Single-step co-deposition of nanostructured tungsten oxide supported gold nanoparticles using a gold–phosphine cluster complex as the gold precursor

The use of a molecular gold organometallic cluster in chemical vapour deposition is reported, and it is utilized, together with a tungsten oxide precursor, for the single-step co-deposition of (nanostructured) tungsten oxide supported gold nanoparticles (NPs). The deposited gold-NP and tungsten oxide supported gold-NP are highly active catalysts for benzyl alcohol oxidation; both show higher activity than SiO2 supported gold-NP synthesized via a solution-phase method, and tungsten oxide supported gold-NP show excellent selectivity for conversion to benzaldehyde.


Introduction
We report a vapour phase method for the one-step preparation of supported gold nanoparticles (NPs), achieved by in situ decomposition of a molecular gold cluster and an organometallic tungsten precursor, and on the catalytic properties of the co-deposited material. Vapour synthesis of gold NPs, e.g. via chemical vapour deposition (CVD), offers advantages over wet-chemical techniques; it circumvents several steps associated with solution methods such as washing, drying, calcination and reduction, and can avoid changes in metal dispersion that can occur during high temperature calcination or reduction steps [1]. The use of well-defined molecular precursors in CVD could potentially provide control of the particle size, a critical parameter for the reactivity of the NP [2], by controlling the number of atoms involved in the particle. However, large molecular precursors are expected to have limited volatility and are therefore unsuitable for use in traditional CVD methods; the use of aerosol assisted (AA)CVD, in which the precursor is delivered as an aerosol of a precursor solution rather than as a vapour, allows a wider range of precursors to be exploited as solubility is the key precursor requirement as opposed to volatility [3]. Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
A variety of synthetic strategies to supported metal NP are described in the literature [4][5][6][7], however the use of single-step routes in which both support and catalytic metal are synthesized together are infrequently reported despite the advantage of fewer processing steps [8]. We have recently reported the successful use of AACVD for the one-step synthesis of metal NPs supported on tungsten oxide, using inorganic metal precursors (H x MCl y , where M = Au, x = 1, y = 4 or M = Pt, x = 2, y = 6) [9,10]. One concern in using these inorganic precursors for synthesis of catalytic metal particles is potential for incorporation of chlorine, which can poison the catalyst. Non-halide containing mononuclear gold-phosphine complexes, for instance [R-Au-PR 3 ] [11,12] and [Me-Au-P(OMe) 2 R′] [13], are known CVD precursors for the deposition of gold films, and the molecular gold cluster [Au 9 (PPh 3 ) 8 ](NO 3 ) 3 has been used as a precursor to supported metallic Au particles via solution-phase impregnation [14], however the use of molecular gold clusters in CVD has not previously been reported.
Coupled differential scanning calorimetry-thermal gravimetric analysis (DSC-TGA) was performed on the gold precursor using a Netzsch STA 449C instrument. The TGA was carried out at atmospheric pressure, under a flow of helium gas (50 cm 3 min −1 ), and the rate of heating was 2°C min −1 . The morphology of the samples was examined using scanning electron microscopy (SEM-Jeol 6301F, 5 keV), the structure using x-ray diffraction (XRD-Bruker, AXD D8 Discover, using Cu Kα radiation, operated at 40 KV and 40 mA) and the chemical composition using wavelength dispersive x-ray spectroscopy (WDX-Philips, XL30ESEM). Transmission electron microscopy (TEM; JEOL JEM-100CX II, 100 kV) and high resolution TEM (HRTEM-Jeol 2100, 200 keV) measurements were performed by removing the film from the substrate by sonication of the substrates in methanol. X-ray photoelectron spectroscopy (XPS) was carried out using a Thermo K-α spectrometer in constant analyser energy mode and monochromated Al Kα radiation. Survey scans were collected over the 0-1400 eV binding energy range with 1 eV resolution and a pass energy of 200 eV. Higher resolution scans (0.1 eV) encompassing the principal peaks of W, O, Au, C, Si and P were also collected at a pass energy of 50 eV. Data was processed using CasaXPS.

Results and discussion
Deposition of gold particles on glass substrates was achieved at 350°C using a solution of [Au 9 (PPh 2 (p-tol)) 8 ] (NO 3 ) 3 in methanol and the partially adherent films were transparent and red/purple in colour. UV-vis spectra (figure 2) displayed a broad absorbance between 500 and 650 nm, attributed to the surface plasmon resonance of the gold particles, with the peak centre at ∼566 nm.
