Determination of CO , H 2 O and H 2 coverage by XANES and EXAFS on Pt and Au during water gas shift reaction

The turn-over-rate (TOR) for the water gas shift (WGS) reaction at 200 1C, 7% CO, 9% CO2, 22% H2O, 37% H2 and balance Ar, of 1.4 nm Au/Al2O3 is approximately 20 times higher than that of 1.6 nm Pt/Al2O3. Operando EXAFS experiments at both the Au and Pt L3 edges reveal that under reaction conditions, the catalysts are fully metallic. In the absence of adsorbates, the metal–metal bond distances of Pt and Au catalysts are 0.07 Å and 0.13 Å smaller than those of bulk Pt and Au foils, respectively. Adsorption of H2 or CO on the Pt catalysts leads to significantly longer Pt–Pt bond distances; while there is little change in Au–Au bond distance with adsorbates. Adsorption of CO, H2 and H2O leads to changes in the XANES spectra that can be used to determine the surface coverage of each adsorbate under reaction conditions. During WGS, the coverage of CO, H2O, and H2 are obtained by the linear combination fitting of the difference XANES, or DXANES, spectra. Pt catalysts adsorb CO, H2, and H2O more strongly than the Au, in agreement with the lower CO reaction order and higher reaction temperatures.


Introduction
The water gas shift (WGS) reaction (eqn (1)), is an industrially important reaction for H 2 production and CO removal. 1,2It is mildly exothermic (DH = À40.6 kJ mol À1 ), thus it is thermodynamically favored at lower temperatures.However, due to kinetic limitations, the reaction is typically conducted at temperatures between 200 and 450 1C.In commercial operation, WGS is, typically, a two-stage process with a high-temperature stage (320-450 1C) employing iron oxide-based catalysts and a low-temperature stage (180-250 1C) employing copper-based catalysts.
CO + H 2 O -CO 2 + H 2 (1)   However, such two-stage process designs are not viable for small-scale applications.It was predicted that until 2030, approximately 10% of the world annual energy consumption will originate from the WGS reaction. 1Thus, the development of new, higher activity low-temperature WGS catalysts is of scientific and practical interest.Although commercial Cu/ZnO/Al 2 O 3 WGS catalysts operate at low temperature, they suffer from poor stability under constant start-up-shutdown conditions and display poor sulfur tolerance. 3,4In the past decade, Pt and Au-based WGS catalysts have received intense attention as alternatives to Cu. [5][6][7][8][9][10][11][12][13][14] The former are not pyrophoric and require no exceptional pretreatment protocol.In general, Au catalysts operate at a lower temperature than Pt.Moreover, kinetic studies have shown different reaction orders on Au than on Pt.The reasons for the high rate per mole of Pt and Au compared to other metals and the differences in their kinetics are not fully understood.
In X-ray Absorption Near Edge Spectroscopy (XANES), the Pt and Au L 3 edges correspond to the 2p -5d electronic transition, and therefore, the intensity and position of the XANES spectra are directly related to the 5d electronic structures.Most often XANES is used to determine the metal oxidation state, fraction of metallic and oxidized metal, formation of alloys, etc.6][17] So far, no studies have reported analysis of the difference XANES spectra for identification of the surface coverage of adsorbates during the WGS reaction in order to elucidate the factors that govern turn-over-rates (TOR) and reaction orders.
In the present study, operando EXAFS and XANES have been used to identify the active sites on Pt nanoparticles supported on SiO 2 and Al 2 O 3 and Au/Al 2 O 3 WGS catalysts.Adsorption of CO, H 2 , and H 2 O induce changes in the shape and position of the Pt and Au L 3 XANES spectra.The XANES difference, or DXANES, method is used to identify adsorbates and quantify their individual coverage and under steady-state WGS reaction conditions (CO + H 2 O + H 2 ).The CO surface coverage determined by FTIR agree well with those obtained by the DXANES analysis.The different reaction orders between Pt and Au are explained by the difference in CO surface coverage.In addition it is suggested that weak adsorption of CO leads to exposed sites on Au at low temperature, resulting in catalytic activity at lower temperature than that on Pt.Finally, the differences in adsorption properties of Pt and Au suggest potential opportunities for the development of new, low-temperature WGS catalysts.

Experimental
2.1 Catalyst preparation 4.3% Pt/SiO 2 : 50 g of Davisil silica (Sigma-Aldrich, 280 m 2 g À1 , 1.1 cc/g) was slurried in 250 mL of deionized (DI) H 2 O with about 2 mL of concentrated NH 4 OH.The pH was approximately 10. 5.0 g of Pt(NH 3 ) 4 (NO 3 ) 2 (PtTA) was dissolved in 250 mL of DI H 2 O and 2 mL of concentrated NH 4 OH was added.The PtTA solution was rapidly added to the stirring silica.After 15 min, the silica was settled and the solution was decanted.The Pt/silica was slurried in 200 mL of cold DI H 2 O for 15 min and the solution was decanted.Following a second wash and decanting of the solution, the wet solid was filtered and washed on the filter with 200 mL of DI H 2 O.The catalyst was dried overnight at 100 1C and calcined at 225 1C in flowing air for 5 h.The elemental composition was determined to be 4.3 wt% Pt by ICP analysis.By means of the double isotherm method, volumetric hydrogen and CO chemisorption indicated a dispersion of 0.95 H/Pt and 0.55 CO/Pt.
2.6% Pt/Al 2 O 3 : The solution of 2.3 g of Pt(NH 3 ) 4 (NO 3 ) 2 (PtTA) in 23 mL of DI H 2 O was added to 45 g of g-alumina (200 m 2 g À1 , 0.5 cc/g).The catalyst was dried overnight at 110 1C and calcined in flowing air at 450 1C for 5 h.The elemental composition was determined to be 2.6 wt% Pt by ICP analysis.The H 2 and CO volumetric chemisorption values were 1.0 H/Pt and 0.85 CO/Pt.0.7% Au/Al 2 O 3 : The 0.71% Au/alumina catalyst (BC17) was provided by the World Gold Council.

