Counting Au catalytic sites for the water-gas shift reaction

Transcription

1 Purdue University Purdue e-pubs Birck and NCN Publications Birck Nanotechnology Center Counting Au catalytic sites for the water-gas shift reaction Mayank Shekhar Purdue University, Jun Wang Purdue University Wen-Sheng Lee Birck Nanotechnology Center, Purdue University, M. Cem Akatay Birck Nanotechnology Center, Purdue University, Eric A. Stach Birck Nanotechnology Center, Purdue University; Brookhaven National Laboratory, See next page for additional authors Follow this and additional works at: Part of the Nanoscience and Nanotechnology Commons Shekhar, Mayank; Wang, Jun; Lee, Wen-Sheng; Akatay, M. Cem; Stach, Eric A.; Delgass, W. Nicholas; and Ribeiro, Fabio H., "Counting Au catalytic sites for the water-gas shift reaction" (2012). Birck and NCN Publications. Paper This document has been made available through Purdue e-pubs, a service of the Purdue University Libraries. Please contact epubs@purdue.edu for additional information.

2 Authors Mayank Shekhar, Jun Wang, Wen-Sheng Lee, M. Cem Akatay, Eric A. Stach, W. Nicholas Delgass, and Fabio H. Ribeiro This article is available at Purdue e-pubs:

3 Journal of Catalysis 293 (2012) Contents lists available at SciVerse ScienceDirect Journal of Catalysis journal homepage: Counting Au catalytic sites for the water gas shift reaction Mayank Shekhar a, Jun Wang a, Wen-Sheng Lee a, M. Cem Akatay b,c, Eric A. Stach b,c,d, W. Nicholas Delgass a, Fabio H. Ribeiro a, a School of Chemical Engineering, Purdue University, West Lafayette, IN 47907, USA b School of Materials Engineering, Purdue University, West Lafayette, IN 47907, USA c Birck Nanotechnology Center, West Lafayette, IN 47907, USA d Center for Functional Nanomaterials, Brookhaven National Laboratory, Upton, NY 11973, USA article info abstract Article history: Received 2 April 2012 Revised 6 June 2012 Accepted 9 June 2012 Available online 20 July 2012 Keywords: Water gas shift Gold Bromine poisoning WGS kinetics Transient isotopic switch Operando FTIR We have developed various techniques to count catalytic sites of Au/TiO 2 catalysts for the water gas shift (WGS) reaction. Addition of Br in an amount that is only 16% of the moles of the surface Au on a 2.3 wt.%au/tio 2 catalyst decreases the majority of its WGS reaction rate per total mole of Au but does not result in an appreciable change in the average Au particle size, Au particle shape, apparent activation energy, or reaction orders. From transient isotopic switch experiments, the WGS turnover frequency (TOF) for Au/TiO 2 catalysts with and without Br, based on the operating active sites counted in the experiment, is 1.6 ± 0.5 s 1 under 6.8% CO, 8.5% CO 2, 11.0% H 2 O, 37.4% H 2 at 120 C. The estimated number of potential active sites, 2% of the total amount of Au on the 2.3 wt.%au/tio 2 catalyst, best correlates with the Au corner atoms (2%) of the cubo-octrahedral particles. From operando FTIR spectroscopy, the normalized IR peak area of CO adsorbed on Au 0 near 2100 cm 1 is proportional to the WGS reaction rate for Au/TiO 2 catalysts with and without Br. Thus, the dominant active sites on Au/TiO 2 catalysts for the WGS reaction are taken to be the metallic corner Au sites with Au Au coordination number of 4. Ó 2012 Elsevier Inc. All rights reserved. 1. Introduction While bulk Au is often regarded to be chemically inert, supported Au nanoparticles catalyze a wide variety of reactions such as CO oxidation [1], water gas shift (WGS) [2], and selective and total oxidation of hydrocarbons [3,4]. Thus, the study of the source of catalytic activity of Au nanoparticles presents an opportunity to understand the unique characteristics possessed by metal nanoparticles. In the literature, the catalytic activity of supported Au nanoparticles is claimed to be due to cationic Au [5], bilayers of Au [6,7], perimeter sites [8], and low coordinated corner sites [2,9]. Fu et al. [5] found that the WGS rate for Au/CeO 2 catalysts was not affected by the removal of metallic Au particles by cyanide leaching and thus that metallic nanoparticles were mere spectators in the WGS reaction. They concluded that the nonmetallic Au species embedded in ceria catalyze the WGS reaction [5]. In the work by Herzing et al. [6], the CO oxidation catalytic rates for Au/FeO x catalysts correlated with the presence of bilayer clusters that are 0.5 nm in diameter and contain only 10 gold atoms. The dependence of the rate on the average Au particle size has been extensively used to determine the identity of active sites on Au/TiO 2 catalysts [2,7 9]. Valden et al. [7] found the CO oxidation rate per total mole of Au for Au/TiO 2 catalysts to be maximum at an Corresponding author. Address: School of Chemical Engineering, Purdue University, 480 Stadium Mall Drive, West Lafayette, IN , USA. address: fabio@purdue.edu (F.H. Ribeiro). average Au particle size of 3 nm. On the contrary, HD exchange [8], CO oxidation [9], and WGS reaction rates per total mole of Au for Au/TiO 2 catalysts were found to increase with decrease in average Au particle size. While the HD exchange rate correlated with the perimeter Au sites [8] as being active, the CO oxidation [9] and WGS [2] rates correlated with the low coordinated corner Au sites being active. Halide poisoning has been used as an effective tool to study the active sites for the CO oxidation reaction and CO adsorption on Au catalysts. Addition of chlorine [10], bromine [11,12], and fluorine [13] has been shown to suppress the catalytic activity of Au catalysts by either promoting agglomeration of Au nanoparticles or poisoning the active Au sites. In the work by Oxford et al. [11], it was shown that when 5 10% of the total moles of Au, or 10 20% of the surface moles of Au, in a Au/TiO 2 catalyst were bound to Br, the CO oxidation catalytic activity was completely blocked, although, at 60 C, 35% of the original CO adsorption capacity remained. The perimeter Au atoms at or near the particle-support interface were claimed to be the potential active sites in the oxidative environment of the CO oxidation reaction [11]. Here, we have used the WGS reaction as a model to study the origin of catalytic activity of supported Au nanoparticles. In this work, we have used poisoning by Br to confirm the identity of the catalytic sites on Au/TiO 2 catalysts. Gold nanoparticles supported on TiO 2 have been shown to have superior catalytic activity for the WGS reaction [2]. Model non-porous and crystalline rutile /$ - see front matter Ó 2012 Elsevier Inc. All rights reserved.

