Au/iron oxide (Fe³⁺) catalyst was prepared from Fe(NO₃)₃·9H₂O (Wako Pure Chemical Industries, Ltd.: Osaka, Japan) and HAuCl₄·4H₂O (Kojima Chemicals Co., Ltd.: Osaka, Japan) as precursor materials. The preparation method has been described in detail in our previous report [21]. The precursors were dissolved in distilled water to prepare 200 mL of 1.69–1.86 ×10² mmol L⁻¹ iron nitrate aqueous solution and 100 mL of 0.15–1.52 × 10 mmol L⁻¹ chlorauric acid aqueous solution for the synthesis of 3 g Au/iron oxide catalyst containing 1, 3, and 10 wt% Au. The pH value of the iron nitrate solution was controlled at 8.2 by adding Na₂CO₃ powder (Wako Pure Chemical Industries, Ltd.: Osaka, Japan) for hydrolysis, and the suspension was vigorously stirred in a water bath at 60 °C. After 1 h of aging, the chlorauric acid solution was added to the iron hydroxide solution, followed by additional 0.5 h stirring. During the aging and mixing processes, the pH value was maintained within 8.0–8.4 by the addition of diluted nitric acid and sodium carbonate. The precipitates were filtered and washed with 1 L of distilled water (60 °C) to eliminate Cl⁻ and Na⁺. Finally, they were vacuum-dried for 12 h at 80 °C, followed by calcination in a heater for 2 h at 200 or 400 °C. The prepared catalysts were sieved to the particle size between 0.3 and 1.0 mm for catalytic activity tests. Following is a summary of the catalysts prepared under different conditions: calcined at 200 °C with 1, 3, and 10 wt% Au and calcined at 400 °C with 3 wt% Au. In addition, iron oxide calcined at 200 °C with no Au content was also prepared by the same procedure as a reference. The catalysts, thus prepared, are collectively denoted as Au/Fe₂O₃, although the iron oxide might contain other oxidation states such as ferrihydrite, Fe₅HO₈·4H₂O.

X-ray diffraction (XRD) patterns of catalysts were measured by an X-ray diffractometer (RIGAKU, multiflex) equipped with a Cu-K_α radiation source at a voltage of 40 kV and a current of 40 mA. The BET surface area and the pore size distribution were obtained from N₂ adsorption data at 77 K (BELJAPAN, Belsorp-mini) using the BET equation and the BJH method, respectively. The morphology and microstructure of catalysts were observed using a transmission electron microscope (TEM; JEOL, JEM-1010) at an accelerating voltage of 100 kV. For the TEM observations, samples were ground to fine powder and dispersed in ethanol using ultrasonic agitation. A drop of the suspension was then transferred to a carbon-coated Cu mesh grid and dried at room temperature. X-ray photoelectron spectroscopy (XPS) spectra were measured with SHIMADZU ESCA-3400 using an Al K_α radiation source at an accelerating voltage of 10 kV and an emission current of 20 mA. The measurements were performed under vacuum pressure of less than 5 × 10⁻⁶ Pa. A correction of binding energy was performed, referencing the position of C 1s peak at 284.8 eV.