XAS reactor description
2.2.1 Operando XAS reactor.The temperature of the operando, fixed-bed plug flow reactor was controlled in an Al heating block.A type K thermocouple was mounted inside the reactor at the top of the catalyst bed to measure the reaction temperature.The design details of the reactor and its validation as a true plug flow reactor can be found elsewhere. 20The flow rates of gases (CO, H 2 , Ar) to the catalyst bed were controlled using mass flow controllers.The concentration of water was controlled by saturation of the reaction gas at different temperatures, thereby varying the partial pressure.The lines from the water saturator to the reactor were heat traced to prevent water condensation.The exit water vapor was condensed in an ice bath prior to injection of the products to the on-line gas chromatograph (GC).
Approximately, 10 mg of the Pt/Al 2 O 3 catalyst was placed on a 3 mm plug of SiO 2 powder (Davisil 644, 280 m 2 g À1 ) supported on Pyrex wool in a quartz tube reactor (O.D.: 0.25 00 , I.D.: 0.123 00 ), to achieve a level catalyst bed and clear visual distinction between the Pyrex wool, silica and catalyst.The catalyst bed height was approximately 6 mm.
The catalysts were pre-reduced at 200 1C and 280 1C for Au and Pt, respectively.The concentrations of the individual reactants for the WGS reaction were 6.8% CO in Ar, 20% H 2 in Ar, or 12.3% H 2 O in Ar.The total gas flow rate over the Pt/Al 2 O 3 catalyst in each experiment was constant at 10 mL min À1 , yielding a flow-rate-to-catalyst mass ratio of 0.91 mL min À1 mg cat À1 .The total gas flow rate over the Au/Al 2 O 3 catalyst in each experiment was 15 mL min À1 giving a flow rate to catalyst mass ratio of 0.27 mL min À1 mg cat À1 .The kinetic rates for WGS were determined by on-line gas chromatography at 120 and 200 1C for Au and 200 and 280 1C for Pt WGS, and agree with those determined in a laboratory reactor, which used larger amounts of catalyst.

XAFS measurements
X-Ray absorption measurements were conducted on the insertion device beamline of the Materials Research Collaborative Access Team (MRCAT, 10-ID) at the Advanced Photon Source (APS), Argonne National Laboratory.Ionization chambers were optimized for the maximum current with linear response (ca. 10 10 photons detected per sec.)using a mixture of N 2 and He in the incident X-ray detector and a mixture of ca.20% Ar in N 2 in the transmission X-ray detector.A third detector in the series collected a reference spectrum (Au or Pt foil) simultaneously with each measurement for energy calibration.The catalyst supports were spray-dried microspheres of about 100-200 mesh, which allowed the samples to be loosely packed without bed plugging.The reactor composition, diameter and wall thickness was chosen to give a total absorbance (mx) at the Pt L 3 (11.56keV) edge or Au L 3 (11.92keV) edge between 1-2 and edge steps (Dmx) between about 0.3-0.5.Three spectra were obtained in quick scan mode in about 4 min and were averaged for data analysis.For both reactors, the EXAFS and XANES spectra of the catalysts with adsorption of CO, H 2 , H 2 O or WGS gas mixture were obtained at room temperature, 120 1C, 200 1C and for Pt 280 1C.The gases were purified to remove traces of oxidants (air) by passing through a Matheson PUR-Gas Triple Purifier Cartridge.

XAFS data analysis
Phase shift and backscattering amplitudes were obtained from the Au and Pt foils for Au-Au and Pt-Pt scattering, respectively.Standard procedures based on WINXAS 3.1 software were used to fit the XAS data.The EXAFS coordination parameters were obtained by a least square fit in q-and r-space of the isolated nearest neighbor, k 2 -weighted Fourier transform data.The quality of the fits were equally good with both k 1 and k 3 weightings.The EXAFS data and fits were obtained for reduced catalysts at 200 1C for Au and 280 1C for Pt and room temperature.A linear temperature dependence of the Debye-Waller factor (DWF) was assumed in order to calculate the values at intermediate temperatures. 21

Difference (D) XANES spectra
The normalized, energy calibrated Pt L 3 and Au L 3 edge XANES spectra were obtained by standard methods.The XANES spectra were fit with a linear combination of the catalyst in He (no adsorbates) and with individual gases.For CO and H 2 the reference XANES spectra correspond to the coverage at room temperature, while for H 2 O the reference XANES spectra correspond to the relative coverage at room temperature on Au, but 280 1C on Pt, i.e., the temperature for the WGS reaction.
The DXANES spectra were obtained by subtracting the XANES spectrum of the supported catalyst in He from that with different reaction gases at various temperatures.For single gases (CO, H 2 , and H 2 O), the difference XANES spectra at elevated temperatures were fit using the room temperature DXANES as references.This gives a relative fraction of the adsorbed gas at that temperature compared to the amount adsorbed at the reference temperature.For mixed gas compositions, the DXANES spectra were fit as a linear combination of the DXANES spectra for each single adsorbate.

Laboratory testing of the WGS reaction
The catalytic activities of the Pt/Al 2 O 3 , Pt/SiO 2 , and Au/Al 2 O 3 catalysts were determined using a plug-flow, laboratory reactor, which has been described elsewhere. 22For each experiment, 250 to 300 mg of the catalyst was pre-treated in situ by drying at 100 1C in a flowing inert atmosphere, followed by heating to 200 1C for Au and 300 1C for Pt at 50 mL min À1 in 25% H 2 /75% Ar with a ramping rate of 5 1C min À1 .In this fashion, the Pt/Al 2 O 3 , Pt/SiO 2 , and Au/Al 2 O 3 catalysts were reduced for two hours.After reduction, the catalysts were exposed to a standard WGS composition of 6.8% CO, 8.5% CO 2 , 21.9% H 2 O, 37.4% H 2 and balance Ar.For all kinetic experiments, the total pressure was kept constant at ambient pressure with a total inlet flow rate of 75.4 mL min À1 .The Pt/Al 2 O 3 , Pt/SiO 2 , and Au/Al 2 O 3 catalysts were stabilized at 300 1C, 260 1C, and 200 1C respectively, under the standard gas composition for 15 to 20 h.Water in the reaction gas was condensed in an ice bath and the dry exit stream from the reactor was periodically injected into an Agilent 6890 GC.The dry inlet gases were analyzed before each injection to determine the response factors of the detectors and ensure precise measurements.The GC is equipped with a thermal conductivity detector and a Carboxen 1000 column operating with helium as a carrier gas.
After the stabilization process, the reactor temperature was adjusted to lower the CO conversion below 10% and maintain differential conditions during kinetic measurements.The rate of CO consumption was used to calculate the WGS rate under differential conditions.Reaction orders for the reactant and product gases were determined by varying each gas concentration independently.The four concentrations were varied over the ranges 4-21% CO, 5-25% CO 2 , 11-34% H 2 O, and 14-55% H 2 .To determine the apparent activation energy, the temperature was varied over a range of 30 1C with the catalysts exposed to the standard gas concentrations.

