Comprehensive Experimental and Theoretical Study of the CO + NO Reaction Catalyzed by Au/Ni Nanoparticles

The catalytic and structural properties of five different nanoparticle catalysts with varying Au/Ni composition were studied by six different methods, including in situ X-ray absorption spectroscopy and density functional theory (DFT) calculations. The as-prepared materials contained substantial amounts of residual capping agent arising from the commonly used synthetic procedure. Thorough removal of this material by oxidation was essential for the acquisition of valid catalytic data. All catalysts were highly selective toward N2 formation, with 50–50 Au:Ni material being best of all. In situ X-ray absorption near edge structure spectroscopy showed that although Au acted to moderate the oxidation state of Ni, there was no clear correlation between catalytic activity and nickel oxidation state. However, in situ extended X-ray absorption fine structure spectroscopy showed a good correlation between Au–Ni coordination number (highest for Ni50Au50) and catalytic activity. Importantly, these measurements also demonstrated substantial and reversible Au/Ni intermixing as a function of temperature between 550 °C (reaction temperature) and 150 °C, underlining the importance of in situ methods to the correct interpretation of reaction data. DFT calculations on smooth, stepped, monometallic and bimetallic surfaces showed that N + N recombination rather than NO dissociation was always rate-determining and that the activation barrier to recombination reaction decreased with increased Au content, thus accounting for the experimental observations. Across the entire composition range, the oxidation state of Ni did not correlate with activity, in disagreement with earlier work, and theory showed that NiO itself should be catalytically inert. Au–Ni interactions were of paramount importance in promoting N + N recombination, the rate-limiting step.

Nanoparticles synthesis: Ni colloidal particles were synthesized as follows; Ni(NO3)2.6H2O (0.136 g, 0.468 mmol) and poly(4-vinylpyridine) (0.492 g, 4.68 mmol) were added to ethylene glycol (60 mL) and stirred overnight at 50 °C. The resulting solution was light blue. The solution was then cooled to 0 °C using an ice bath and dioxane (50 mL) was added. The pH was then adjusted to 9-10 by the addition of aqueous NaOH (4 mL, 1 M). The temperature was raised to 180 °C and stirred for 30 min., this caused the solution to undergo a colour change from light blue to yellow. Aqueous NaBH4 (2 mL, 0.5 M) was added to the solution at 180 °C, this lead to a final colour change from yellow to black. The colloid solution was stirred and cooled to room temperature. The resulting nanoparticle solution was purified as described below.
Monometallic Au particles were synthesized in a similar way, poly(4-vinylpyridine) (0.492 g, 4.68 mmol) was added to ethylene glycol (60 mL) and stirred at 80 °C for 3 hours. The temperature was increased to 100 °C for 5 minutes and the pH was adjusted to 9-10 by the addition of aqueous NaOH (3 mL, 1 M), producing a colour change from colourless to light yellow. At this time a solution of NaAuCl4.2H2O (0.186 g, 0.468 mmol) in Milli-Q water (20 mL) was added under vigorous stirring, the resulting colour turned an immediate deep purple, the temperature remained at 100 o C for a further 2 hours and allowed to cool to room temperature under sustained stirring. The resulting nanoparticle solution was purified as described below.
AuxNi(100-x) alloy nanoparticles were synthesized by combining elements from the two methods previously mentioned, various metal contents were used to generate alloys of different metal compositions. The quantities and metal concentrations can be seen in Table S1; Ni(NO3)2.6H2O and poly (4-vinylpyridine) were added to ethylene glycol (56.2 mL) in a 1:10 molar ratio and stirred overnight at 50 °C, the resulting solution was light blue. This was then cooled to 0°C using an ice bath. The pH was then adjusted to 9-10 by the addition of aqueous NaOH (3 mL, 1 M). At this time a solution of NaAuCl4.2H2O in Milli-Q water (23.4 mL) was added under vigorous stirring, the temperature was maintained at 0 o C for 6 hours, this induced a series of colour changes from blue/yellow to red/purple. The colloidal solution was allowed to warm to room temperature while stirring. The resulting nanoparticle solutions were purified as described below. Nanoparticle purification and redispersion: Excess acetone was added to the colloid in a 10:1 volume ratio and shaken for 30 seconds. The mixture was left at room temperature overnight without stirring and the particles sedimented. If the particles did not sediment overnight, the sample was centrifuged in 50 mL aliquots at 6000 rpm for 15 minutes and the sediment isolated by discarding the supernatant using a Pasteur pipette. In either case, the sediment was redispersed in the minimum amount of ethanol using an ultrasonic bath, and the process was repeated twice. The final sediment was redispersed in ethanol. All samples could be purified immediately or after a delay of up to six weeks with no observable change to the product.
Catalyst synthesis from purified nanoparticle solutions: The γ-alumina support was provided by Toyota Motor Corporation (TMC). The specific surface area of the γ-alumina was 160 m 2 g -1 .
To load nanoparticles onto the support, the alumina was added to a reaction vessel under air and some quantity of the nanoparticle solution added. The volume of the suspension was kept constant between samples. The suspension was stirred for 2 h at room temperature and then sonicated for ten minutes using an ultrasonic bath operating at 200 W. The solvent was removed by rotary evaporation and the resulting powder was obtained.