Estimation of gold particle size from the peak absorbance of the plasmon observed using UV-vis spectroscopy (figure 2), using multipole scattering theory as described by Haiss et al [18] and assuming the gold particles were spherical and uncoated and the refractive index of the medium in contact with the particles was dominated by the glass substrate (1.517 at 546 nm), gave a value of 70 nm.
WDX of the gold coatings showed considerable carbon (∼19 at%) but no phosphorus contamination, suggesting the phosphine ligands were eliminated and the residual carbon was due to aerosol solvent pyrolysis. XRD revealed only peaks assigned to cubic gold (Fm-3m space group, a = b = c = 4.070 (1) Å; ICDD card no. 00-04-0784 a = b = c = 4.0786 Å). SEM analysis showed the film to be composed of roughly spherical particles of diameter up to ∼50 nm diameter, which was confirmed by TEM ( figure 3) with a wide dispersion of particle sizes observed between 5 nm and 58 nm (average = 16 ± 1 nm; n = 100). HRTEM images (figure 3 top and inset) of the gold particles exhibited highly ordered planar spacing of 2.39 ± 0.02 Å, consistent with the internal lattice spacing of the (111) plane of gold (d = 2.355 Å).
The size of these particles demonstrate that coalescence of the initial gold cluster cores (∼0.6 nm) [19] has occurred, although it is not possible to discern whether this is as a result of gas-phase reaction or aggregation on the substrate surface. Previous use of [Au 9 (PPh 3 ) 8 ](NO 3 ) 3 in solution-phase preparation of supported gold particles produced NP of ∼16 nm size [14], consistent with the average gold particle size found by TEM (16 ± 1 nm) for these AACVD deposited NP.
WDX indicated a sub-stoichiometric WO 2.69 composition, consistent with the blue colour of the films, with gold detected at ∼0.6 at% (Au/W atom ratio: ∼1/11 in precursors, ∼1/40 in film), phosphorus at <0.1 at% (P/W atom ratio: ∼1/ 13 in precursors, <1/500 in film), and carbon of ∼4 at% (C/W atom ratio: ∼40/1 in precursors, ∼1/7 in film). From these results we note the relatively poor incorporation of gold (∼25% efficiency), albeit higher than the 5-10% observed using inorganic precursors [10], and that the gold-phosphorus bonds are cleaved (P/Au ratio ∼1/1 in precursor, ∼1/17 in film). The dramatic reduction in carbon between precursor and film also indicates the precursors decompose as desired to eliminate unwanted contaminant elements under these conditions. XPS spectra revealed three W 4f 7/2 environments observed at 34.9 eV (oxidation state +6, 66%), 35.8 eV (oxidation state +5, 22%) and 36.6 eV (variation in vacuumlevel attributed to surface defects, 12%) indicating a WO 2.88 stoichiometry [20]. A single Au 4f doublet was observed with Au 4f 7 at 84.2 eV indicating the presence of metallic gold only and an Au/W ratio of ∼1/50, consistent with that found by WDX. No phosphorus was detectable indicating any contamination was below the detection limit of the instrument.
SEM showed the tungsten oxide had a nanorod morphology similar to that observed previously when depositing tungsten oxide alone under identical conditions [9], although some larger agglomerated structures were also observed. Films deposited both without and with (figure 5) the gold precursor present appeared similar in SEM images and no obvious gold particles were observed for films co-deposited with [Au 9 (PPh 2 (p-tol)) 8 ](NO 3 ) 3 .
In contrast TEM images clearly showed the presence of small (∼4-8 nm diameter) particles in co-deposited samples with the particles well distributed along the needle length ( figure 6(b)). These particles are absent in the sample deposited without gold precursor ( figure 6(a)). HRTEM images (figure 6(c)) exhibit highly ordered crystalline needle structures with marked planar spacing of 3.80 ± 0.04 Å, consistent with tungsten oxide ((001 or (010) planes) [10], with the decorating particles having planar spacing of 2.40 ± 0.03 Å, consistent with the (111) lattice spacing of gold. The gold particle sizes were on average slightly smaller and possessed a marginally narrower size distribution (4-11 nm diameter) than we observed using HAuCl 4 as the gold precursor [10].  Table 1 shows the catalytic properties of the AACVD deposited samples, an SiO 2 blank and an Au/SiO 2 catalyst synthesized via solution-phase calcination/impregnation using the same gold precursor, which includes turnover numbers (TONs) for conversion of benzyl alcohol for either the total amount of catalyst (including support) or for the amount of gold only.