Diffuse reflectance FTIR
Infrared spectra were obtained with a Nicolet Magna 550 FTIR augmented with a Thermo Spectra Tech Collector II diffuse reflectance (DR) mode attachment and equipped with a high-temperature, high-pressure environmental chamber.All DR spectra were collected in situ at a total flow rate of 50 mL min À1 with approximately 10 mg of finely ground sample catalyst.The samples were each exposed to 2 different gas compositions: 6.8% CO and WGS reaction mixture of 6.8% CO, 37.3% CO, 4% H 2 O and 8.6% CO 2 .The carrier gas, Ar, was fed through a Matheson PUR-Gas Triple Purifier Cartridge to remove trace O 2 , hydrocarbons and moisture.DR spectra were collected at 3 different temperatures always in order of decreasing temperature: 200, 120 1C and RT for Au/Al 2 O 3 and 280, 120 1C and RT for the Pt samples.The maximum temperatures of 200 1C and 280 1C were chosen to correlate with the temperature used to collect laboratory kinetic data for Au and the two Pt samples, respectively.At each temperature, ten minutes were allowed for equilibration.Before each catalyst was exposed to the adsorbates it underwent surface cleaning by oxidation with 10% O 2 for 10 min at 200 1C and 300 1C for the Au and Pt samples, respectively, followed by reduction in 25% H 2 for 30 min.Background spectra (256 scans) were collected in Ar at each adsorption temperature as the catalyst was cooled to RT from the reduction temperature.Then the catalyst was heated to the highest adsorption temperature, the adsorbing gases were introduced and a DR spectrum was collected with respect to the Ar background.The catalyst was cooled to the next adsorption temperature without changing the gas-phase conditions and a new spectrum was collected.Background spectra for WGS reaction mixture were obtained while flowing Ar through a bubble saturator with H 2 O at RT. Collecting backgrounds with H 2 O allows for subtraction of gas phase H 2 O bands.Spectra were averaged over 32 scans with 4 cm À1 resolution.
Peak fitting and data analysis were completed in CasaXPS v. 2.3.14 slightly modified for IR data compatibility.As adsorbed hydrogen is not visible in IR, and as adsorbed H 2 O is not distinguishable from gas phase H 2 O, IR characterization was limited to adsorbed CO.CO has been extensively used as a catalyst probe molecule, including during in situ WGS. 5,234][25][26][27][28][29][30][31][32] The spectra were background subtracted using a straight-line background from 2144 cm À1 (the valley between the P and R branches of the gas phase CO bands) to approximately 1700 cm À1 .As the gas phase CO bands overlap chemisorbed CO, a model peak for the P branch of gas phase CO was created from a spectrum of metal-free Al 2 O 3 support under WGS conditions.The model peak was highly constrained in shape (FWHM) and position allowing for fits of the very small peaks (linearly adsorbed CO on Au), which would otherwise be obscured by the significantly larger gas phase CO spectrum.For clarity, the CO gas peak was subtracted from the catalyst spectra.All fits assumed Gaussian peaks in positions known for CO adsorbed on Pt or Au.The fits were optimized via a Levenberg-Marquardt algorithm in the CasaXPS software.

Transmission electron microscopy
The Pt catalyst samples were dispersed in ethanol and sonicated for 10 min, dispersed on 200 mesh carbon-coated copper grids, and dried for 15 min at room temperature.The Z-contrast imaging was done by using an electron microscope (JEM-2010F FasTEMm FEI) manufactured by JEOL, USA operated at 200 kV and an extracting voltage of 4500 V. Since most of the particles were circular in shape, an electronic grid was place around the nanoparticle image in order to determine the diameter.The particle size distribution was obtained by measuring the diameter of approximately 1000 particles.

Catalyst characterization
It was previously reported that adsorption of Pt(NH 3 ) 4 (NO 3 ) 2 on SiO 2 at strongly basic pH followed by calcination at 225 1C gives 1-2 nm metallic particles with Pt loadings up to 2%. 33or the higher loading of this Pt/SiO 2 catalyst, similarly small sizes were obtained.Impregnation of PtTA on Al 2 O 3 followed by calcination at 450 1C also gives small metallic nanoparticles upon reduction.Generally, Al 2 O 3 -supported catalysts are less prone to sintering than those supported on SiO 2 .Assuming a H-to-surface Pt atomic ratio of 1, the dispersions estimated from hydrogen chemisorption experiments of both catalysts are near 1.0, i.e., every atom is at the surface of the particles.
Dark-field STEM images are shown in Fig. 1(a) and (b) for Pt/Al 2 O 3 and Pt/SiO 2 , respectively.For Pt/SiO 2 , the particle size distribution (not shown) indicates an average size of 1.7 nm with about 5% of the particles larger than 2.5 nm.Similarly, the average particle size on Pt/Al 2 O 3 was 1.5 nm with less than 5% of particles larger than 2 nm.
Additional information on the Pt particle sizes was determined by EXAFS spectroscopy at room temperature after the catalysts were reduced in 4% H 2 /He at 300 1C for 30 min.The isolated first-shell EXAFS spectra were obtained by a Fourier transform of the k 2 -weighted data from 2.75 to 12.2 A ˚À1 , followed by an inverse Fourier transform from 1.6 to 3.2 A ˚.The fit parameters were determined by fitting both the real and imaginary parts of the Fourier transform of the isolated k 2 -weighted EXAFS spectra and are summarized in Table 1.Assuming spherical nanoparticles, 34,35 the average sizes, determined based on previous correlations of the coordination number, N Pt-Pt , with dispersion, 36 give estimates of 1.5 nm for both Pt catalysts, in agreement with STEM and hydrogen chemisorption size estimates.
For these small Pt nanoparticles without adsorbates, there is a contraction in the Pt bond distances ( The 1% Au/Al 2 O 3 catalyst was a commercial sample provided by the World Gold Council.Au is well known to adsorb little H 2 or CO, ruling out the possibility of using chemisorption to estimate the dispersion and particle size.The Au particle size distribution by TEM was unsuccessful due to the poor contrast between the Au nanoparticles and alumina support.Therefore, the particle size was determined by EXAFS spectroscopy.As shown in Table 1, an estimate from N Au-Au also gives an average size of about 1.5 nm.In addition, the contraction of the Au bond distance to 2.77 A ˚is consistent with particles of this size. 36Although there is a small increase of about 0.02 and 0.03 A ˚, upon adsorption of H 2 and CO, respectively, the change is much smaller than that on Pt and the Au-Au bond distance is still significantly shorter than in Au foil, e.g., 2.88 A ˚.

