Combined in Situ XAFS/DRIFTS Studies of the Evolution of Nanoparticle Structures from Molecular Precursors

: The rational design of catalyst materials is of great industrial signi ﬁ cance, yet there is a fundamental lack of knowledge in some of the most well-established processes, e.g. formation of supported nanoparticle structures through impregnation. Here, the choice of precursor has a signi ﬁ cant in ﬂ uence on the resulting catalytic properties of the end material, yet the chemistry that governs the transformation from de ﬁ ned molecular systems to dispersed nanoparticles is often overlooked. A spectroscopic method for advanced in situ characterization is employed to capture the formation of PdO nanoparticles supported on γ -Al 2 O 3 from two alternative molecular precursors - Pd(NO 3 ) 2 and Pd(NH 3 ) 4 (OH) 2 . Time-resolved di ﬀ use re ﬂ ectance infrared Fourier transform spectroscopy is able to identify the temperature assisted pathway for ligand decomposition, showing that NH 3 ligands are oxidized to N 2 O and NO − species, whereas NO 3 − ligands assist in joining Pd centers via bidentate bridging coordination. Combining with simultaneous X-ray absorption ﬁ ne structure spectroscopy, the resulting nucleation and growth mechanisms of the precious metal oxide nanoparticles are resolved. The bridging ability of palladium nitrate aids formation and growth of larger PdO nanoparticles at lower onset temperature (<250 ° C). Conversely, impregnation from [Pd(NH 3 ) 4 ] 2+ results in well-isolated Pd centers anchored to the support which requires a higher temperature (>360 ° C) for migration to form observable Pd − Pd distances of PdO nanoparticles. These smaller nanoparticles have improved dispersion and an increased number of step and edge sites compared to those formed from the conventional Pd(NO 3 ) 2 salt, favoring a lower light o ﬀ temperature for complete methane oxidation. kV and 15 mA and using a 1D Si strip detector. Sherrer calculations of the average crystallite sizes used the full width half-maximum of the di ﬀ raction peaks located at the 2 θ angle 55 ° , assuming the Sherrer constant to be 0.94. Catalytic light o ﬀ curves for methane oxidation were obtained using JM automotive test rig, using a synthetic natural gas mixture (0.5% CH 4 , 10% O 2 , 0.01% C 2 H 6 , 0.01% C 3 H 8 , 0.01% NO, 10% CO 2 , and 10% H 2 O), a gas hourly space velocity of 90 000 h − 1 , and a temperature ramp rate of 15 ° C min − 1 . E ﬄ uent gases were analyzed by a MKS 2000 multigas FTIR analyzer.


■ INTRODUCTION
Supported metal nanoparticles are a cornerstone of heterogeneous catalysis, and by extension, the chemical industry. Optimizing the preparation method for smaller nanoparticle size and improved dispersion is a common theme in heterogeneous catalyst design: to improve specific surface area, lower precious metal content, and increase metal−support interfacial regions while maintaining thermostability toward sintering. There are various methods for the preparation of supported metal catalysts such as controlled colloidal routes, 1,2 deposition precipitation, 3 grafting techniques, 4 and atomic deposition, all of which result in nanoparticle materials with differing properties. Despite these efforts, the conventional route of incipient wetness impregnation is regarded as a reliable route, taking advantage of accessible precursor materials (inorganic metal salt) in a one-step batch process that can be scaled up for industrial production. 5 Adjustments to the impregnation method by using alternative metal precursors result in differing properties of the final nanoparticle catalyst together with any changes to temperature and atmospheric conditions of the proceeding drying and calcination steps. 6,7 The amphoteric nature of common metal oxide supports means that the pH of the impregnating solution is an important factor. Charged surface sites (e.g., Al−O − or Al−OH + ) can be created when the pH is above or below the isoelectric point of the support, which allow the impregnation of metal salts to occur via ion exchange or ligand substitution mechanisms. 8 Nanoparticles prepared from chloride containing precursors are found to suffer from residual Cl − blocking surface catalytic sites, 9−11 whereas organic precursors such as acetate or acetylacetonate as well as ammonia containing salts and nitrate precursors have been reported to yield greater dispersion over the support. 6,9,12 Despite investigations into the influence of preparation methods on the final metal dispersion and catalytic activity, 7,13 there is often lack of evidence and reasoning for the chemical processes that govern the nanoparticle formation.