The average mass of the co-deposited films was ∼2 mg, of which ∼1.8 wt% (as determined by WDX) was Au, i.e.
∼0.04 mg gold. The films deposited from the gold precursor alone had a mass of approximately 0.1 mg.
Gold-NP supported on SiO 2 (figure 7, 200 mg-equivalent to 8 mg Au), used as a standard for solution-phase synthesized material, gave 53% conversion of benzyl alcohol after 6 h with TON(gold) = 580. Tungsten oxide supported gold-NP (2 mg-equivalent to 0.04 mg Au) gave 23% conversion, however calculation of TON is complicated in this case as the tungsten oxide 'support' itself was also found to be active for benzyl alcohol oxidation (15% conversion, TON (catalyst) = 650). Tungsten oxides are versatile catalytic materials [21] and the oxidation of benzyl alcohol to benzaldehyde over tungsten oxide has been observed previously [22] with the oxidizing power of tungsten oxide attributed to the presence of oxygen defects and low valent tungsten centres [23], hence fully oxidized WO 3 was found to be rather inactive. Therefore the activity found for the tungsten oxide support is consistent with the sub-stoichiometric composition as determined by WDX and XPS. It is worth noting that the activity of these tungsten oxide samples (TON (catalyst) = 650) is comparable to that of the gold-NP supported on SiO 2 (TON(gold) = 580). Estimating the TON due to gold only for gold-NP supported on tungsten oxide, by subtraction of the conversion due to the support, gives TON (gold) = 17 000, whilst gold-NP deposited alone via AACVD were equally active with TON(gold) = 23 500, i.e. the CVD deposited gold-NP (either supported or 'bare') were an order of magnitude more active than gold-NP deposited from the same precursor via a solution phase method. It has been found previously that higher catalytic performance is observed in CVD deposited materials than those produced using solutionphase methods [24], although it should also be noted that the particle sizes (up to 50 nm) for the solution phase gold-NP were considerably larger than those for CVD deposited gold-NP (16 nm) or tungsten oxide supported gold-NP (∼6 nm).
The selective catalytic oxidation of hydrocarbons to their respective aldehydes is a major challenge for industry and reactions are frequently performed at very low conversion rates in order to avoid the formation of carboxylic acids [25]. Oxidation of benzyl alcohol to benzaldehyde is a useful test reaction as it is easily over-oxidized to benzoic acid and therefore provides a direct measure of the selectivity of the catalyst. Whilst both the SiO 2 supported and AACVD deposited 'bare' gold-NP showed good selectivity to benzaldehyde (97% and 94% respectively) the tungsten oxide supported particles provided 100% selectivity. A support can act as a promoter to enhance the activity and selectivity of a reaction [26], for instance Pt-NPs supported on tungsten oxide exhibit excellent CO tolerance [27,28] due to the presence of OH ads on the tungsten oxide surface facilitating the oxidation of passivating CO ads [29], hence the improvement in selectivity observed may be attributable to interaction between the tungsten oxide support and the gold NPs although further work, including operando studies, is ongoing to prove this.

Conclusions
In summary the AACVD co-deposition of gold-NP and gold-NP supported on nanostructured tungsten oxide, utilizing a molecular gold cluster precursor, produces an extremely active catalyst compared to a traditionally prepared catalyst of gold NPs on microstructured silica using the same gold precursor. In addition increased selectivity in oxidation of benzyl alcohol to benzaldehyde is observed for AACVD co-deposited gold NP supported on tungsten oxide, attributed to interaction between the support and the metal NP. AACVD is readily adaptable to other nanostructured metal oxides and metal NPs and therefore this technique can be applied to a wide variety of not only catalytic materials, but also materials for use in other fields, such as plasmonic nanostructures and solar cells.    Table 1. Catalytic performance of tungsten oxide, gold and tungsten oxide supported gold films compared to silica supported gold particles produced using the same gold precursor via solution-phase synthesis.