WGS reaction kinetics
Results of the kinetic measurements are summarized in  been shown to depend strongly on the particle size. 40,41onsequently, it is difficult to make a direct comparison between our measured rates and those of others.3][44] Small variations are based on factors such as the concentration of feed gases, temperature, or composition of catalysts.For example, in the study by Grenoble et al., 43 the CO and H 2 O reaction orders were À0.21 and 0.75 for platinum and 0.74 and 0.13 for Au/Al 2 O 3 .However, in that study, hydrogen inhibition of the reaction was not taken into account.
The H 2 O and CO 2 reaction orders are very similar for the catalysts reported here.These catalysts, however, do display a variation in the H 2 reaction order, especially for the commercial Cu catalyst.The largest difference between Au and Pt is the CO reaction order.In general, Au catalysts have a CO reaction order near unity, while those for Pt are closer to zero.

XANES spectra of adsorbates on Pt WGS catalysts
Typically, the L 3 XANES spectra are used to determine the oxidation state, or fraction of metallic and oxidized Pt.Smaller changes in L 3 edge XANES spectra, however, also occur with chemisorption of gases, e.g., H 2 and CO. 15,16,45 Thus, such changes induced by reactant and product chemisorption can be used to determine adsorbate identity.
In addition, the change in Pt L 3 XANES intensity was shown to be linearly dependent on the amount of adsorbed H 2 . 18hus, the coverage of gases under WGS reaction conditions can also be determined.

CO adsorption.
Adsorption of CO on Pt leads to significant changes in the position, intensity and shape of the Pt L 3 XANES spectrum.Fig. 2(a) shows a comparison of the L 3 edge XANES of 4.3% Pt/SiO 2 in He at 280 1C, and in 6.8% CO from room temperature to 280 1C.Upon adsorption there is a shift in the edge position to higher energy.In addition, there is an increase in intensity up to about 20 eV above the edge.With increasing temperature up to 280 1C, there is only a small decrease in intensity, suggesting little CO desorbs at high temperature.
If one subtracts the XANES in He from that with adsorbed CO, the difference (or DXANES) shows how the shape and intensity of the edge changes with adsorption of CO.Fig. 2(b) shows the DXANES spectra of CO adsorption at room temperature, 120 1C, 200 1C, and 280 1C for Pt/SiO 2 .A very similar series of DXANES (not shown) are observed for Pt/Al 2 O 3 .It can be seen that the shape of the DXANES is very similar for all temperatures and the spectra differ primarily in intensity.As the temperature increases, the magnitude of the DXANES decreases, corresponding to less adsorbed CO.Using the DXANES spectrum at RT as the reference, the relative fraction of CO at each temperature can be determined.The results of the DXANES fits are given in Table 3.Similar results for single adsorbates are obtained from the usual linear combination fit using the XANES spectra with He and RT CO.Since CO saturates the surface of Pt nanoparticles at RT, the relative coverage also corresponds to the fractional surface coverage.Upon heating to 280 1C, the surface CO coverage is about 70% of that at RT. Very similar CO coverages are obtained on Pt/Al 2 O 3 .3.3.2H 2 adsorption.[45][46][47][48][49]  Using the DXANES spectrum at RT as the reference, the fraction of H 2 at each temperature can be calculated. 15The fitting results from the DXANES spectra are given in Table 3.At 280 1C, the relative coverage decreases to about 27% on Pt/SiO 2 and to 18% on Pt/Al 2 O 3 .Since at room temperature each surface Pt has one adsorbed H atom, the fractional fits at high temperature correspond to the H 2 surface coverage.Compared to CO, H 2 desorbs more easily, which is in agreement with the lower heat of adsorption.

XANES spectra of adsorbates on 1.5 nm Au WGS catalysts
Consistent with the lower adsorption capacity of Au compared to Pt, the changes in XANES spectra with adsorbed gases are much smaller. 50,51Because of these small changes, careful calibration of the reference foil and energy correction of the data files are required to avoid small artifacts in the DXANES spectra.
3.4.1 CO adsorption.Fig. 5(a) shows the Au L 3 edge XANES spectra of Au/Al 2 O 3 in He at 120 1C, in 6.8% CO at room temperature, in 6.8% CO followed by He purge at room temperature, and in 6.8% CO at 120 1C.Compared to adsorption on metallic Pt nanoparticles, the changes due to CO adsorption are very small and there is no shift in the edge position. 50Fig.5(b) shows the DXANES spectra on Au/Al 2 O 3 at room temperature in 6.8% CO, room temperature with He purge (0% gas phase CO), and at 120 1C in 6.8% CO.In the DXANES it is evident that the changes in the XANES shape occur above the edge, i.e., at higher than 11.92 keV.With increasing temperature, there is less adsorbed CO as expected.In addition, removal of gas-phase CO by purging with He also leads to a decrease in adsorbed CO indicating that even at room temperature, some CO is weakly bound.
The fits of relative CO coverage (compared to that at RT) from the DXANES adsorption are given in Table 4. Very similar fits were obtained by using a linear combination of Au with out adsorbate (He only) and that with adsorbate at RT.At RT, approximately 40% of the adsorbed CO desorbs with a He purge, and at 120 1C, the relative CO coverage (in flowing CO) is about 35% of that at RT.At 200 1C (spectra not shown), there is about 5% adsorbed CO on these Au nanoparticles.