In this study, genesis of PdO nanoparticles from two different impregnated Pd precursors, Pd(NO 3 ) 2 and Pd-(NH 3 ) 4 (OH) 2 , are investigated. The development of PdO nanoparticles on high surface area γ-Al 2 O 3 has been studied extensively for a number of catalytic applications: CO oxidation, 14 H 2 O 2 synthesis, 15 upgrading bio-oils, 16 and automotive three-way catalysts. 17 Of recent industrial interest is the use of Pd/γ-Al 2 O 3 in catalyzing low temperature methane oxidation for the clean and efficient operation of compressed natural gas engines (CNGs). 18−21 Mechanistic studies propose the oxidation of CH 4 to occur on a reducible PdO nanoparticle surface, close to the Al 2 O 3 interface where oxygen migration from the support can assist in reoxidation of the Pd nanoparticle surface via a Mars−Van Krevelen mechanism. 21 Therefore, the preparation of Pd/γ-Al 2 O 3 with greatest dispersion of PdO for increased interfacial contact with the support is most desirable.
The advantage of synchrotron-based techniques applied to investigate industrial catalyst preparation is highlighted in work by Espinosa-Alonso et al. for the formation of supported Ni materials. 22 In this study, by combining two noninvasive, timeresolved spectroscopic techniques, the surface and bulk properties of the precursor materials are probed in situ under a controlled calcination environment. While diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) is used to follow the vibrational modes of adsorbed and coordinated inorganic molecular species at the surface, X-ray absorption fine structure (XAFS) is used to follow the local coordination environment and oxidation state of Pd throughout the bulk of the sample. Using both techniques at the same time with online mass spectrometry of the effluent gas gives the advantage of confidently assigning features that change with response to sample environment and allow interpretation of the mechanisms involved in ligand decomposition and metal nanoparticle formation.
■ EXPERIMENTAL SECTION Sample Preparation. Two 3 wt % Pd/γ-Al 2 O 3 catalysts were prepared by incipient wetness impregnation of an aqueous solution of Pd salt onto a γ-Al 2 O 3 support (SASOL), followed by calcination (500°C , 2 h). The first sample was prepared from an acidified aqueous solution of palladium nitrate (15.11  In Situ XAFS/DRIFTS Measurements. In situ XAFS and DRIFTS measurements were conducted at B18, Diamond Light Source, Didcot, UK of the impregnated, dried precursor samples during the calcination procedure using the modified Harrick DRIFTS cell. A DaVinci arm fitted with praying mantis optics was used to refocus the IR beam outside the FTIR spectrometer for positioning of the Harrick DRIFTS cell in the X-ray beam. The Harrick XAFS/DRIFTS cell, which has been used and reported previously, 23 has an X-ray path length of 3.17 mm positioned 1.04 mm below the surface of the catalyst bed. The sample was heated to 500°C (10°C min −1 under a constant controlled flow of air (30 mL min −1 ) and maintained at 500°C for a 2h dwell period before being cooled back to room temperature. XAFS and DRIFTS spectra were recorded continuously throughout this period, and the effluent gas was monitored online by a Hiden QGA mass spectrometer. XAFS measurements were performed at the Pd Kedge in transmission mode using QEXAFS setup with a fast scanning Si(311) double crystal monochromator. All XAFS spectra were acquired with a Pd foil placed between I t and I ref , and the time resolution was 21 s spectrum −1 (k max = 15.9). DRIFTS spectra were collected with an Agilent Carey 680 FTIR spectrometer taking 64 scans with a resolution of 4 cm −1 using the liquid nitrogen cooled MCT detector. The time resolution of DRIFTS was 30 s spectrum −1 .
EXAFS Analysis. Processing extended X-ray absorption fine structure (EXAFS) data was performed using IFEFFIT with the Horae package (Athena and Artemis). Athena was used to calibrate, align, and normalize the spectra with respect to the Pd foil, for which E 0 was set at 24 358 eV. EXAFS data processing of k 2 data used an appropriate k range for the data (3.2−13.4 Å −1 ). Analysis of the Fourier transformed data was limited to fitting with the first two coordination shells (denoted Pd (O)1 , Pd (O)2 , Pd (N)1 , Pd (N)2 , Pd (Pd)1 , and Pd (Pd)2 ), using cif files of Pd(NO 3 ) 2 and PdO. The amplitude reduction factor, S 0 2 , was derived from fitting the Pd foil using a coordination number of 12 to give a value of 0.8. The value of σ 2 , the mean square relative displacement of absorber and backscatter atoms, is known to increase with temperature. 24 Consequently, the linear dependence of σ 2 with temperature was fitted using nonphase corrected Fourier transformed Pd K-edge data collected of the calcined PdO/γ-Al 2 O 3 between 500 to 20°C, where all other parameters (delr and S 0 2 ) are assumed to be fixed. This variance in σ 2 with temperature, reported in Figure S1 of the Supporting Information, was used to fix σ 2 for the fitting of the EXAFS data collected during the calcination temperature ramp period.