4 M. Shekhar et al. / Journal of Catalysis 293 (2012) TiO 2 support with a BET surface area of 28 m 2 g 1 was used to study the active sites of Au/TiO 2 catalysts for the WGS reaction. Since the support was non-porous, all the active Au nanoparticles were accessible on the surface of the support. Further, the support s crystallinity enhanced the contrast between the Au nanoparticles and the support in transmission electron microscopy (TEM) images, allowing precise determination of the Au particle size distribution and the Au particle shape. A 2.3%Au/TiO 2 catalyst, prepared using this model support, was poisoned by KBr to prepare 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts such that the total moles of Br present were 4% and 8% of the total moles of Au, respectively. A physical model of Au nanoparticles as truncated cubo-octahedra was used to determine the percentage Au that are surface sites, perimeter sites and corner sites from the entire Au nanoparticle size distribution determined from TEM. The Au content of these catalysts was determined to be 2.3 ± 0.2 wt.% by atomic absorption. Due to the strong dependence of the rate on the Au particle size of Au/TiO 2 catalysts, it was essential to compare the rates of the brominated Au/TiO 2 catalysts to Au/ TiO 2 catalysts at the same particle size. While poisoning by Br and transient isotopic switch experiments were used as tools to count Au catalytic sites, operando Fourier transform infrared (FTIR) spectroscopy was used to determine the nature of Au catalytic sites TEM The TEM images illustrate the imaging advantages of having high Z contrast between a metal and its non-porous support. The use of the non-porous support material ensures that no metal particles are hidden from view within a pore structure. Au particle size can change during exposure to WGS reaction conditions due to sintering. Thus, used samples (samples after the kinetic measurements) were imaged by TEM using an kv S/TEM FEI Titan operating at 300 kv. Prior to the TEM experiments, the catalyst samples were dispersed in ethanol and sonicated for 10 min. The suspensions were then dropped on 200 mesh lacey carbon coated copper grids. The grids were dried in air for 15 min at room temperature. The size of each gold cluster was determined from the longest measureable distance for that cluster. The particle size distributions of the used Au/TiO 2 and brominated Au/TiO 2 catalysts were calculated from the acquired TEM images. From the particle size distributions, the number (d), surface (d s ), and volume (d v ) average Au particle sizes were determined using the following equations: P i d ¼ d i n ; d s ¼ P i d3 i P i d2 i P i ; d v ¼ d4 i P i d3 i 2. Experimental methods 2.1. Catalyst preparation The Au/TiO 2 catalysts were prepared by the deposition precipitation (DP) method. HAuCl 4 3H 2 O was used as the Au precursor and was added to deionized water to give a M Au solution. A solution of 0.1 N NaOH was added dropwise to the Au solution so that the mixture maintained a ph = 6 at 35 C for approximately 6 h. The support material was then added to the solution, and the suspension was heated to 85 C at a rate of 1.7 C per minute and maintained at 85 C for 1 h. The mixture was then cooled, centrifuged, washed, and dried. The detailed procedure is discussed in our previous work [2]. Non-porous and crystalline, TEM friendly rutile TiO 2 support was used to ensure that all Au was deposited on the outside of the support, accessible to TEM analysis. The rutile TiO 2 support used was corporation lot number E from Sachtleben Chemie GmbH, Germany, and had a stable BET surface area of 28 m 2 per gram (after steaming at 500 C using a 30% water in air mixture for 48 h). One gram of 2.3%Au/TiO 2 catalyst (synthesized by the procedure stated above) was impregnated by the incipient wetness method. A KBr solution containing 4% and 8% Br of the total moles of Au was added dropwise to the solid catalyst while it was stirred to enhance the mixing in order to synthesize 2.3%Au (1Br:25Au)/ TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts, respectively. The impregnation was followed by drying the catalyst overnight under vacuum at room temperature. Since the brominated Au catalysts are prone to sintering with time at room temperature, these materials were immediately (in a period of 1 h) transferred to the tubular reactor unit for kinetic measurements. The Au loadings of the catalysts were determined by atomic absorption spectroscopy (AAS), performed on each sample with a Perkin Elmer AAnalyst 300 instrument. Prior to AAS measurements, the catalysts were digested (2 ml/1 ml/100 mg = aqua regia/hf/catalyst) in a Nalgene Ò amber high-density polyethylene bottle for at least 3 days, and this solution was then diluted to the desired concentration for the AAS measurement. The absorption results were compared to those of known standards to obtain the Au content. We note that the Br content was not measured. Thus, the nominal Br loadings represent upper bounds on the Br content. Here, d i is the Au particle diameter (defined as the longest distance measure for each particle), n is the total number of Au particles counted from the TEM images of a given sample, and the summation is performed over the entire particle population identified by the TEM images. These equations are also provided in the literature [14] Operando FTIR spectroscopy and transient isotopic switch experiments The FTIR spectrometer was a Bruker Vertex 70 FTIR and the transmission IR cell was a homemade reactor, the details of which are provided elsewhere [15,16]. About 70 mg of catalyst sample was pressed in the form of a thin wafer with diameter about 2 cm for the transmission IR study. The