H 2 and H 2 O adsorption.
The Au L 3 XANES spectra of Au/Al 2 O 3 in He at 120 1C, in 20% H 2 at room temperature, and in 20% H 2 at 120 1C are given in Fig. 6(a).Similar to the adsorption of CO, the changes in the XANES spectra upon H 2 adsorption are very small, and there is no shift of the edge position. 51The DXANES spectra for adsorbed H 2 at room temperature and at 120 1C are shown in Fig. 6(b).The changes in the XANES spectra also occur above the edge.The fits of the relative coverages (compared to that at RT) are given in Table 4. Similar to the adsorption of CO, H 2 binds weakly to Au with a relative coverage of about 50% at 120 1C and 15% at 200 1C.Adsorption of H 2 O at RT and 120 1C on Au/Al 2 O 3 also gives small changes in the XANES spectra [Fig.7(a)].The DXANES spectra are shown in Fig. 7(b) and the relative coverages are given in Table 4.The relative surface coverage decreases rapidly with increasing temperature, and little is adsorbed at 200 1C.Although the changes in Au XANES with adsorbates are much smaller and indicate weaker adsorption energies than on metallic Pt, the XANES shows that under WGS reaction temperatures, CO, H 2 and H 2 O all chemisorb on the Au nanoparticle surface.In addition, despite the very low adsorbate coverage on Au at 200 1C, the WGS TOR is significantly higher than that on Pt, e.g., Table 2.

FTIR of adsorbed CO
Since the infrared spectra of adsorbed CO are very similar for Pt/Al 2 O 3 and Pt/SiO 2 , 23 only the results for Pt/SiO 2 are presented here.Fig. 8 shows the diffuse reflectance IR spectra over a range from 2145 to 1700 cm À1 for 6.8% CO on 4.3% Pt/SiO 2 at different temperatures.At RT, the data shows three clear peaks, a sharp shoulder at 2088 cm À1 , a strong peak at 2070 cm À1 with an asymmetric tail and a broad low intensity peak at 1763 cm À1 .The largest peak at 2070 cm À1 is designated linear-bonded CO, while the asymmetric tail likely represents a distribution of sites that bond CO in a bridgedbonded conformation.While bridged CO is generally expected between 1900-1800 cm À1 , Sheppard and Nguyen propose that bridged-bonded CO species occur in the region from about 2000-1800 cm À1 . 52The low-frequency peak is assigned as three-fold bonded CO. [23][24][25][30][31][32]53 The sharp shoulder is still under debate in the literature and is thought to be due to CO adsorbed in small islands or arrays, 31 or CO adsorbed on Pt d+ . 29,30 Astemperature increases to 280 1C, the sharp shoulder shifts to 2078 cm À1 , the linear peak shifts to 2060 cm À1 , the bridged species becomes more prominent and the three-fold bonded CO peak shifts to 1762 cm À1 .It is well known that these shifts are due to a decrease in CO surface coverage leading to less dipole-dipole coupling and a shift to lower frequency.28,30 Under WGS conditions at 280 1C, Fig. 8b, there is a small decrease in the linear-bonded CO and a small shifts in the sharp shoulder, and intensity of the three-fold CO peak.The CO coverage measured by FTIR does not change significantly with changing CO or under WGS (CO + H 2 O + H 2 ) reaction mixture.
Fig. 9(a) shows the DRIFTS spectra of CO on Au/Al 2 O 3 from 2145 to 1950 cm À1 at different temperatures, with the CO gas-phase peak subtracted from the catalyst spectra.Since the peak intensities of CO on Au/Al 2 O 3 are a factor of 10 smaller than those on Pt/SiO 2 , the gas-phase CO bands obscure the peaks of CO adsorbed on the Au nanoparticles.The subtraction process was only used in presentation of the data, not the true data analysis.The difference spectra, however, allow for better determination of the peak shape, position and intensity.The linear CO adsorption peak is located at 2097 cm À1 is similar to previous studies on metallic Au 26,27,54 and the position does not change with temperature.Fig. 9(b) shows the DRIFTS of adsorbed CO on Au under WGS reaction conditions.At both 200 and 120 1C the peak has red-shifted to 2094 cm À1 , while at RT the peak has blue-shifted to 2100 cm À1 .The blue-shift may result from competitive adsorption of H 2 O on the metal as discussed below.At 200 1C the peak areas with 6.8% CO only and during WGS reaction are similarly small each showing a coverage of about 5% that for CO only at RT, while at 120 1C the relative coverage changes to 50% and 40% for CO only and under WGS reaction conditions, respectively.For the WGS gas mixture at RT, the relative coverage is also roughly 40%, a significant decrease from the higher CO-only coverage.The relative coverages determined by IR spectroscopy are very similar to those determined by XANES and DXANES fits in Table 4.