Prepared catalysts were loaded into a quartz tube reactor with a 7 mm inner diameter for the catalytic activity test. The reaction temperature was controlled with an infrared gold image furnace by monitoring the temperature at inlet and outlet of the catalyst bed. Two kinds of WGS tests were conducted for examining catalytic activity: one is the temperature dependence test using as-prepared catalysts; and the other is the WGS at 80 °C using the samples reduced in various ways from Au/Fe₂O₃ catalyst (Au: 3 wt%) calcined at 200 °C. In the temperature dependence test, the prepared Au/Fe₂O₃ catalysts were used without pre-reduction. The catalyst bed was heated to the starting temperature, 140 °C, under an Ar stream and kept for 0.5 h to remove air and moisture, followed by the introduction of the reactant gas. The reactant gas (H₂ 45.9%, CO 8.3%, CO₂ 4.1%, H₂O 41.7%) was fed to 0.4 g of the catalyst bed at gas hourly space velocity (GHSV) of 2600 h⁻¹. In the WGS reaction test at 80 °C, the catalyst was reduced prior to the reaction. The pre-reduction was conducted under three different reducing gas streams, CO/N₂, H₂/Ar, and CO/H₂O/N₂, at two different temperatures, 140 and 200 °C. The flow rates of the reducing gases were 20, 20, and 26 (steam/carbon ratio, S/C = 10) mL min⁻¹, respectively, over 0.3 g catalyst, where the flow rates of reducing gases, CO and H₂, were unified to 0.62 mL min⁻¹. After a 2 h pre-reduction, the temperature was decreased under an Ar stream to room temperature for characterizations of the reduced samples or to 80 °C for catalytic activity tests. In the WGS test at 80 °C, the reactant (CO/H₂O/N₂ = 2.2/30.3/67.5) was introduced into the catalyst bed at GHSV = 2800 h⁻¹. Since the partial pressure of H₂O, about 30 kPa, is sufficiently lower than the saturation vapor pressure at 80 °C, 47.4 kPa, the vapor was not condensed to liquid.

The effluent gas from the reactor was diverted through a cold trap to remove H₂O, mixed with an Ar flow of 50 mL min⁻¹ for dilution, and then analyzed using an on-stream gas chromatograph (VARIAN, CP-4900 Micro-GC) equipped with a thermal conductivity detector and two columns, a Molsieve 5A and a PoraPLOT Q. Analytical values were expressed as the conversions of CO and H₂, and the yields of CO₂ and H₂. The conversions were calculated from the ratio of the inlet and outlet molar flow rates as 1 – F_outlet/F_inlet, where F stands for the molar flow rate of CO or H₂. The yields were obtained from the ratio of the outlet molar flow rate of CO₂ or H₂ over the inlet molar flow rate of CO.

Figure 1 compares the catalytic activities of the prepared Au/Fe₂O₃ catalysts in WGS. All of the samples were tested in the same temperature profile, wherein the temperature was raised at a rate of 10 °C min⁻¹ and maintained for 40 min at the given reaction temperature. An average value of conversion was calculated from three data collected during the last 10 min of the constant temperature period. Thermodynamically calculated equilibrium CO conversion curve is also drawn.

The CO conversions reached the equilibrium values below 200 °C in the samples loaded with 10 and 3 wt% Au and calcined at 200 °C. The CO conversion using these samples gradually decreased with increasing temperature and deviated from the equilibrium one because of a catalytic deactivation. In terms of the Au content, 3 wt% was enough to convert CO at low temperatures. However, the effect of Au content on CO conversion was significantly reduced at higher temperatures. The sample calcined at 400 °C showed a lower activity at low temperatures, while the difference of conversion among the gold catalysts was reduced with the increase in reaction temperature.

These results reflect the reaction system of WGS using an Au/iron oxide catalyst; the catalytic activity is caused by small Au particles activated by synergism with the support at low temperatures, gradually shifting to be support-dependent at high temperatures [20]. The catalytic activity of gold is diminished by the agglomeration of Au particles with the increase in reaction temperature. On the other hand, the reaction involves the reduction of Fe₂O₃ to Fe₃O₄ phase, which is a catalytically active oxidation state [19,20] The shift in catalytic activity from Au-dependent to iron oxide dependent is clearly seen in the conversion curve of 1 wt% Au catalyst that showed low conversion of CO at 140 °C because of the low content of Au. The catalytic activity of Au increased with temperature up to 180 °C and then decreased due to the agglomeration, while the active site shifted to iron oxide at around 250 °C, leading to the increase in conversion at 300 °C.