Ex Situ Characterization. Thermogravimetric analysis was carried out in a TA thermogravimetric analyzer (instrument Q50/DSC Auto Q20) with a sample mass of 10 mg placed in a platinum crucible under continuous flow of 40% N 2 and 60% air and ramped at 10°C min −1 . Transmission electron microscopy (TEM) images of the samples were obtained using a JEOL JEM 2100 transmission electron microscope. Samples were prepared for TEM analysis by dispersing in high-purity ethanol using ultrasonication. Ten microliters of the sonicated suspension were pipetted onto a holey carbon supporting Cu grid, and the solvent evaporated. Ex situ Pd K-edge XAFS spectra of the calcined catalyst samples after reducing in H 2 (at 100°C) were taken in transmission mode on B18 at Diamond Light Source. CO chemisorption of the calcined catalyst samples was performed with FTIR analysis in transmission mode. Samples (20 mg) were pressed into self-supporting wafers (1.35 cm 2 ), mounted in the transmission FTIR cell, and dried at 100°C under constant flow of He. Several pulses of CO were introduced into the cell until saturation. The CO adsorption FTIR spectra were recorded (40−4500 cm −1 , 4 cm −1 resolution) after purging the cell with He (15 min). Raman spectra were recorded using a Bruker Senterra microscope at B22, Diamond Light Source, using 532 nm laser at 5 mW. XRD diffraction patterns were collected using a Rigaku Miniflex diffractometer (ISIS Materials Characterization Lab) with a 600 mW Cuα tube source operated at 45

Chemistry of Materials
Article kV and 15 mA and using a 1D Si strip detector. Sherrer calculations of the average crystallite sizes used the full width half-maximum of the diffraction peaks located at the 2θ angle 55°, assuming the Sherrer constant to be 0.94. Catalytic light off curves for methane oxidation were obtained using JM automotive test rig, using a synthetic natural gas mixture (0.5% CH 4 , 10% O 2 , 0.01% C 2 H 6 , 0.01% C 3 H 8 , 0.01% NO, 10% CO 2 , and 10% H 2 O), a gas hourly space velocity of 90 000 h −1 , and a temperature ramp rate of 15°C min −1 . Effluent gases were analyzed by a MKS 2000 multigas FTIR analyzer.

■ RESULTS AND DISCUSSION
Ex Situ Characterization. Two 3 wt % PdO catalysts supported on γ-Al 2 O 3 were prepared by the impregnation and calcination from Pd(NO 3 ) 2 and Pd(NH 3 ) 4 (OH) 2 to give catalysts Pd/γ-Al 2 O 3 -ex(NO 3 ) and Pd/γ-Al 2 O 3 -ex(NH 3 ), respectively. TEM images of the calcined catalysts ( Figure 1) show dark Pd nanoparticles dispersed upon the contrasting lighter colored γ-Al 2 O 3 support. Particle size analysis conducted over 100 measured particles revealed smaller average particle diameter (3.1 nm) for catalyst Pd/γ-Al 2 O 3 -ex(NH 3 ) compared to that of Pd/γ-Al 2 O 3 -ex(NO 3 ) (4.4 nm). This is consistent with the presence of larger agglomerated particles (>15 nm) observed only for Pd/γ-Al 2 O 3 -ex(NO 3 ), as shown clearly in the corresponding histogram of particle size distribution. Although only 5% of the 100 measured particles have a diameter greater than 15 nm, this is significant when considering the mass fraction of Pd located at the core of larger particles rather than contributing to catalysis at the surface.