Structure of the Pt and Au WGS catalysts
It has been previously reported that the electronic structures of the atoms in metal nanoparticles differ from the bulk due to the rehybridization of the spd orbitals. 55The rehybridization results in an increase in the local electron density between metal atoms, which in turn leads to a higher bond order and subsequently a contraction of the metal-metal bond distances. 15,36,51,55,56In the present study, EXAFS experiments reveal that the reduced 1.6 nm supported Pt catalysts have a Pt-Pt bond distance of 2.69 A ˚(0.08 A ˚shorter than bulk), and the 1.4 nm supported Au catalyst has a Au-Au bond distance of 2.75 A ˚(0.13 A ˚shorter than bulk).This contraction, however, occurs only in the absence of adsorbates.Upon H 2 or CO adsorption, the bond distance increases, as shown in Tables 1 and 3.
From fitting the XANES (or DXANES) with adsorbed hydrogen at different temperatures (RT, 120, 200 and 280 1C), the surface coverage was determined (Table 3).Over this temperature range, the EXAFS indicates there is little change in N Pt-Pt , or particle size.While the size does not change, the  Pt bond distance increases with decreasing temperature.Fig. 10 shows that the Pt-Pt bond distance is linearly correlated with the hydrogen coverage.In addition, extrapolation to zero coverage gives an adsorbate-free bond distance of 2.69 A ˚, which is the measured bond distance in He at RT and 280 1C (Table 1).It is thought that when H 2 is adsorbed on Pt nanoparticles, the Pt-Pt valence electron density is shared by the Pt-H bond making the Pt-Pt bonds weaker and the bond distance longer. 15,51,56n Table 3, it can be seen that the Pt-Pt bond distance also increases upon adsorption of CO.Because CO has a higher heat of adsorption, the surface coverage is higher than that of   H 2 at the same temperature.At the same surface coverage, however, the contraction in the Pt-Pt bond distance is similar for CO and H 2 .For example, at 70% surface coverage (120 1C for H 2 and 280 1C for CO) on Pt/SiO 2 , the Pt-Pt bond distance is 2.73 A ˚for both adsorbates.At 280 1C and under WGS reaction conditions, the Pt-Pt bond distance is larger than that in He, but is shorter than that of Pt foil.Thus, the Pt-Pt bond distance in small Pt nanoparticles is dependent on the fractional coverage of adsorbates, at least for CO and H 2 .
On Au nanoparticles the surface coverage of adsorbates is low.Thus, there is little change in the metallic bond distance in the presence or absence of adsorbates.The metallic bond distance is similar in He, H 2 , CO and under WGS reaction and is significantly shorter than that of Au foil.
The EXAFS of both Pt and Au indicate that the active phase is metallic, and for each adsorbate or under WGS reaction condition, there is no indication of oxidized metal as has been suggested for Au on CeO 2 . 11,132 The effect of adsorbates on the L 3 XANES spectra Pt and Au L 3 XANES correspond to the dipole allowed 2p -5d electronic transition, and is most often used to determine the oxidation state of the metal.As shown in this and other studies, [15][16][17][18][45][46][47][48][49][50][51] the intensity and position of the L 3 XANES spectra are affected by the surface coverage of adsorbates.Although one can determine the metal oxidation states from the K-edge XANES, similar changes in the XANES shape and intensity induced by adsorbates are not observed.57 For Pt, the changes in XANES position and intensity are sufficiently different to be used to identify the type of adsorbates.As shown in Fig. 11, the DXANES spectrum of H 2 O adsorption at 280 1C for Pt/SiO 2 has a shape distinct from that generated by CO and H 2 adsorption.In an effort to understand these differences in shape, we have performed calculations using the CASTEP code to simulate the XANES spectra.58 Briefly, the CASTEP simulation calculates the XANES (or EELS) spectra from matrix elements defined by an initial core state using a one-atom all electron calculation and a final state based upon an ''on-the-fly'' pseudopotential calculation.A 1Â1 unit cell was utilized in order to reduce computational effort.Calculations were performed without the inclusion of a core hole due to the electron shielding of platinum.The calculated spectra are all shifted in energy by a constant value (11 564 eV) so that the calculated edge onset is consistent with experimental  results.Gaussian broadening of 3.0 eV is applied to mimic the instrument broadening.We approximate the surface coverage by assuming one monolayer of CO in the 3-fold hollow sites of Pt(111) (this is the favored site using the PW-91 functional due to the well known over-coordination problem of GGA functionals 59 ).Since molecular water does not adsorb strongly on metal surfaces and has a calculated adsorption energy of 0.30 eV, 60 we model the effect of water adsorption as one monolayer of OH at atop sites on the Pt(111). In both cass, we have chosen coverages that are substantially higher than the experimentally determined values.However, our models, although not quantitative, will allow us to assess the effects of water and CO adsorption from a qualitative standpoint allowing for identification of trends.Fig. 12(a) shows the simulated Pt L edge XANES (the calculations do not include spin-orbit coupling so we cannot differentiate between the L 2 and L 3 edges) for a platinum atom at the surface of Pt(111) as well as Pt with a monolayer of adsorbed CO and OH.The absorption spectrum shifts by 0.6 eV to lower energies with adsorbed OH and by 1.3 eV to higher energy with adsorbed CO.Fig. 12(b) shows the simulated DXANES spectra for CO and OH.It is clear that the simulated spectra replicate many of the features of experimental spectra, showing the positive peak for OH and the negative peak for CO associated with the edge shift.Recent work from Schweitzer et al. 61 suggests that the position of the edge is directly related to the density of states at the Fermi level (and not due to the charge transfer effects).Fig. 13 shows the partial d-electron density of states for surface atoms of Pt with and without OH, CO adsorbates.Upon adsorption of OH and CO, two effects are observed.First, the d-electron density is significantly depleted in the presence of an adsorbate.Second, the center of the d-band is shifted away from the Fermi level in the presence of the adsorbate.The d-band center shifts from À2.64 eV for Pt(111) to À4.17 eV for CO/Pt(111) and to À3.39 eV for OH/Pt(111). Hover, important differences exist as well.While the d-electron density is strongly depleted near the Fermi edge when CO is adsorbed, there is still significant density near the edge when OH is adsorbed.This implies that d-orbitals of different symmetry are involved in the bonding of CO versus OH to the Pt(111) surface.62   In addition, in Fig. 12(a) there is a sharp increase in the XANES intensity above the edge when either CO or OH is adsorbed.We hypothesize that the increase in intensity is related to the presence of empty (presumably antibonding) states above the Fermi level that are created from hybridization of d-states of the metal with adsorbate states.As proposed by Hoffmann, 63 both bonding and antibonding states will be created due to orbital overlap between filled states of the metal and both filled and empty states of the adsorbate.These hybrid states are now allowed transitions since they possess some d-symmetry.