IR backgrounds of 100 scans were collected when the catalyst was exposed to 11% H 2 O, balance Ar, and He at desired temperatures. All spectra taken at WGS steady-state conditions were collected at a resolution of 4cm 1 and averaged over 50 scans. For isotopic transient experiments, the IR spectra were collected in the rapid scan, time-resolved mode with the scan rate about 8 spectra per second. For the operando studies, the CO, H 2, and Ar in He balance were bubbled through a H 2 O saturator heated to a temperature at which the vapor pressure gave the desired concentration (6.8% CO, 11.0% H 2 O, 8.5% CO 2, and 37.4% H 2 ). The concentrations of CO, CO 2, and H 2 were the same as for the kinetic measurements done in the tubular reactor unit in this study, and the total flow rate was 50 sccm. For the transient isotopic switch experiments, the details are provided elsewhere [16]. In short, CO + Ar in the reaction mixture was switched to 13 CO + Ne at the same flow rate and pressure, with Ar and Ne acting as tracers. An Agilent 5973 N mass spectrometer (MS) was used to track the gas phase changes during the isotopic transient experiments. To determine the error associated with these measurements, the procedure was repeated three times for each catalyst. The standard deviation calculated from the three measurements is the reported error. Fresh catalyst from the same batch used in the kinetic measurements was loaded into the IR reactor cell. It was pretreated with the same procedure as used in the kinetic measurements. The total flow rate of gases through the IR cell was kept constant at 50 sccm throughout the experiments. The IR spectra and concentration

5 96 M. Shekhar et al. / Journal of Catalysis 293 (2012) measurements by gas chromatography during the pretreatment and under steady-state reaction conditions were then collected. CasaXPS version was used for integration of the IR peak areas in the CO stretching region. Gaussian Lorentzian symmetric line-shape curves GL (30), that is, 70% Lorentzian and 30% Gaussian, were used. The peak position, area, and Full Width at Half Maximum (FWHM) were optimized by minimizing the root mean square (RMS) error through Levenberg Marquardt iterations in CasaXPS Kinetic measurements tubular reactor unit For each of the kinetic experiments, mg of catalyst was added to a reactor in our automated, four independent parallel tubular plug flow reactor setup, described elsewhere [17]. The 2.3%Au/TiO 2, 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/ TiO 2 catalysts were reduced in a 25% H 2, 75% Ar mixture with a flow rate of 50 sccm at 200 C for 2 h followed by a pretreatment at the standard WGS conditions (6.8% CO, 21.9% H 2 O, 8.5% CO 2, 37.4% H 2, and balance Ar) with a flow rate of 75.4 sccm at 140 C for 20 h. After this WGS pretreatment, the temperature was lowered to 120 C so that conversion was less than 10% and the WGS rates were determined. We refer to the rates measured immediately after pretreatment as the initial rates, and, for uniformity, these are the values presented as the results. The apparent reaction orders with respect to the reactants and products were measured by varying one gas concentration at a time (4 21% CO, 5 25% CO 2, 11 34% H 2 O, and 14 55% H 2 ) at 120 C, and the apparent activation energy was measured by varying the temperature over a range of 30 C, with the concentrations kept at the standard conditions. The WGS reaction rate for the catalysts reported here decays by less than 5% of the initial rate during the kinetic measurements. The catalysts were then exposed to Ar gas as the temperature was lowered to room temperature. Once at room temperature, the catalysts were passivated with a 2% O 2 in Ar mixture for 2 h. A more detailed discussion of the procedure for the WGS kinetic measurements is provided in our earlier work [2]. Since the brominated Au catalysts are prone to sintering with time at room temperature [11], these passivated samples were analyzed by TEM on the same day that the kinetic measurement experiments ended. 3. Results 3.1. Determination of Au particle sizes The particle size distributions of the used Au/TiO 2 and brominated Au/TiO 2 catalysts were calculated from their TEM images, typical examples of which are shown in Fig. 1. The number average Au particle sizes of the used 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/ TiO 2 catalysts after PFR and operando FTIR reactor measurements are reported in Tables 1 and 2, respectively. The number, surface, and volume average Au particle sizes of the used 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts after PFR and operando FTIR reactor measurements are reported in Fig. 2 and S1. The number, surface, and volume average Au particle size of the used Fig. 1. Typical TEM images of used (A) Au/TiO 2, (B) 2.3%Au (1Br:25Au)/TiO 2 and (C) 2.3%Au (1Br:12Au)/TiO 2 catalysts used to determine the Au particle size distributions. Table 1 Summary of PFR results for 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts at 120 C, 6.8% CO, 21.9% H 2 O, 8.5% CO 2, and 37.4% H 2. Catalyst Number average Au particle diameter (nm) Rate/10 3 (mol H 2 ) (mol Au) 1 s 1 E a (kj mol 1 ) Apparent reaction orders H 2 O CO 2 CO H 2 2.3%Au/TiO ± ± ± ± ± ± ± %Au (1Br:25Au)/ 4.0 ± ± ± ± ± ± ± 0.05 TiO 2 Table 2 Summary of operando FTIR reactor results for 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts at 120 C, 6.8% CO, 11.0% H 2 O, 8.5% CO 2, and 37.4% H 2. Sample Number average Au particle diameter (nm) Rate/10 3 (mol H 2 ) (mol Au) 1 s 1 Operating active sites as% of total Au a (%) E a (kj mol 1 ) TOF a (s 1 ) 2.3%Au/TiO ± ± ± ± ± %Au (1Br:25Au)/ TiO ± ± ± ± ± 0.5 a Determined by transient isotopic switch experiments.