Temperature programmed oxidation
Temperature programmed thermogravimetric analysis experiments were performed using a Perkin Elmer Pyris 1 TGA-MS, equipped with a 500 amu Hiden Mass spectrometer. ~ 5 mg of the as-prepared catalyst (Au75Ni25/α-Al2O3) was heated in flowing 20% O2, He balance (30 mL min -1 ) ramping temperature at 3 °C min -1 , while monitoring the mass loss and m/z 18 and 44 downstream of the sample via a capillary inlet to the mass spectrometer.

Catalyst Testing
Prior to testing the catalysts were calcined in in synthetic air (O2 21%, He balance) at 500 °C (heating rate = 10 °C min -1 , hold time = 1 h) followed by reduction in 5% H2/He mixture of a total flow of 50 mL min -1 . Reduction was conducted in situ by heating at 10 °C min -1 to 550 °C and cooling under 5% H2 in He before admitting the reactive gases. The catalysts were tested in a single-pass, fixed-bed reactor operating at atmospheric pressure with a GSV = 1.6 litres min -1 gcat -1 . Reactant and product analyses were performed using a Siemens Ultramat NDIR NO/N2O detector and an MKS quadrupole mass spectrometer. The reactor was operated with total gas flowrates in the range 100 ml min -1 . The gas flow in a typical run consisted of 1000 ppm of NO and 1000 ppm of CO diluted in He. The temperature measurement was by means of a K-type thermocouple in contact with the reactor wall and the heating rate during light-off measurements was 3 °C/min.

Electron microscopy
Au, Ni, and AuNi nanoparticles and the corresponding supported catalysts were analysed by TEM at the Electron Microscopy Suite, Department of Physics, Mott Building, University of Cambridge. Analysis used either a FEI Philips Tecnai f20 or an FEI Philips Tecnai 20 highresolution transmission electron microscope operating at 200 keV. Sample preparation was by droplet coating of ethanol suspensions on holey carbon films on 300 mesh copper grids (AGS147-3H, Agar Scientific). Electron optical parameters: CS = 0.6 mm, CC = 1.2 mm, electron energy spread = 1.5 eV, beam divergence semi-angle = 1 mrad). Particle sizes were analysed using ImageJ by counting the diameters of more than 100 particles in multiple lower magnification images, where the diameter was measured along the longest axis and size distributions were produced using Microsoft Excel.

ED Spectroscopy
EDS analysis was performed with an Oxford Instruments Inca analyser fitted in a Jeol 2010 transmission electron microscope running at 200 kV using a Gatan Ultrascan 4000 digital camera for image acquisition.