Changes in the XANES spectra during WGS reaction
The Pt L 3 XANES spectra at 280 1C under WGS reaction and CO are given in Fig. 14(a).The similarity of these indicates that the major surface species under WGS reaction is CO at about the same coverage as CO only.By subtracting the XANES spectrum under CO from that under WGS conditions, a small residual XANES feature is obtained and is shown in Fig. 14(b).Although this peak is very small, it has the same shape as the DXANES of H 2 O. Using the DXANES of H 2 O at 280 1C as the reference, the fit indicates that there is approximately 10% as much adsorbed H 2 O under WGS as for H 2 O only.The absolute coverage is unknown since it was not possible to quantitatively determine the amount of adsorbed H 2 O on the Pt nanoparticles at 280 1C.Much of the reduced surface coverage is due to the high CO coverage, which is about 70%, essentially the same as for CO alone.
While the DXANES on Pt are sufficiently different to identify the adsorbed species in a mixture, the changes on Au are much smaller, and the position and shape are more similar.As a result, identification of the adsorbed species is less reliable.Fig. 15 shows the DXANES for adsorbed CO, H 2 and H 2 O at RT on Au/Al 2 O 3 .While there are subtle differences in shape between CO and H 2 or H 2 O, the DXANES spectra of H 2 and H 2 O are identical.Fig. 16 compares the DXANES of the WGS gases and that of CO only at RT.The small difference indicates a second contribution.A fit of the WGS DXANES indicates that there is about 65% relative CO coverage and 35% relative H 2 O (or H 2 ) coverage, Table 4.The relative CO and H 2 O coverages decrease at 120 1C and there is only a trace amount of adsorbed CO at 200 1C.Small amounts of H 2 O (or H 2 ) are likely present at 200 1C, but are too small to detect.Despite the low surface coverage of reactants on Au, this catalyst has a significantly higher TOR than Pt.

Comparison of the amount of adsorbed CO by DXANES and FTIR
FTIR can be used to quantify CO coverage on supported catalysts where the integrated absorption coefficient (IAC) has been determined.However, the application of IACs is not without difficulties.Hollins and Pritchard 28 have summarized some of the corrections that are required in order to be quantitative.For example, there is a transfer of peak intensity from low frequency to higher frequency, a blue-shift due to increased dipole coupling as coverage increases, and an overall intensity loss at high surface coverages. 64As a result, significant errors in estimates of the amount of adsorbed CO can occur at high coverage.For our samples, the intensity loss at high  coverage is especially significant on Pt/Al 2 O 3 , but occurs to a lesser extent on Pt/SiO 2 .While this makes quantification difficult, these effects allow for some qualitative conclusions.For example, the dipole shift to higher wavenumber at lower adsorption temperature in Fig. 8 is an indication of an increase in CO coverage, despite the slight decrease in integrated area that is observed by us and others in the literature. 30n the Au sample, coverage-dependent effects are not observed due to the low coverage at every temperature and integrated areas can be used for quantification of the relative surface coverages.Using FTIR CO coverage under 6.8% CO as a reference, we compute the Au CO coverages at different temperature and gas mixtures to be very similar to those determined by XANES (or DXANES).At 200 1C, on Au there is agreement by IR, XANES and chemisorption that the CO coverage is low.Also, at lower temperatures there is competitive adsorption of CO with H 2 O, i.e., the CO coverage is lower with the WGS gas mixture than for CO only.The competitive adsorption of CO with H 2 O at low temperature is very different on Au versus Pt.
Although the amounts of adsorbed CO are similar by IR and XANES, DRIFTS affords the additional advantage of distinguishing between different CO adsorption conformations, and hence different CO adsorption sites.The advantage of the DXANES analysis is that one can identify and quantify IR inactive adsorbates (such as H 2 ), or highly IR absorbing molecules (such as H 2 O).Furthermore, the changes in the XANES imply bond formation at the catalytic site.Although the WGS reaction conditions are strongly reducing, i.e., CO and H 2 , the active site for the WGS reaction has been suggested to be ionic Au. 11,13 While there is no evidence for Au oxide or ionic Au in the EXAFS of this catalyst, the increase in XANES intensity with adsorbates might reflect oxidation of a small number of metallic Au atoms under reaction conditions.Comparison of the DXANES for the catalyst oxidized in air and with adsorbed H 2 O at RT is shown in Fig. 17      and is nearly completely desorbed at 200 1C.By contrast, the oxygen coverage on ionic Au does not change with temperature except at temperatures greater than about 350 1C. 65Thus, the shape of the XANES spectra and differences in chemical properties indicate that the changes in the Au XANES are due to adsorbate-metal bond formation, rather than oxidation of the metal nanoparticle surface under WGS conditions. 14ig. 18 shows a similar comparison of the DXANES of oxidized Pt and with adsorbed H 2 O.The oxidized Pt DXANES was obtained by air oxidation of a 9 nm Pt nanoparticle supported on SBA-15 at RT. Oxidation leads to surface PtO, i.e., Pt 2+ , with a metallic core.Similar to the effect of H 2 O on Au, the change in the DXANES intensity for H 2 O on Pt increases with increasing partial pressure and decreases with increasing temperature.By contrast the amount of oxidized Pt does not reversibly change with O 2 partial pressure and cannot be desorbed in He until temperatures greater than about 500 1C. 33he conclusion from both EXAFS and XANES under operando conditions, i.e., high reaction rate, is that metallic Au and Pt are the active sites for WGS in these catalysts.

4.5.2
The influence of adsorption on the observed WGS kinetics.The results from this study also indicate possible reasons why Au has a higher apparent TOR at low reaction temperature, higher CO reaction order and, generally, operates at a lower temperature than Pt.Under WGS reaction conditions, the CO coverage on Pt is high, about 70% at 280 1C.Increasing the partial pressure of CO has a minimal effect on increasing the surface coverage and rate.Thus, the reaction order is near zero.Similarly, as one lowers the reaction temperature on Pt, the CO surface coverage increases to near saturation at room temperature.At low reaction temperatures, therefore, there are few sites to adsorb H 2 O.In order to create exposed sites for H 2 O adsorption, higher reaction temperatures are required.Finally, since the CO surface coverage is high, especially at low temperature, increasing the CO partial pressure would most likely lead to a decrease in rate by complete saturation of the Pt surface.It is suggested that strong CO adsorption and high surface coverage leads to few exposed catalytic sites and contributes to inhibition of the TOR at low temperature, low reaction order and higher reaction temperatures on Pt.
On Au, the heat of CO adsorption is much lower, leading to significantly lower surface coverage, more exposed catalytic sites and a CO reaction order close to 1. Since the Au surface is not saturated, catalytic activity is possible at lower temperatures than for Pt.In addition, increasing CO partial pressure would increase the surface coverage and reaction rate.At higher pressure, it may even be possible to conduct the WGS reaction at a lower temperature than was done here (1 atm and 200 1C).
There is a continual need to develop WGS catalysts with higher rates per unit volume and which operate at lower temperature where thermodynamics favors H 2 production and lower levels of CO.The results from this study suggest that adsorption is an important factor in determination of the apparent TOR, reaction order of CO and temperature of the reaction.It is expected that improved catalysts will be those having lower heats of CO adsorption than Pt.To be active, the next generation of catalysts will have partial CO coverages at low, or even room temperature.