X-ray diffraction patterns of both calcined catalyst samples show diffraction peaks assigned to cubic γ-Al 2 O 3 (JPDS card no. 00-002-1420) similar to that of the fresh support, labeled with arrows in Figure 2, proving that structural integrity of the support is maintained after impregnation. Additional diffraction peaks located at 2θ angles of 42.6, 55.6, and 72,7°, labeled with asterisks in Figure 2 26 Ex situ Pd K-edge XAFS spectra of the calcined catalyst samples were collected for compositional and quantitative information on the Pd environment in each case. The Pd Kedge XANES show the Pd environment of each sample to be broadly consistent with PdO at room temperature (Supporting Information, Figure S3) with small differences in amplitude attributed to particle size effects. Under an H 2 atmosphere, PdO nanoparticles of both samples were completely reduced to metallic Pd. A fitting model to the EXAFS data (Figure 3a) was achieved using the first shell scattering path, Pd−Pd, of metallic Pd with σ 2 value of 0.0065 Å 2 . The amplitude of this fitted scattering feature gave Pd−Pd coordination numbers of 10.3 (4) for Pd/γ-Al 2 O 3 -ex(NO 3 ) and 7.5(2) for Pd/γ-Al 2 O 3 -ex(NH 3 ), corresponding to average particle diameters of 2.7(6) and 1.1(1) nm, respectively, using a previously reported method detailed in the Supporting Information (Tables S1 and S2). 27 CO can be used as a probe molecule in FTIR adsorption experiments to investigate surface adsorption properties of the catalyst materials. 28,29 Linear and bridge bonding of CO adsorbed on both Pd/γ-Al 2 O 3 samples were identified by the sharp band at 2071 cm −1 and the broad band between 1969 and 1830 cm −1 , respectively (Figure 3b). 30−33 The significant difference in intensity of CO absorption bands from the two catalyst samples is consistent with Pd/γ-Al 2 O 3 -ex(NH 3 ) having larger surface area due to smaller particle size. Additionally, the type of surface sites can be clarified from the relative linear to

Chemistry of Materials
Article bridge bonding adsorption band intensity ratio, reported in Table 1. The greater ratio of linear to bridge bonded CO on Pd/γ-Al 2 O 3 -ex(NH 3 ) is indicative of a rough Pd surface with greater number of step and edge sites. 32,34 The discrepancy between the average particle size calculated by the different methods, reported in Table 1, is not surprising. The TEM is limited to the resolution and field of view of the microscope; XRD is limited to assess particles with sufficiently large crystalline domains, while EXAFS probes all Pd species in the sample. Despite these differences, there is consistent evidence for Pd/γ-Al 2 O 3 -ex(NH 3 ) providing smaller average PdO crystallite size and greater dispersion over the support compared to Pd/γ-Al 2 O 3 -ex(NO 3 ). Light off curves for CH 4 conversion (Supporting Information, Figure S4) confirm that the catalyst with improved dispersion and smaller particle size, Pd/γ-Al 2 O 3 -ex(NH 3 ), is able to catalyze the complete oxidation of methane at a temperature that is lower than that of the catalyst possessing some larger agglomerated particles, Pd/γ-Al 2 O 3 -ex(NO 3 ).