ICP-OES of unsupported nanoparticles
To prepare samples of the colloids for inductively coupled plasma optical emission spectroscopy (ICP-OES), 1 mL of sample was syringed into 1 mL of aqua regia (7:3 mixture of hydrochloric and nitric acid, respectively) and then 8 mL of Milli-Q water was added. A rubber crimp-top vial was used to reduce the risk of leakage. The aqua regia solution turned yellow immediately upon mixing the acids, and then turned gradually dark yellow over the course of 30 minutes due to the generation of nitrosyl chloride and chlorine gas. Upon addition of the colloid, a colour change occured that is characteristic of the metal ion being dissolved. The acidic solution was left for one hour to ensure complete dissolution of the nanoparticles. Commercial standards were used to calibrate for all metals. Au and Ni metal contents were determined by ICP-OES (PerkinElmer Optical Emission Spectrometer Optima 5300 DV).

Powder X-ray diffraction
PXRD patterns were acquired on a PANalytical Empyrean diffractometer fitted with an X'celerator detector and using a Cu-Kα1 source operating at 40 kV and 40 mA with a step size of 0.02°. Samples were powdered with a pestle and mortar and loaded onto a silicon zero background holder for analysis. PXRD patterns of samples 1-15 were collected, and select samples were then dried in an oven and patterns reacquired to ensure that moisture had not interfered with the data collection. In these cases, no change was confirmed. Scherrer analysis was performed on Au nanoparticles.

X-ray photoelectron spectroscopy
XPS was performed in a customized system incorporating a hemispherical analyser (SPECS Phoibos 100), a non-monochromatized x-ray source (Al Kα; 1486.6 eV) and a high temperature-high pressure cell (SPECSHPC-20) with infrared sample heating. The analyser was operated at a fixed transmission and 50 eV pass energy with energy step of 0.1 eV. Binding energies were calibrated using Al 2p (74.0 eV) as an internal reference. The high temperaturehigh pressure cell design allowed sample heating up to 800 °C, under flow or static conditions, at pressures up to 20 bar. This arrangement enabled fast post-reaction sample transfer from the reaction chamber to the spectrometer chamber, whilst maintaining reaction gas or UHV conditions -i.e. without exposure to laboratory air. Prior to analysis, samples were evacuated to a base vacuum of 10 −7 mbar at room temperature. In a typical experiment that utilised the cell, the catalyst sample was initially placed in the spectrometer chamber and XP spectra were acquired. The sample was then transferred under vacuum to the high pressure cell were it was exposed to the reactive gases and heated to the appropriate temperature. After the treatment the sample was cooled to room temperature in reaction gas atmosphere, evacuated and then it was transferred back to the spectrometer chamber for analysis, without exposure to laboratory air.