The catalytic activity was thus affected by complicated structural change during the reaction. A temperature-programmed reduction (TPR) study of Au/iron oxide is well established, demonstrating that the reduction of iron oxide by H₂ at low temperature is accelerated by the existence of Au and by the lower crystallinity of iron oxide [22,23]. These studies suggest that the gold catalysts calcined at 200 °C of this study were activated to the form of Au/Fe₃O₄ at the first test point of 140 °C. Namely, the control of the structure of Au/Fe₃O₄ is essential to enhance the catalytic activity at low temperatures. In the following sections, we discussed the low-temperature pre-reduction for Au/Fe₂O₃ catalyst (Au 3 wt%) calcined at 200 °C, which showed high catalytic activity, to develop more efficient catalyst for low-temperature WGS.

The reduction temperatures employed were 140 and 200 °C. As reducing gases, CO and H₂ were used because they are possible reactants involved in WGS to reduce iron oxide, and CO/H₂O was also used as a WGS condition. Using these conditions, the pre-reductions were conducted for 2 h, producing six kinds of reduced samples. The reduced catalysts were represented by reduction temperature and reducing gas, e.g., 140-CO.

First, the reaction mechanism during the pre-reduction of Au/Fe₂O₃ calcined at 200 °C was qualitatively discussed by analyzing the product gas. The produced gas was analyzed every 3 min during the treatment, and the conversions of CO or H₂ and the yields of H₂ and CO₂ were calculated. Figure 2 shows the change in the product gases for each treatment at 140 °C; (a) H₂ reduction, (b) CO/H₂O reduction, and (c) CO reduction. In the pre-reduction treatments at 200 °C, similar results were obtained. The amounts of formed or consumed CO, H₂, and CO₂ were calculated by integrating the conversions and yields, and summarized in Table 1. In H₂ pre-reduction treatment, the amount of H₂ consumed was nearly equal to the stoichiometrically calculated value, 1.91 mmol g-cat⁻¹, based on the equation: Fe₂O₃ + 1/3H₂ → 2/3Fe₃O₄ + 1/3H₂O. The reduction of iron oxide by H₂ is believed to progress through a dissociation of H₂ into hydrogen atoms on Au particles and the following spillover on iron oxide to drive the reduction as: Fe³⁺OH⁻ + H → Fe²⁺ + H₂O [22]. In CO/H₂O pre-reduction treatment, CO was completely consumed throughout the operating period, and CO₂ was maintained at stoichiometric yield; however, it took about 40 min for H₂ to reach the stoichiometric yield. This result suggests that H₂ produced by WGS at the initial stage of the treatment was consumed for reducing iron oxide. The amount of H₂ consumed during CO/H₂O pre-reduction was almost the same with that during H₂ pre-reduction. From these results, it was clarified that H₂ had a significant role to reduce iron oxide in the reduction. On the other hand, in CO pre-reduction treatment, the amount of CO consumed was about two-fold greater than the amount of H₂ consumed in other two pre-reductions, indicating that the CO pre-reduction cannot be represented simply by the equation, Fe₂O₃ + 1/3CO → 2/3Fe₃O₄ + 1/3CO₂.

Andreeva et al. [14] proposed the reaction mechanism of WGS on the Au/α-Fe₂O₃ catalyst: the redox transfer cycle between Fe²⁺ and Fe³⁺ by ad/desorption of carbon species and the dissociative adsorption of H₂O on Au particles. In this scheme, formate is formed from the adsorbed CO and the adjacent hydroxyl group, followed by desorption of CO₂ and H atom accompanying the reduction of Fe³⁺ to Fe²⁺ as: AuOH⁻Fe³⁺ + CO → Au□Fe³⁺-COOH⁻ → Au□Fe²⁺ + CO₂ + H, where the symbol “□” expresses the site of iron oxide adjacent to Au, which presumably activates the Au for the dissociative adsorption of H₂O.