In Situ XAFS/DRIFTS of the Preparation Route. Pd Kedge EXAFS analysis of the impregnated precursor samples, Pd(NO 3 ) 2 /γ-Al 2 O 3 and Pd(NH 3 ) 4 (OH) 2 /γ-Al 2 O 3 gave structural information on the initial Pd coordination environment after impregnation onto the γ-Al 2 O 3 at room temperature. The k 2 weighted χ(k) EXAFS spectrum of the impregnate Pd(NH 3 ) 4 (OH) 2 /γ-Al 2 O 3 compares closely to that of the unsupported aqueous Pd(NH 3 ) 4 (OH) 2(aq) solution ( Figure  4a). Conversely, the k 2 weighted χ(k) EXAFS spectra of the impregnated precursor Pd(NO 3 ) 2 /γ-Al 2 O 3 shows greater longrange structure from EXAFS oscillations extending to greater k range (∼10.4 Å −1 ) with additional scattering feature at 9 Å −1 (annotated with an asterisk in Figure 4b) that is absent from the χ(k) EXAFS spectrum of the unsupported Pd-(NO 3 ) 2 (H 2 O) x reference. A simple fitting model was constructed from one calculated scattering path representative of four light atomic neighboring atoms (14 ≤ atomic number, Z ≤ 16) located at 2.049 Å from the absorbing Pd atom. This model was sufficient to fit the Fourier transformed EXAFS data of unsupported Pd(NH 3 ) 4 (OH) 2(aq) and impregnated Pd-(NH 3 ) 4 (OH) 2 /γ-Al 2 O 3 (Figure 4c), showing that the Pd 2+ center in each case is coordinated to four oxygen or nitrogen atoms of ammonia, hydroxyl, or water ligands with little or no longer range structure, consistent with a [Pd(NH 3 ) 4 ] 2+ complex. This model was, however, inadequate to fit that of the impregnated Pd(NO 3 ) 2 /γ-Al 2 O 3. In this case, additional structure observed in the EXAFS data requires an additional scattering path to refine the EXAFS fitting model to this sample. A refined fit to Pd(NO 3 ) 2 /γ-Al 2 O 3 , displayed as the red dashed line of Figure 4d, includes an additional scattering contribution from neighboring Pd atoms located 3.43 Å from absorbing Pd. The distance is consistent with a Pd atom in a second coordination shell from the absorbing Pd, which corresponds to neighboring Pd centers adjoined by bridging ligand coordination. It is noted that such a Pd distance (3.43 Å) is close to that of the second coordination shell Pd atom in the crystal structure of tetrahedral PdO.
DRIFTS spectra of both precursor samples before calcination are reported in Figure 5a. The broad band centered at 3500 cm −1 can be assigned to symmetric and antisymmetric O−H stretching vibrations of lattice water, and a sharp band at 1420 cm −1 is assigned to carbonate ions on the oxide support. 35 Figure 5b). The shift in nitrate bands to greater separation is consistent with increased bidentate nitrate coordination to a metal center. 38,39 The early preassociation of Pd species of Pd(NO 3 ) 2 /γ-Al 2 O 3 , as indicated from EXAFS, can therefore also be supported by the DRIFTS spectra which provide evidence for neighboring Pd centers to be linked by bridging nitrate coordination. DRIFTS spectra of the precursor sample Pd(NH 3 ) 4 (OH) 2 or Pd(NO 3 ) 2 /γ-Al 2 O 3 at room temperature show H−N−H vibrational modes of surface NH 3 species as bands at 1490 and 1345 cm −1 . Upon increasing temperature, the positions of the bands remain but also show steady decrease in intensity until complete removal by 225°C. The detection of N 2 O in the effluent gas by online mass spectrometry up to this temperature (Supporting Information, Figure S6b) shows that the NH 3 species are oxidized before leaving the chamber, consistent with previously reported catalytic activity for low temperature NH 3 oxidation at Pd surface. Following the removal of NH 3 , weak intensity absorption bands for NO stretching vibrations of surface nitrosyl (NO − ads ) are revealed at 1510 and 1556 cm −1 , respectively, highlighted by the red circle in Figure 5c. The formation of NO − can be attributed to oxidative decomposition of NH 3 ligands, which then form strong interaction with the γ-Al 2 O 3 surface. The nitrosyl absorption bands are present in the DRIFTS spectra up to 400°C, which is consistent with previous temperature studies for stability of NO − adsorbed onto the γ-Al 2 O 3 surface. 40−42 The M/Z 30 signal detected by online mass spectrometry of the effluent gas from sample Pd(NH 3 ) 4 (OH) 2 /γ-Al 2 O 3 between 390−450°C (Supporting Information, Figure S6b) indicates a release of NO − (ads) as NO gas at this temperature. Note that in the case of the Pd(NO 3 ) 2 / γ-Al 2 O 3 sample, the formation of NO − species is unlikely as the

Chemistry of Materials
Article nitrogen of the nitrate ligands is already in higher oxidation state and thus is released in the oxidizing environment directly as NO gas (or NO 2 ), which is detected as M/Z 30 between 100 and 400°C (Supporting Information, Figure S6a).