X-ray absorption spectroscopy
XAS experiments were performed at beamline B18, Diamond Light Source, Oxfordshire, UK. A double-crystal Si(111) monochromator was used to scan X-ray energy from -100 to 800 eV relative to Ni K-edge (8,333 eV) and -200 to 900 eV relative to the Au L III -edge (11,919 eV). Each sample (sufficient that the absorption coefficient jump at the edge was satisfactory for transmission measurements) was prepared by pressing finely ground powders into selfsupporting disks using an IR die press and then inserting in a modified commercial infrared in situ reaction chamber (Specac), able to work up to 800 °C under a controlled atmosphere. 1 Gas flow was controlled by means of mass flow controllers, calibrated volumetrically for the specific gas mixtures immediately prior to use. The in situ cell was then employed for transmission XAS measurements. Each sample was measured in a sequential series of gas and temperature conditions, following the ex situ calcination treatment as used for the samples employed in catalyst evaluation. This enabled the reduction step and subsequent behaviour in reactive gases during catalysis to be monitored. The sequence was: (1) as prepared; (2) 50% H2, 550 °C; (3) 50% H2, cooled to 150 °C; (4) 1% CO, 1% NO 150 °C; (5) 1% CO, 1% NO 300 °C; (6) 1% CO, 1% NO 400 °C; (7) 1% CO, 1% NO 550 °C; (8) 1% CO, 1% NO cooled to 150 °C (He balance with total gas flow rate 100 mL min -1 in all cases).
A minimum of three spectra are acquired at each condition following 10 minutes stabilising time and the spectra compared visually to ensure the catalyst is stable. Pure metal foils of the respective elements measured in tandem between a second and third ionisation chamber, enabled X-ray energy calibration and data alignment. XANES spectra were aligned by calibration of the first zero crossing point of the second derivative of the reference metal foil reference spectra in each case. XAS data processing was performed using IFEFFIT 2 with the Horae package 3 (Athena and Artemis). The amplitude reduction factor, S0 2 was derived from EXAFS data analysis of the respective metal foil reference spectra (for which the co-ordination numbers of the f.c.c. metal are known), yielding a value of 0.85 for both elements. This was then fixed in the analysis of sample spectra. The parameters corresponding to the correction to the photoelectron energy origin, co-ordination numbers, bond lengths, and mean-squared relative deviation of atoms around absorbing atoms were then varied during fitting. For Au-Ni or Ni-Au co-ordination, as a starting point an input file was used for which the host metal was modified by placing an appropriate number of the other metal in the atoms lists and modifying the lattice parameter based on Vergard's law, as is seen to be the case for known AuNi alloys. 4,5 For the Ni K-edge Ni-Ni, Ni-Au and Ni-O co-ordination was considered based on f.c.c. Ni metal, Au-modified Ni (described above) or cubic NiO (for the first co-ordination shell using other morphologies of NiO makes no difference). Ni-O-Ni from the NiO was also considered in low temperature samples where it was not possible to fit the data reasonably without this contribution. For the Au L III -edge Au-Au and Au-Ni first neighbour co-ordination was used to fit the first co-ordination shell. Multiple scattering paths were considered, but found to contribute minimally in the first co-ordination shell region due to the longer effective path length. In order to ensure the number of fit-parameters remained significantly below the Nyquist number of independent data points, where necessary, the photoelectron energy origin for higher temperature samples (>150 °C) was fixed based on the average of all conditions at 150 °C, and similarly the mean-squared relative deviation of atoms around the absorber was assumed to vary by 10 -5 Å 2 K -1 . This linear variation of the mean squared displacement is consistent with reported Deybe correlated model data for Pt and Ge; 6 an approach recently used for AuPd bimetallic nanoparticle analysis 7 and the data obtained in the present study for the pure monometallic Au and Ni particles as a function of temperature. Whilst fixing parameters in this way may introduce small systematic errors, it enables more reliable trends to be seen without over-interpretation of the data. Physically reasonable constraints were applied such that the co-ordination numbers for individual contributing paths should not exceed their crystal close packed structure and the total co-ordination should not exceed 120% of the contribution found for their respective crystal structures (slightly over 100% to allow for the magnitude of typical fitting errors and / or disorder). All fits were performed using multiple k-weight fitting, the k 3 -weighted data is shown in all figures.