Taking this scheme into consideration, an additional experiment was conducted for the CO-reduced sample as follows; the sample after 140-CO treatment was kept at the reduction temperature for 30 min under an Ar stream to completely purge CO, and then H₂O with carrier Ar (H₂O/Ar = 20/80) was introduced to the catalyst bed. Figure 3 shows the yield of product gas from the H₂O introduction. H₂ began to form at the moment of H₂O introduction, and the yield reached the maximum at 9 min and decreased to 0 within 40 min. Simultaneously, a small amount of CO₂ was formed. The amounts of H₂ and CO₂ formed through this operation were 1.34 and 0.38 mmol g-cat⁻¹, respectively. This result confirms that the sample reduced by CO has a surprisingly active site for dissociative adsorption of H₂O, expressed by the symbol “ □”. The total amount of CO₂ formed through the successive treatment of CO reduction and H₂O introduction was balanced with the amount of CO consumed in the reduction treatment. As for H₂, the total amount was almost identical to the amount of H₂ consumed in H₂ pre-reduction and equal to half of that of CO₂ formed. Therefore, we can describe the scheme of the CO reduction treatment as: AuOFe³⁺ + CO → Au□Fe²⁺ + CO₂. The following steam treatment involves the dissociative adsorption of H₂O; therefore, the scheme is presumed as: Au□Fe²⁺ + H₂O → Au-H⁺OH⁻ □Fe²⁺ → AuOH⁻Fe³⁺ + 1/2H₂, thus leading to the initiation of the aforementioned WGS cycle.

The cause for the small amount of H₂ formed in the reduction treatment (Figure 2(c)) and CO₂ formed in the H₂O treatment (Figure 3) cannot be specified, but minor reactions occurred by regional physical or chemical structure may account for this.

Analysis of the product gas from pre-reductions implied that iron oxide in the catalyst was reduced to Fe₃O₄ phase even at 140 °C. Structural characteristics and XRD patterns of the reduced samples are shown in Table 2 and Figure 4, respectively. The BET surface area of the non-treated sample was about 200 m² g⁻¹, with micropores having peak radius of 1.7 nm.

The surface area significantly decreased with the increase in pore peak radius by the pre-reduction. The sample before the pre-reduction showed no distinct peaks in the XRD pattern (data not shown). On the other hand, XRD patterns of the reduced samples showed that the iron oxide was in the form of Fe₃O₄, although no peak appeared in the pattern of 140-CO. The ratio of intensities at the peaks of 30.1° and 35.5° (the most intense peak of magnetite), I_30.1°/I_35.5°, is 30 for pure Fe₃O₄ [24]. When the crystallized iron oxide contains α-Fe₂O₃ phase, the intensity ratio decreases because α-Fe₂O₃ has the second most intense peak at 35.7° (overlapping with the most intense peak of magnetite) but does not have in the vicinity of 30.1°. The I_30.1°/I_35.5° values of the reduced samples were between 28 and 35, with the exception of 140-CO that had undetectable crystallinity level. These results confirm that the majority of amorphous iron oxide was reduced to Fe₃O₄ and crystallized during the pre-reduction. The highest level of crystallization occurred when CO/H₂O was used as reducing gas, followed by H₂. The reduction temperature also affected the crystallization and reduction ability. The surface area of the samples reduced at 140 °C was about 1.7 times higher than that of the samples reduced at 200 °C for all reducing gases. The reason for the different structure of 140-CO is still not clear, but is believed to be related to the oxidation state of iron oxide. In other words, the transformation to Au□Fe²⁺ by CO pre-reduction has less impact on the structure of iron oxide, while the overall formation of Fe₃O₄ phase by H₂O treatment can cause crystallization accompanied by decrease in surface area. In fact, when the 140-CO was treated with steam at 80 °C for 40 min, the surface area decreased to 71 m² g⁻¹.