6 M. Shekhar et al. / Journal of Catalysis 293 (2012) Rate / (mol H 2 ) (mol Au) -1 (s) -1 at 120 o C 1.E-01 1.E-02 1.E-03 2x lower rate 2.3% Au/TiO 2 IR 2.3% Au/TiO 2 PFR 2.3%Au-(1Br:25Au)/TiO 2 IR 6xlower rate 2.3%Au-(1Br:25Au)/TiO 2 PFR Au/TiO 2 Rate = 0.22d -2.7±0.1 1.E Number Average Au Particle Size / nm Fig. 2. WGS reaction rate per total mole of Au calculated at 120 C, 6.8% CO, 21.9% H 2 O, 8.5% CO 2, and 37.4% H 2 versus number average Au particle size for Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts in PFR and operando FTIR reactor measurements. The data for Au/TiO 2 catalysts (blue squares) was first reported in our earlier work [2], and it has been re-plotted here for comparison with catalysts presented in this work. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 2.3%Au (1Br:25Au)/TiO 2 is 4.0 nm, 4.2 nm, and 4.5 nm and 2.3%Au (1Br:12Au)/TiO 2 is 2.6 nm, 2.7 nm, and 2.7 nm after plug flow reactor (PFR) measurements, respectively. The WGS kinetic measurements lasted 120 and 20 h for 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts, respectively. Therefore, the average Au particle size after the PFR measurements was found to increase with time on stream. The WGS reaction rate per total mole of Au for Au/TiO 2 catalysts varies with the number (d), surface (d s ) and volume (d v ) average Au particle size as d 2.7±0.1, d 2:80:1 s and d 2:90:1 v, respectively (Fig. 2 and S1). Figure S1 shows that the difference between the values of d, d s, and d v for the used Au/TiO 2, 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts is small. This confirms a narrow Au particle size distribution on these catalysts. To eliminate the errors associated with using an average Au particle size, the entire Au particle size distribution identified by TEM was used to compute the percentage of corner, perimeter, and surface sites on these catalysts. This was done by performing a summation of the fraction of corner, perimeter, and surface sites, determined from the truncated cubo-octahedra geometry [2], over the entire particle size distribution. The addition of halides to supported Au catalysts is commonly reported [10,11,13] to promote sintering of Au nanoparticles. We observed that the addition of an amount of Br equal to 16% of the total moles of Au, that is, 2.3%Au (1Br:6Au)/TiO 2 catalyst, resulted in a 6.7 nm number, 7.0 nm surface, and 7.2 nm volume average Au particle size after 20 h under WGS reaction mixture. These average particle sizes ( nm) are significantly higher than those caused by the addition of Br equal to 4% and 8% of the total moles Au, that is, the 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts ( nm) after 120 and 20 h under WGS, respectively. Therefore, the addition of Br at 4% and 8% of the total moles Au does not result in sintering of Au particles, possibly due to the low content of Br on them Determination of Au particle shape Fig. 3 shows typical high resolution (HR)-TEM images used to determine the Au nanoparticle shapes of the Au/TiO 2 and brominated Au/TiO 2 catalysts. From the HR-TEM images, the Au nanoparticles on these catalysts formed well faceted truncated cubooctahedrons. A truncated cubo-octahedral shape for Au nanoparticles was reported in our previous work on Au/Rutile [2] and Au/ Al 2 O 3 [15] catalysts. Based on this result, the geometry of Au nanoparticles was assumed to be truncated cubo-octahedral, that is, cubo-octahedral geometry terminated at the midline, to estimate the percentage of corner, perimeter, and surface sites on Au/TiO 2 and brominated Au/TiO 2 catalysts. It should be noted that Fig. 3B shows a Au nanoparticle with a cubo-octahedral geometry truncated at approximately 3/4th its height. However, the density of such nanoparticles was less than 10% for the catalysts used in this work (3 out of 31 Au nanoparticles observed in HR-TEM). The remaining Au nanoparticles possessed cubo-octahedral geometry truncated approximately at the midline Kinetic studies on Au/TiO 2 and brominated Au/TiO 2 catalysts The WGS kinetics in the PFR s was measured under 6.8% CO, 21.9% H 2 O, 8.5% CO 2, and 37.4% H 2. In order to compare the rates measured in the operando FTIR under 6.8% CO, 11.0% H 2 O, 8.5% CO 2, and 37.4% H 2 to PFRs, the WGS reaction rates measured in the operando FTIR reactors were adjusted to the WGS reaction conditions in PFRs (Fig. 2 and S1), using the kinetic parameters reported in Table 1. In general, this procedure required only accounting for an increase in water concentration using the water order and small adjustments to the temperature using the apparent activation energy. Fig. 2 shows the calculated WGS reaction rate per total mole of Au at 120 C under 6.8% CO, 21.9% H 2 O, 8.5% CO 2, and 37.4% H 2 in PFR and operando FTIR reactors for Au/ TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts. The WGS reaction rate per total mole of Au for Au/TiO 2 catalysts varies with the number Fig. 3. Typical HR-TEM images of used (A) Au/TiO 2, (B) 2.3%Au (1Br:25Au)/TiO 2 and (C) 2.3%Au (1Br:12Au)/TiO 2 catalysts, respectively, used to determine the particle shape.

7 98 M. Shekhar et al. / Journal of Catalysis 293 (2012) Ln [Rate / (mol H 2 ) (mol Au) -1 (s) -1 ] %Au/TiO 2 E app = 60 ± 3 kj (mol) %Au-(1Br:25Au)/TiO 2 E app = 57 ± 3 kj (mol) T -1 / K -1 Fig. 4. Arrhenius plots for 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts in PFR measurements. Ln [Rate / (mol H 2 ) (mol Au) -1 (s) -1 at 120 C] CO order = 0.75 CO 2 order = CO order = 0.85 H 2 O order = H 2 order = H 2 O order = CO 2 order = H 2 order = Ln [Pressure / atm] Fig. 5. Plots used to calculate the apparent reaction orders for 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts in PFR measurements. (d), surface (d s ), and volume (d v ) average Au particle size as d 2.7±0.1, d 2:80:1 s and d 2:90:1 v, respectively (Fig. 2 and S1). The data for Au/TiO 2 catalysts was first reported in our earlier work [2], and it has been re-plotted in Fig. 2 and S1 for comparison with catalysts presented in this work. The Au/TiO 2 catalyst with a number average Au particle size of 1.2 nm has a 100 times higher rate per total mole of Au than that with a number average Au particle size of 6.7 nm. The variation of rate with Au particle size has been used to conclude that the potential active species for Au/TiO 2 catalysts are the low coordinated Au sites for both the CO oxidation [9,18] and the WGS reactions [2]. The rate for the 2.3%Au/TiO 2 catalyst used in this work (prior to addition of Br) at 120 C is similar to that for Au/TiO 2 catalysts at the same number, surface or volume average Au particle size in PFR, and operando FTIR reactor measurements (Fig. 2). The rate for the 2.3%Au (1Br:25Au)/TiO 2 catalyst at 120 C is 6 and 2 times lower than for Au/TiO 2 catalysts at the same number, surface or volume average Au particle size in PFR, and operando FTIR reactor measurements, respectively (Fig. 2 and S1). The rate for the 2.3%Au (1Br:12Au)/TiO 2 catalyst was undetectable at 120 C in PFR and operando FTIR measurements. Fig. 4 shows the Arrhenius plots for 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts in PFR measurements. The apparent activation energies of 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts near 120 C are 60 and 57 kj (mol) 1 in the PFR measurements (Table 1 and Fig. 4) and 56 and 52 kj (mol) 1 in operando FTIR reactor measurements (Table 2), respectively. Fig. 5 shows the plot used to determine the apparent reaction orders for 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts in the PFR measurements. The H 2 O, CO 2, CO, and H 2 reaction orders of 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts at 120 C in PFR measurements are 0.30 and 0.35, 0.10 and 0.05, 0.75 and 0.85, and 0.20 and 0.15, respectively (Table 1 and Fig. 5). It was not possible to measure the apparent activation energy and reaction orders for the 2.3%Au (1Br:12Au)/TiO 2 catalyst because its WGS reaction rate was undetectable. The WGS reaction kinetics for the 2.3%Au/TiO 2 catalyst measured in the PFR and operando FTIR reactors is in the kinetically controlled regime [16]. This was established by satisfying the Koros Nowak (K N) criterion, that is, the WGS reaction rate was found to be proportional to the concentration of active sites for a series of Au/TiO 2 catalysts [16]. Additionally, the TiO 2 used is non-porous [16] and thus should be free of internal pore diffusion transport limitations Operando FTIR spectroscopy and transient isotopic switch experiments Fig. 6 shows the mass spectrometer (MS) response and time-resolved FTIR spectra during transient isotopic switch ( 12 CO to 13 CO) experiments on 2.3%Au (1Br:25Au)/TiO 2 catalyst during WGS at 120 C. The area between the envelope made by the Ne and 13 CO 2 can be used to estimate the operating and potential active sites and will be discussed later. The area between the envelope made by the Ne and 13 CO is small due to rapid switch of isotopic 13 CO with CO and the relatively small uptake of CO on the catalyst. The time-resolved FTIR spectra during transient isotopic switch in Fig. 6 shows the fast switch of CO adsorbed on Au 0 and Au d (IR peaks near 2100 cm 1 and 2030 cm 1, respectively, assigned in the literature [20] to CO adsorbed on Au 0 and Au d, respectively) to isotopic 13 CO adsorbed on Au 0 and Au d (IR peaks near 2052 cm 1 and 1992 cm 1, respectively). Fig. 7 shows the FTIR spectra (CO region) for 2.3%Au/TiO 2, 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts during WGS after 20 h exposure a.u. / MS signal Normalized Ne 13 CO ads Au 0 Absorbance / a.u. 13 CO ads Au δ CO CO-Au 0 13 CO-Au CO CO-Au δ- 13 CO-Au δ Wavenumber / cm Time / s Fig. 6. Comparison of normalized MS response and IR band area during transient isotopic switch experiments of 2.3%Au (1Br:25Au)/TiO 2 catalyst during WGS at 120 C: Ne (red solid line), 13 CO (blue solid line) and 13 CO 2 (pink solid line); and surface species (normalized IR band area): 13 CO adsorption on Au 0 ( 13 CO ads Au 0, green square with dotted line, 2052 cm 1 ), 13 CO adsorption on Au d ( 13 CO ads Au d orange triangle with dashed line, 1992 cm 1 ). Feed: 11% H 2 O, 37.5% H 2, 6.8% 13 CO, 13.5%Ne in He. The synchronization between IR and MS signals was done by shifting the response curve of the normalized IR band intensities of 13 CO gas phase to best match the response curve of the normalized Ne MS signal. The synchronization procedure is detailed in our previous work [16]. Inserted figure shows the time-resolved FTIR spectra while the isotopic transient experiment is switching from CO to 13 CO at time = 0, 0.28, 0.42, 0.56, 0.7, 0.84, 0.98, 1.12, and 2 s, relative to the switch. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Normalized IR band intensity / a.u.

8 M. Shekhar et al. / Journal of Catalysis 293 (2012) Absorbance / a.u to the reaction mixture at 200 C. While the peak area of the IR peak near 2100 cm 1 decreases, the area of the IR peak near 2030 cm 1 increases upon addition of more Br to 2.3%Au/TiO 2 catalyst. 4. Discussion %Au/TiO Counting Au catalytic sites 2.3%Au-(1Br:25Au)/TiO 2 2.3%Au-(1Br:12Au)/TiO Wavenumber / cm Fig. 7. FTIR spectra of 2.3%Au/TiO 2, 2.3%Au (1Br:25Au)/TiO 2, and 2.3%Au (1Br:12Au)/TiO 2 catalysts during WGS at 200 C Bromine poisoning Fig. 3 shows that the Au particles on 2.3%Au/TiO 2, 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts formed well faceted shapes that resembled truncated cubo-octahedra. Thus, a physical model of Au nanoparticles as truncated cubo-octahedron, similar to that used by Williams et al. [2], can be used to determine the percentage Au that are surface, perimeter, and corner sites as a function of Au nanoparticle size distribution according to the equation: P sðdþ Percentage of site of interest ¼ P tðdþ here s(d) is the number of atoms for a particle of diameter d that correspond to the site of interest, t(d) is the total number of atoms for that cluster, and the summation was carried out over all the Au particles identified from the TEM images of that catalyst. Thus, we estimate that the Au clusters in 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts after PFR measurements had 1% and 3% of the total Au atoms as corner sites in contact with support, 3% and 8% as perimeter sites and 25% and 40% as surface sites, respectively. The Au clusters in 2.3%Au (1Br:25Au)/TiO 2 catalyst after operando FTIR reactor measurements had 2% of the total Au atoms as corner sites in contact with support, 6% as perimeter sites and 35% as surface sites. The 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts had a number of moles of Br equal to 4% and 8% of the total moles of Au. Therefore, if we assume that 1 Br atom blocks 1 Au atom, the 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts after PFR measurements had sufficient Br to cover all the perimeter sites and 16% and 20% surface Au atoms, respectively, whereas the 2.3%Au (1Br:25Au)/TiO 2 catalyst after operando FTIR