Computational details
DFT calculations were performed using the VASP 5.4 program. 8,9,10 In these calculations the energy was computed using the GGA functional as proposed by Perdew et al. (PBE). 11 The effect of the core electrons on the valence states was represented with the projector-augmented wave approach. 12 The valence electronic states were expanded using a plane wave basis set with a kinetic energy cutoff of 400 eV, which ensures adequate convergence of energetic and structural parameters with respect to basis set size. All calculations on Ni surfaces were performed sampling the Brillouin zone using a Γ centered 4×4×1 mesh of points. 13 Iterative relaxation of the atomic positions was stopped when the forces on the ions were <0.01 eV Å -1 .
To obtain faster convergence, thermal smearing of the one-electron states (kBT = 0.2 eV) was allowed, using the Gaussian smearing method to define the partial occupancies. Transition state calculations were undertaken using the climbing image version of the nudged elastic band algorithm. 14 The Ni (111) surface was modeled by a (3×3×5) supercell with a unit cell parameter of a = 3.530 Å containing 45 atoms and a 15 Å vacuum space between the slab surfaces. This model is large enough to accommodate reactants and products and to test different Au/Ni substitutions. In the geometry optimizations all layers of the slab were allowed to relax. The Au11Ni34 model was constructed by performing one Au/Ni atom substitution at a time checking all possible distributions and selecting the more stable one at each step. The model contains four Au atoms in the top layer, another four Au atoms in the first layer below the surface and three Au atoms in the second sublayer (Main Text, Figure 9). The possible effect of the relaxation of the a and b cell parameters was evaluated in the models with the higher Au content and it was found to be negligible. The Ni (111) stepped surface was modeled by a (6×3×5) supercell and taking out three consecutive rows of Ni atoms along the [110] direction. This creates two steps showing on the riser the (100) face (left side of Figure S1) and a surface equivalent to the (111) face (left side of Figure S1) -the reactivity of the step was only examined on the step that shows the (100) face. The Au20Ni61 model was constructed following the procedure described for the Au11Ni34 flat surface model, i.e. performing one Au/Ni atom substitution at a time on various configurations and selecting the more stable one for the next step. The unit cell for NiO is of halite type, with a small rhombohedral distortion introduced by an antiferromagnetic ordering of the Ni atoms' magnetic moments along the [111] direction. To model the NiO (100) surface we have employed a (2×2×4) supercell slab model, using an optimized cell parameter of a = 4.162 Å (experimental value, a = 4.1684 Å) 15 and an effective Hubbard U (Ueff = U-J) of 5 eV, similar to other values reported in the literature. 16 In all NiO surface calculations, Brillouin zone integrations were performed at the Γ point with a kinetic energy cutoff of 500 eV.

C 1s XP spectra after exposure to pre-treatment and reactive gases
The following spectra show a comparison of Au50Ni50/α-Al2O3 to Ni100/α-Al2O3 after calcination as a function of exposure to reactive gases using a high pressure cell connected directly to the XP spectrometer to prevent air exposure and contamination as a result of sample transfer. Figure S8: C 1s XP spectra as a function of the conditions the sample has been exposed to: (1) after calcination as loaded into the XP spectrometer (denoted "Calc"); (2) after exposure to the reduction conditions of 550 °C in H2; (3) after exposure to reaction conditions at (380 °C and then) 550 °C in 1000 ppm NO, 1000 ppm CO. Au50Ni50 and Ni100/α-Al2O3 supported catalyst samples were studied, shown along with the conditions in the figure legend. The spectra show a small amount of residual surface carbon after high temperature calcination, but strikingly almost no carbon at all after the exposure to reaction conditions. The spectra are calibrated to the Al 2p, which has a value of 74.0 eV for Al in the Al2O3 support. Signal intensities are as acquired and not normalised.
XP spectra after calcination, before reduction

Ni K-edge and Au L-edge co-ordination number summary plots based on EXAFS fitting
Ni K-edge (note the sample with 25% Ni could not be reliably fitted):

EXAFS fitting parameters
Ni K-edge:

EXAFS fits
Each figure shows the name of the relevant data set at the top of the figure and contains the kspace data and fit (top left), R-space magnitude data and fit (bottom left) and the real part of the data and fit, along with the contributing paths (right). The paths are named such that the material is given square brackets ('nickel' = f.c.c. nickel metal; 'gold' = f.c.c. metallic gold; 'NiOx' = nickel oxide) followed by the identity of the scattering atom (the '1.1' can be ignored, but denotes the first shell of that type of atoms in the structure). If the material is based on an alloy where an intermediate combination exists the alloy composition is given as the material in square brackets.

Ni K-edge data
Pure Ni sample Figure S25: Pure Ni sample after calcination and before treatment with gases or temperature, Ni K-edge EXAFS spectroscopy fitted data.       Figure S32: Au25Ni75 sample after calcination and before treatment with gases or temperature, Ni K-edge EXAFS spectroscopy fitted data.        Figure S40: Au50Ni50 sample after calcination and before treatment with gases or temperature, Ni K-edge EXAFS spectroscopy fitted data.                              Figure S78. Computed reaction pathway for the 2Nads (adjacent) → 2 − ads chemical step (see Figure 9, main text) for lower Au content. The final state is taken as a reference at 0 eV. The legend indicates the ratio "Au/Ni" used in the calculations.