radicals that can interact with the solubilised methane, whose concentration may also be enhanced by the hydrophobic alkyl chains in the organic layer around the metal surface.
In conclusion, we have investigated the role of the PVP ligand shell and TiO 2 support in controlling the reactions of H 2 O 2 on AuPd-PVP colloidal particles. These particles display significant activity towards H 2 O 2 direct synthesis, hydrogenation and decomposition as well as methane oxidation using H 2 O 2 and O 2 . We have shown that supporting the PVP stabilised AuPd nanoparticles on TiO 2 increases

Catalyst preparation
Bimetallic Au-Pd nanoparticles were prepared by standard colloidal methods. An aqueous solution of HAuCl 4 precursor (Strem Chemicals) and acidic solution of PdCl 2 (Sigma Aldrich) precursor (in 0.58 M HCl) were dissolved in 800 mL of de-ionized water (Au : Pd = 1 : 1 by moles) to give a total metal concentration of 0.16 mmol L −1 . Polyvinylpyrrolidone (PVP, average molecular weight 1 300 000, Sigma Aldrich) was added as a stabilizer to give the required metal-to-PVP weight ratio (typically 1 : 1.2 (wt/wt)). After 2-3 min of stirring, freshly prepared 0.1 M sodium borohydride (NaBH 4 , Sigma Aldrich) solution was added such that the molar ratio of NaBH 4 -tometal was 5 : 1 (mol/mol). This produced a dark brown colloid which was then left stirring for 30 minutes to ensure that all the metal precursors were reduced to the metallic form. The colloid was concentrated using a roto-evaporator to give a nominal metal loading of 0.66 mmol L −1 . The colloid was stored in glass media bottles prior to use. For supported catalysts, the sol prepared as described above was immobilized onto a TiO 2 (P25, Degussa, 1.98 g) or activated carbon (Darco G60) (1.98 g) in the following manner. A sufficient amount of support material was added to ensure a 1 wt% metal loading and the solution was acidified to pH 1 using sulphuric acid to enhance and achieve more homogeneous deposition of nanoparticles. The supernatant solution became clear over a 1 h period after support addition, indicating the deposition process was complete. The sol-immobilized catalyst was then filtered, washed thoroughly with distilled water and then left to dry in an oven at 110°C for 16 h.

Direct synthesis of H 2 O 2
Hydrogen peroxide synthesis was evaluated using a Parr Instruments stainless steel autoclave with a nominal volume of 100 mL and a maximum working pressure of 14 MPa according to our previous optimisation studies. 46 To test each catalyst for H 2 O 2 synthesis, the autoclave was charged with catalyst (in each case 0.66 μmol), 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 (7 bar) before filling with 5% H 2 /CO 2 (29 bar), followed by the addition of 25% O 2 /CO 2 (11 bar). The temperature was then decreased to 2°C followed by stirring (1200 rpm) of the reaction mixture for 0.5 h. H 2 O 2 productivity was determined by titrating aliquots of the final   (29 bar) represents the sum of both hydrogenation and decomposition pathways, while the use of 25% O 2 /CO 2 (29 bar) allows for catalytic activity towards H 2 O 2 decomposition alone to be determined. All reaction conditions were as described above, apart from the gas atmosphere for the H 2 O 2 decomposition studies.

Methane oxidation reaction conditions
Methane oxidation was carried out in a 50 mL glass-lined stainless steel Parr autoclave reactor. The reactor was charged with either colloidal or supported catalyst (0.66 μmol metal per reaction) and H 2 O 2 (amount defined in figure captions, Sigma Aldrich, 50 wt% in water). The charged autoclave was sealed and purged three times with methane (10 bar 99.999%, Air Products). It was then pressurized with methane (30 bar) and in some cases oxygen (5 bar, BOC). The mixture was stirred at 1500 rpm and heated to 50°C at a ramp rate of 2.25°C min −1 and maintained at the reaction temperature for 30 min. At the end of the reaction, the autoclave was cooled in ice to a temperature below 10°C in order to minimize the loss of volatile products. The reaction gas was removed for analysis in a gas sampling bag.

Product analysis for the methane oxidation reaction
Liquid phase product analysis was carried out using 1 H-NMR on a Bruker 500 MHz instrument equipped with a solvent suppression system. Tetramethylsilane (TMS) in CDCl 3 was used as an internal standard. The H 2 O 2 concentration was determined using a titanium oxalate spectrophotometric method (Agilent, Cary 60). In this procedure, 0.05 to 1.0 mL of reaction sample was acidified using dilute H 2 SO 4 before adding potassium titanium oxalate solution (0.5 wt% in water, Sigma Aldrich) to form the yellow pertitanic acid complex with a characteristic absorption peak at 390 nm. Gaseous products were quantified using a Varian 450-GC fitted with a CP-Sil 5CB capillary column (50 m length, 0.32 mm diameter, carrier gas = He), a methaniser unit and both FID and TCD detectors.

Catalyst characterization
TEM. Materials for TEM analysis were prepared by dispersing the colloidal or supported catalyst onto a continuous carbon film supported on a 300-mesh copper TEM grid and allowing the solvent to evaporate. Specimens were examined using the bright-field imaging mode in a JEOL 2000FX transmission electron microscope operating at 200 kV equipped with an Oxford Instruments X-ray energy dispersive (XEDS) spectrometer system.
Explanation of determination of number of surface atoms is reported in the ESI † material. The same approach was used for each catalyst material for comparative purposes assuming spherical particles for colloidal nanoparticles and hemispherical particles for supported nanoparticles.

Conflicts of interest
There are no conflicts to declare.