reactor measurements had sufficient Br to cover only 66% of the perimeter sites and 11% of the surface Au atoms. In the physical model, the corner sites in contact with the support have Au Au coordination number of 4, and the perimeter sites in contact with support include the corner sites and the remaining perimeter sites that have Au Au coordination number of 5 [2]. These low coordinated Au sites have been shown to bind more strongly to CO and O as compared to bulk Au [9]. Fig. 2 shows the WGS reaction rate per total mole of Au at 120 C under 6.8% CO, 21.9% H 2 O, 8.5% CO 2, and 37.4% H 2 in PFR and operando FTIR reactors and the number average Au particle size for Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts. The WGS reaction rate per total mole of Au for Au/TiO 2 catalysts varies with the average Au particle size (d) asd 2.7±0.1 at 120 C (Fig. 2) and correlates with the fraction of corner Au sites (Au Au coordination number of four) to the total Au sites that vary as d 2.9 in a truncated cubo-octahedral geometry truncated at the midline [2]. The rate per total mole of Au for 2.3%Au/TiO 2 catalyst (prior to addition of Br) at 120 C is similar to Au/TiO 2 catalysts at the same number, surface or volume average Au particle size in PFR, and operando FTIR reactor measurements (Fig. 2). As already mentioned, it is essential to compare the rates of the brominated Au/TiO 2 catalysts to Au/TiO 2 catalysts at the same particle size. The WGS reaction rate per total mole of Au for 2.3%Au (1Br:25Au)/TiO 2 catalyst at 120 C under 6.8% CO, 21.9% H 2 O, 8.5% CO 2, and 37.4% H 2 is 6 and 2 times lower than for Au/TiO 2 catalysts at the same number, surface or volume average Au particle size in PFR, and operando FTIR reactor measurements, respectively. The rate for 2.3%Au (1Br:12Au)/TiO 2 catalyst is undetectable at 120 C in PFR and operando FTIR measurements. From the density functional theory (DFT) results provided in the literature [12], bromine is unlikely to be mobile during the WGS reaction over Au/TiO 2 catalysts because it binds more strongly to the coordinatively unsaturated Au sites than to the coordinatively saturated Au sites. The WGS reaction rate for the 2.3% Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts decays by about 15% in the first 20 h under WGS reaction mixture. After this initial deactivation, the rate remains stable and decays by less than 5% during the kinetic measurements that last for 4 5 days. Thus, based on the DFT computations and the stability of our catalysts, Br is not mobile during WGS on Au/TiO 2 catalysts. Potassium has been reported [21] to be a promoter of the WGS reaction rate for Pt/Al 2 O 3 and Pt/SiO 2 catalysts. The effect of addition of K on Au/TiO 2 catalyst was determined by impregnating a 2.3%Au/TiO 2 catalysts with a solution containing 16% moles KNO 3 of total moles Au, that is, a 2.3%Au (1KNO 3 :6Au)/TiO 2 catalyst. The WGS reaction rate per total mole of Au for 2.3%Au (1KNO 3 :6Au)/TiO 2 catalyst, with a number average Au particle size of 2.7 nm, is mol H 2 (mol Au) 1 s 1 at 120 C. From the dependence of WGS reaction rate per total mole of Au with number average Au particle size (Fig. 2), a Au/TiO 2 catalyst with a number average Au particle size of 2.7 nm should have a WGS reaction rate per total mole of Au of mol H 2 (mol Au) 1 s 1 at 120 C. Therefore, K does not promote the WGS reaction rate of Au/TiO 2 catalysts. The apparent activation energies of 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts near 120 C are similar in PFR measurements at 60 and 57 kj (mol) 1, respectively (Table 1 and Fig. 4) and operando FTIR reactor measurements at 56 and 52 kj (mol) 1, respectively (Table 2). The H 2 O, CO 2, CO, and H 2 reaction orders of 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts at 120 C in PFR measurements are similar at 0.30 and 0.35, 0.10 and 0.05, 0.75 and 0.85, and 0.20 and 0.15, respectively (Table 1 and Fig. 5). Therefore, these catalysts have a different number of active sites but with the same chemical nature. Thus, the 6 and 2 times decrease in the WGS reaction rate per total mole of Au for the 2.3%Au (1Br:25Au)/TiO 2 catalyst in PFR and operando FTIR reactor measurements, respectively, is primarily due to poisoning of the active sites by Br. Since there is not enough bromine to poison all Au sites and Br may interact with sites on the titania and Au surface atoms that are not catalytically active, only some active sites are poisoned by Br while others remain unpoisoned. The

9 100 M. Shekhar et al. / Journal of Catalysis 293 (2012) residual rate of 2.3%Au (1Br:25Au)/TiO 2 catalyst is due to the unpoisoned active sites. In summary, although the location of the Br on 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts is unknown, it is established that addition of an amount of Br corresponding to the number of low coordinated perimeter atoms decreases most of the catalytic activity exhibited by the 2.3%Au/ TiO 2 catalyst without altering the apparent activation energies and reaction orders. Thus, bromine poisons the active sites of the 2.3%Au/TiO 2 catalyst, and it is clear that not all surface atoms of Au exhibit the same rate Transient isotopic switch experiments Fig. 6 shows the mass spectrometer (MS) response and time-resolved FTIR spectra during transient isotopic ( 12 CO to 13 CO) switch experiments of the 2.3%Au (1Br:25Au)/TiO 2 catalyst during WGS at 120 C. If it is assumed that the WGS reaction rate is proportional to the coverage of 13 C on the surface and that the reaction pathway is irreversible, a 13 C mass balance on the reactor shows that the area of the envelope made by the Ne (scaled to represent 13 CO) and the 13 CO 2 MS response curves multiplied by the WGS reaction rate per total mole of Au is the fraction of the total moles of Au that are the operating active sites on the catalyst [16,19]. The operating active sites are defined as the multiplication of all potential active (L) sites and h 13C. From the transient isotopic switch experiments, 0.74 ± 0.10% and 0.37 ± 0.07% of the total Au atoms were determined by the procedure above to be the amounts of operating active Au sites on 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalyst, respectively (Table 2). These used catalysts have a similar number, surface, and volume average Au particle size, 3.1 nm, 3.2 nm and 3.4 nm for 2.3%Au/TiO 2 and 2.9 nm, 3.0 nm and 3.1 nm for 2.3%Au (1Br:25Au)/TiO 2, respectively, after operando FTIR reactor