Conclusions
Low-temperature Pt and Au WGS catalysts have been investigated by EXAFS and XANES spectroscopy in order to identify the structure of the active site, and origin of high activity.Using a plug-flow, kinetic-EXAFS reactor, the EXAFS, XANES and kinetic measurements were simultaneously measured.The EXAFS of both Pt and Au indicate the active phase is fully reduced, metallic nanoparticles.At the L 3 edge, the surface-adsorbed species alter the position and shape of the XANES spectra.For Pt, subtraction of the adsorption edge free of adsorbate from that with adsorbate gives a DXANES spectrum, which has a unique shape for each adsorbate.Thus, the Pt DXANES can be fit to identify the type and coverage of adsorbed species under reaction conditions.During WGS reaction at 280 1C, the CO surface coverage on Pt is approximately 70% similar to that of CO only.While the primary surface species is CO, a small amount of adsorbed H 2 O is also observed.On Au, the changes in the XANES spectra are much smaller, consistent with low coverage of reactants and products.At low temperatures, CO and H 2 O are adsorbed, but under reaction temperatures near 200 1C, there is little adsorbed CO, H 2 or H 2 O. Nevertheless, the Au TOR is significantly higher than that for Pt.The high TOR of Au is suggested to result from the weak adsorption of CO and the availability of free catalytic sites, even at low temperature, while on Pt, high CO surface coverage leaves few exposed active sites resulting in low rates.The implication is that enthalpy of CO adsorption is an important factor in development of low-temperature WGS activity.
Fig. 3(a) shows a comparison of the Pt L 3 XANES spectra of 4.3% Pt/SiO 2 in He at 280 1C, in 20% H 2 at room temperature, 120 1C, and 280 1C.Fig. 3(b) shows the corresponding DXANES spectra for Pt/SiO 2 .Similar changes occur on Pt/Al 2 O 3 with H 2 adsorption (not shown).As with CO, as the temperature increases, the magnitude of the DXANES spectra decreases.

Fig. 10
Fig. 10 Dependence of Pt-Pt bond distance on H 2 coverage; the surface coverage was determined for 20% flowing H 2 at RT, 120, 200 and 280 1C.

4. 5
Implications for WGS catalysts 4.5.1 Identification of the active site in Pt and Au WGS catalysts.
and indicates that the shape and position of oxidized Au and with adsorbed H 2 O are substantially different.In addition, as shown in
2.69 and 2.68 A ˚for Pt/Al 2 O 3 and Pt/SiO 2 , respectively) compared to Pt foil (2.77A ˚).This shortened bond distance, however, increases upon adsorption of 4% H 2 (2.69 to 2.73 A ˚on Pt/Al 2 O 3 and 2.68 to 2.75 A ˚on Pt/SiO 2 ).Similarly, there is a slightly larger increase in Pt bond distance for 7% CO (2.75A ˚for Pt/Al 2 O 3 and Pt/SiO 2 ).

Table 2
[37][38][39]ion energy of approximately 10 kJ mol À1 was observed for the Au/Al 2 O 3 catalyst.Although this result is considerably lower than that for the Pt WGS catalysts, e.g., 70-80 kJ mol À1 , it is typical of previously reported activation energies for Au/Al 2 O 3 .[37][38][39]Therate for Au WGS catalysts has Here we report the turn-over-rate (TOR) of Au/Al 2 O 3 and those on Pt/Al 2 O 3 and Pt/SiO 2 with very similar size (and that of a commercial Cu catalysts with much larger size).The TOR is calculated as the moles of CO reacted per mole of surface metal (Pt, Au, or Cu) per second.At 200 1C the Au/Al 2 O 3 has a TOR, which is about 20 times that of Pt/Al 2 O 3 and has a similar TOR to that of a commercial Cu/ZnO/Al 2 O 3 catalyst.In addition, Pt/SiO 2 has a TOR, which is about six times higher than Pt/Al 2 O 3 , which indicates the well-known support effect for the WGS reaction.

Table 1
EXAFS fit parameters for the reduced supported Pt and Au catalysts under different treatment conditions at room temperature

Table 2
Summary of catalytic performances of the supported Pt and Au catalysts and a commercial CuZn catalyst a Temperature at which the reaction orders were determined.b Dispersion measured from EXAFS.c Dispersion measured by CO chemisorption.d CO chemisorption after N 2 O oxidation. 66e Rates calculated at 200 1C, 7% CO, 22% H 2 O, 8.5% CO 2 , 37% H 2 .
3.3.3H 2 O adsorption.Compared to CO and H 2 , adsorption of H 2 O gives smaller changes in the Pt L 3 XANES.Fig. 4(a) compares the XANES spectra of Pt/SiO 2 in He at 200 1C and with adsorbed H 2 O at two H 2 O vapor pressures (also at 200 1C).Since H 2 O condenses and adsorbs on the supports, H 2 O adsorption was conducted at elevated temperature.In addition to smaller changes in intensity, there is a slight shift to lower energy upon H 2 O adsorption.Fig. 4(b) shows the L 3 DXANES of H 2 O adsorption at different H 2 O vapor pressures for Pt/SiO 2 .Consistent with the higher H 2 O vapor, there is an increase in the DXANES intensity.Since Pt L 3 XANES intensity and position are dependent on the d orbital occupancy, these changes indicate that H 2 O is chemically bonded, i.e. transfer of some electron density between Pt and H 2 O, at this temperature.Since it was not possible to quantify the amount of chemisorbed H 2 O on the Pt nanoparticles separately from that physisorbed on the support, the relative coverage (given in parenthesis in Table 3) at 280 1C (relative to that at 280 1C in the absence of other WGS adsorbates) was determined for comparison with H 2 O coverage under WGS reaction (discussed below).

Table 3
Temperature-dependent EXAFS and XANES fit parameters for Pt WGS catalysts with adsorbed gases

Table 4
Temperature-dependent EXAFS and XANES fit parameters for the Au/Al 2 O 3 catalyst with adsorbed WGS gases

Table 4 ,
the amount of adsorbed H 2 O decreases with increasing temperature