The nucleation and growth of PdO nanoparticles from each impregnated precursor can be followed from the amplitude of the scattering features of the Fourier transform Pd K-edge EXAFS data collected during the calcination ramp period, shown in Figure 6. The scattering features at 2.5 and 3 Å correspond to scattering from nearest neighbor Pd atoms (located at real distances of 3.05 and 3.43 Å) in the tetrahedral PdO crystal structure. 43,44 An increase in amplitude of these features at higher temperatures, despite increasing thermal motion and thus increased dampening from thermal disorder, indicates an increase in coordination to neighboring Pd atoms and thus growth of the PdO structure. For Pd(NO 3 ) 2 /γ-Al 2 O 3 , the scattering feature at 3 Å in the Fourier transform EXAFS data is first observed at room temperature ( Figure 4d) and    Figure 6b by the formation and growth in amplitude of the scattering feature at 3 Å. The coordination number of neighboring palladium atoms located at 3.43 Å from the absorbing palladium is plotted against temperature for both samples (Supporting Information, Figure S7).  45 It is worth noting that the temperature for NO − desorption from this sample, evidenced by loss of NO stretching bands in DRIFTS spectra and presence of NO (g) in effluent gas between 360−400°C, correlates with the temperature at which Pd−Pd scattering paths were first identified from the EXAFS data. The stabilizing effect of NO species has been previously postulated in multiple studies by Sietsma et al., whereby introducing NO gas into the calcination atmosphere was shown to suppress nanoparticle agglomeration over metal oxide support. 46,47 It is possible that a combination of both high initial Pd dispersion and NO − (ads) stabilization contributed to the later formation and suppressed growth of PdO nanoparticles from the Pd(NH 3 ) 4 (OH) 2 precursor.

■ CONCLUSIONS
Through use of a combined XAFS/DRIFTS approach to study the formation of supported nanoparticles during calcination of impregnated metal oxides, a molecular insight into the role of different inorganic precursors was achieved. Pd/γ-Al 2 O 3 prepared from incipient wetness impregnation of aqueous Pd(NH 3 ) 4 (OH) 2 solution benefits from smaller Pd particle size and improved dispersion over the support, which in turn favors a lower light off temperature for CH 4 oxidation compared to that of the catalyst prepared from Pd(NO 3 ) 2 .
The nucleation and growth of PdO nanoparticles during calcination was identified from the EXAFS data by an increase in the Pd−Pd scattering contribution located at a distance (3.43 Å) consistent with a Pd−Pd distance of crystalline PdO. For Pd(NO 3 ) 2 /γ-Al 2 O 3 , early association of Pd neighbors at this distance (3.43 Å) was observed in the EXAFS data of the sample before heat treatment; the bridging nature of nitrate ligands in this sample, evidenced by NO and O−N−O nitrate stretching frequencies in DRIFTS spectra, provides a rationale for the early assembly of Pd centers via bridging mechanism. Upon calcination of the Pd(NO 3 ) 2 /γ-Al 2 O 3 sample, there was a pronounced increase in the Pd−Pd scattering component at 3.43 Å between 200−250°C, indicative of significant growth of PdO nanoparticles at this temperature. Conversely, the Pd centers from precursor Pd(NH 3 ) 4 (OH) 2 /γ-Al 2 O 3 appear to adopt isolated sites upon impregnation, evidenced by the absence of any Pd−Pd distances from the Pd EXAFS data, which can be explained by fast ion exchange of [Pd(NH 3 ) 4 ] 2+ with deprotonated Al− O − surface sites of the support. Although NH 3 ligands were shown to decompose from the impregnated sample below 200°C , the formation of PdO nanoparticles (observed by Pd−Pd scattering from EXAFS data) did not occur until temperatures beyond 360°C. Most NH 3 was partially oxidized and released as N 2 O gas at 210°C, but small amounts of NO − were captured by the γ-Al The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.chemmater.7b02552.
EXAFS fitting parameters, Raman emission spectra, XANES spectra, EXAFS particle size calculations, catalytic light off curves, reference DRIFTS spectra, TGA curves, and mass spectrometry data (PDF) this study. Mark Frogley at B22, Diamond Light Source is thanked for assistance with operating the Raman microscope. Johnson Matthey is acknowledged for their provision of precursor materials and catalyst testing facility. The UK Catalysis Hub is kindly thanked for resources and support provided via our membership of the UK Catalysis Hub Consortium. UCL and EPSRC are thanked for the iCASE studentship of E.K.D.

■ ABBREVIATIONS
DRIFTS, diffuse reflectance infrared Fourier transform spectroscopy EXAFS, extended X-ray absorption fine structure XANES, X-ray absorption near-edge structure TEM, transmission electron microscopy FTIR, Fourier transformed infrared XAFS, X-ray absorption fine structure XRD, X-ray diffraction