measurements. Although, during operando FTIR reactor measurements, the WGS reaction rate per total mole of Au for 2.3%Au (1Br:25Au)/ TiO 2 catalyst is half that of 2.3%Au/TiO 2 catalyst (Fig. 2), the turnover frequency (TOF) calculated based on the number of operating active sites determined by transient isotopic switch experiments on the 2.3%Au (1Br:25Au)/TiO 2 catalyst is 1.6 ± 0.5 s 1 at 120 C, which is within experimental error of the equivalent TOF of 2.3%Au/TiO 2 catalyst (1.5 ± 0.4 s 1 ) at 120 C (Table 2). Thus, the decrease in the WGS reaction rate upon the addition of Br to 2.3%Au/TiO 2 catalyst is primarily due to the poisoning (blocking) of the active sites by adsorbed Br. It is noted that the particle size distribution for the Au in 2.3%Au (1Br:25Au)/TiO 2 catalyst after operando FTIR reactor measurements yields corner sites, perimeter sites, and surface sites equal to 2%, 6%, and 37% of the total number of Au atoms. Thus, there is not enough Br (at 4% of the total number of Au atoms) to cover all of the perimeter or surface sites for 2.3%Au (1Br:25Au)/TiO 2 catalyst after the operando FTIR reactor measurements. The percentage of Au operating sites determined by transient isotopic switch experiments was used to estimate the percentage of potential active sites on the 2.3%Au/TiO 2 catalyst by the following reasoning. The coverage of CO (h CO ) for Au/Al 2 O 3 catalysts at 120 C was reported to be 0.35 ± 0.04 of the adsorption of CO on that catalyst at room temperature in our previous work [22]. The adsorption strength of CO for Au catalysts has been shown to be independent of the support by CO TPD [23] and the fact that the CO order is in the range of for gold on alumina and on titania in WGS kinetic measurements [15]. The CO reaction order on the 2.3%Au/TiO 2 catalyst at 120 C of 0.85 (Table 1) then indicates weak adsorption of CO on the surface and is consistent with loss of 2/3rd of the CO coverage when the temperature was increased from room temperature to 120 C, as was the case for Au on alumina. The area between the envelope made by the Ne and 13 CO in transient isotopic switch experiments is small (Fig. 6) due to rapid and reversible adsorption of isotopic 13 CO and a small uptake capacity of the catalyst for CO, again in agreement with the weak adsorption on the Au. We assume further, on the basis of the fact that operando IR showed no other carbon-containing intermediates than adsorbed CO, that the adsorbed CO represents an upper bound on the number of potentially active sites and that CO covers all the potentially active sites at room temperature. On the basis of the discussion above, that, at 120 C, only about 1/3rd of those sites are occupied, the percentage of the total Au atoms that can be potential active sites on the 2.3%Au/TiO 2 catalyst can then be estimated to be the measured number operating sites at 120 C (0.74% of the total Au atoms) divided by 1/3, or 2%. Thus, the percentage of the total Au atoms that are potential active sites on the 2.3%Au/TiO 2 catalyst best correlates with the percentage of corner atoms in contact with the support (also 2%) determined from the truncated cubo-octahedron model. It is noted that less than 10% of the Au nanoparticles in these catalysts, observed by HR-TEM, possess cubo-octahedral geometry truncated at approximately 3/4th its height while the remaining are truncated at approximately the midline. This conclusion regarding the number of sites, however, does not change with the choice of geometry of Au nanoparticles because the fraction of corner sites in contact with the support to the total Au sites for a cubo-octahedral geometry truncated at 3/4th its height is approximately 3/5th than that for the cubo-octahedral geometry truncated at the midline. Therefore, our estimation of corner sites in contact with the support can vary within a factor of two at most. It has been argued that the calculated percentage of operating sites may include contributions from readsorption or secondary reactions of CO 2 [24,25]. Due to these effects, the calculated percentage of operating active sites presented here could be overestimated and can be considered an upper bound. It has been shown in our previous work [16] that the measured average residence time of surface intermediates (s) for the 2.3%Au/TiO 2 catalyst when cofeeding 8% CO 2 was 0.6 ± 0.1 s at 120 C, which is within experimental error of the calculated s without co-feeding CO 2, 0.65 ± 0.05 s. In addition, the calculated s in transient isotopic (CO 2 to 13 CO 2 ) switch experiments for 2.3%Au/TiO 2 catalyst was as low as 0.05 s, indicating an insignificant contribution from readsorption of CO 2. Further, the similar TOF, calculated based on the number of operating active sites, for 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts indicates that readsorption or secondary reactions of CO 2 did not significantly affect our results. To conclude, the TOF calculated based on the number of operating active sites determined on the 2.3%Au/TiO 2 and 2.3%Au (1Br:25Au)/TiO 2 catalysts is similar at 1.6 ± 0.5 s -1 at 120 C, indicating poisoning of the potential active sites by Br. The estimated potential active sites (2% of the total number of Au atoms) on the 2.3%Au/TiO 2 catalyst best correlates with the corner atoms of the cubo-octahedral particles being active Operando FTIR spectroscopy Fig. 7 shows the CO region of the steady-state operando FTIR spectra of the 2.3%Au/TiO 2, 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts at 200 C. The IR peaks observed near 2100 cm 1 and near 2030 cm 1 are assigned in the literature to CO adsorbed on Au 0 and Au d, respectively [20]. The normalized area of the IR peak observed near 2100 cm 1 was found to correlate with the WGS reaction rate on Au/TiO 2 catalysts with different average Au particle sizes [2], during deactivation of Au/Al 2 O 3 [15] and Au/CeZrO 4 [20] catalysts. Fig. 8 shows that the normalized peak area of CO adsorbed on Au 0 near 2100 cm 1 determined by operando FTIR passes through the origin and varies linearly with the steady-state WGS rate at 200 C on 2.3%Au/TiO 2, 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/TiO 2 catalysts. The nor-

10 M. Shekhar et al. / Journal of Catalysis 293 (2012) Rate / 10-2 (mol H 2 ) (mol Au) -1 s Series1 CO on Au 0 Series2 CO on Au δ- Series3 Total CO adsorbed 2.3%Au/TiO 2 2.3%Au-(1Br:25Au)/TiO 2 2.3%Au-(1Br:12Au)/TiO Normalized CO peak areas from operando IR / % Fig. 8. WGS reaction rate per total mole of Au versus normalized FTIR peak area of CO adsorbed on Au 0,Au d, and total CO adsorbed during operando FTIR reactor measurements on 2.3%Au/TiO 2, 2.3%Au (1Br:25Au)/TiO 2 and 2.3%Au (1Br:12Au)/ TiO 2 catalysts at 200 C. malized peak area of CO adsorbed on Au d near 2030 cm 1 and the total CO adsorbed also vary linearly with the rate but do not pass through the origin. The normalized peak areas were calculated by assigning the peak areas of the 2.3%Au/TiO 2 catalyst as 100%. Thus, the rate correlates with the concentration of CO on Au 0 indicating that the active Au sites, viz. the low coordinated corner Au sites, are metallic in nature. For the 2.3%Au (1Br:12Au)/TiO 2 catalyst, the WGS reaction rate was undetectable and only 5% of the original CO adsorbed on Au 0 remained. Therefore, the assumption that at 200 C, all of the CO adsorption on the 2.3%Au/TiO 2 catalyst occurs on the active low coordinated corner Au atoms is good to within 5%. We note further that the decrease in area of the 2100 cm 1 CO peak to 50% for the (1Br:25Au) sample versus the unpoisoned sample is in good agreement with the 50% decrease in the number of operating active sites measured in the isotope switch experiment. In addition, the CO vibrational frequency does not vary in the operando FTIR measurements on the poisoned and unpoisoned catalysts. This constant frequency with changes in CO coverage indicates the absence of dipole dipole interactions for CO on these catalysts. This result could be related to the low coverage of CO on these catalysts and that CO molecules are adsorbed on sites separated from each other, as would be expected for the corner atom sites. Counting active sites can depend on how each reactant is accommodated on the surface of the catalyst. However, for our specific case, that is, the WGS reaction over Au/TiO 2 catalysts, it is shown from Figs. 7 and 8 that the WGS reaction rate scales linearly with the CO adsorbed on Au 0, and therefore, since the coordinatively unsaturated corner atoms are the rate controlling active sites, the counting of such sites is not affected by the accommodation of H 2 O on the surface. The analysis of the reaction orders for the catalysts reported here shows that H 2 O is more strongly bound to the surface than CO because the apparent H 2 O order is 0.3 compared to the CO order that is 0.8. Negative H 2 O order implies a high relative surface coverage of hydroxyl species such as H 2 O, OH, and O. Since we have already shown that the rate is proportional the number of Au corner atoms, this information that the water activation sites are highly covered shows that water activation sites do not become the rate controlling site population over our range of operating conditions. Thus, while the water mechanism is an interesting and still unanswered question, whether the water is competitive with CO on the Au sites or resides on TiO 2 sites proximate to the Au does not change the fact that the rate is proportional to the number of coordinatively unsaturated Au species. In our work here, we conclude that the WGS reaction rates are proportional to the number of metallic corner Au sites. However, based on the data presented here, we cannot draw a precise model for TiO 2 showing sites responsible for water activation. In a recently published paper from our group [15] and DFT calculations in the literature [26], the support sites are important for H 2 O adsorption and dissociation over Au/TiO 2 catalysts. The apparent H 2 O order varies with the support and correlates with the reaction rate for Au/TiO 2, Au/Al 2 O 3 and Au/Al 2 O 3 -World Gold Council (WGC) catalysts [15]. The activation barrier for dissociation of water on Au(100) is 1.5 ev [27], onau 29 nanoparticle is 1.3 ev [27] and on TiO 2 Au interface of a TiO 2 /Au(111) system is 0.6 ev [26]. We note, however, that the dependence of the rate on Au particle size and on the amount of CO adsorbed on coordinatively unsaturated Au, together with the low water order that indicates high coverage of water activation sites, indicate that the Au corner atoms are the sites that control the rate over our entire operating space and that this conclusion is, therefore, not affected by the mechanism of water activation. 5. Conclusions The addition of Br at a level of 4% of the total moles of Au to a 2.3%Au/TiO 2 catalyst decreased its WGS rate by six times in PFR and 2 times in operando FTIR reactor measurements as compared to Au/TiO 2 catalysts at the same number, surface or volume average Au particle size. The addition of Br did not result in an appreciable change in the apparent activation energy or the reaction orders. Operando FTIR and transient isotopic experiments together with the Br poisoning results were used to confirm that the dominant active sites on Au/TiO 2 catalysts for the WGS reaction are the low coordinated metallic corner Au sites. The TOF for Au/TiO 2 catalysts, based on the operating active sites (corner Au atoms), determined from transient isotopic switch experiments under 6.8% CO, 8.5% CO 2, 11.0% H 2 O, 37.4% H 2, and atmospheric pressure, was determined to be 1.6 ± 0.5s 1 at 120 C. The low coordinated corner Au sites are present due to the truncated cubo-octahedra geometry of the Au nanoparticles. For Au/ TiO 2 catalysts with an average Au particle size of 3 nm or above, less than 1 percent of the total Au atoms in the Au nanoparticles are responsible for most of the catalytic activity. Although these conclusions pertain to the WGS catalysis over supported Au nanoparticles, this knowledge can be extended to develop a twofold strategy to further improve the catalytic rate for Au catalysts for reactions catalyzed by Au. First, it is important to increase the percentage of low coordinated Au atoms in Au nanoparticles by synthesizing them either with smaller size or with different nanoparticle shapes. Secondly, from the Sabatier principle, developing supported Au species with Au Au coordination less than 4 is needed to determine the optimum catalytic rates exhibited by supported Au nanoparticles. Acknowledgments Support for this research was provided by the U.S. Department of Energy, Office of Basic Energy Sciences, through the Catalysis Science Grant No. DE-FG02-03ER The authors would like to thank Mr. Leonardo Maciel for his help in conducting the WGS kinetic experiments. E.A.S. acknowledges additional support through the U.S. Department of Energy, Office of Basic Energy Sciences, under Contract No. DE-AC02-98CH10886.