Figure 5 shows TEM photographs of the prepared Au/Fe₃O₄ catalysts. The TEM observation confirms the difference between Fe₃O₄ particles produced by each reduction treatment. The Fe₃O₄ particles of 140-CO were less than 5 nm with their appearance being almost the same as that of the sample before pre-reduction, and the other catalyst also had small Fe₃O₄ particles of about 20 nm. These particles of Fe₃O₄ are surprisingly small compared to the direct synthesis of Fe₃O₄ in liquid phase from Fe²⁺/Fe³⁺ precursor by precipitation [25,26]. Most of the Fe₃O₄ particles formed through the precipitation method are on a micrometer scale because the particle growth after the nucleation of Fe₃O₄ by the dehydration reaction of Fe²⁺ and Fe³⁺ hydroxides cannot be controlled in the liquid phase synthesis. On the other hand, the formation of Fe₃O₄ particles in this study progressed in the gas phase. The crystal growth in the gas phase can be thermally controlled, as obtained for the samples reduced at 140 °C and 200 °C. However, if the calcination temperature of Au/Fe₂O₃ is the commonly-used 400 °C, the formed Fe₃O₄ particles would be large because the reduction of well-crystallized Fe₂O₃ requires far higher temperatures. Moreover, when Au particles are not loaded on the iron oxide, the small Fe₃O₄ particles would not be obtained because Au plays a role on decreasing the reduction temperature [22,23]. Thus, the small Fe₃O₄ particles were obtained only by reducing amorphous iron oxide loading of Au at low temperatures as proposed in this study, and the particle size was controllable with the reducing gas and reduction temperature.

TEM photographs also showed Au particles having a diameter below 4 nm. However, it was difficult to identify the Au particles in 140-CO, 200-CO, and 140-H₂. To investigate the state of Au loaded on the Fe₃O₄ particles, XPS analysis was performed. Au 4f XPS spectra of the reduced samples are shown in Figure 6. Although the bulk Au shows a peak of Au 4f_7/2 electron binding energy of core levels at 84.0 eV, the value negatively shifts in general if it is a supported small nano-particle [27,28], as found at around 83.8 eV in this study. On the other hand, the binding energy of the cationic Au positively shifts by 0.8–1.4 and 2.8–3.5 eV for Au¹⁺ and Au³⁺, respectively [27–31]. These tendencies are applicable to Au 4f_5/2 having a peak at 88.0 eV for the bulk Au. These facts indicate that the partial existence of the cationic Au lowers the overall spectral intensity of Au 4f and gives shoulders at a binding energy above that of metallic Au [20]. The intensity of the spectrum in this study was varied among the reduced samples, and the differences showed a broad correlation with the structural characteristics of iron oxide. The sample 200-CO/H₂O, having the lowest surface area, showed the highest intensity; in contrast, the intensity of 140-CO was the lowest. This indicates that the crystallization and agglomeration of Fe₃O₄ particles were accompanied by the reduction of the cationic Au, Au³⁺ and Au¹⁺, to metallic Au. Because the samples exhibiting a lower intensity of Au 4f peaks correspond to the samples having unobservable Au by TEM observation, the cationic Au was expected to be highly dispersed on the support in an undetectable state for the observation.

As shown in Figure 1, the Au/iron oxide catalyst calcined at 200 °C has sufficient catalytic activity to completely remove CO in the WGS at 140 °C. Therefore, we performed the WGS tests using the pre-reduced samples, namely Au/Fe₃O₄ catalysts, at the low temperature of 80 °C to differentiate conversions of the samples and to determine whether the reaction progresses at such a low temperature. Prior to the tests, the reduced samples were purged under an Ar stream for 15 min at the reduction temperature, and the temperature was decreased to 80 °C, followed by H₂O treatment for 40 min.

The performances of each Au/Fe₃O₄ catalyst are shown in Figure 7 as CO conversion against reaction time for a period of 80 min. For all samples, stoichiometric amounts of H₂ and CO₂ were produced from H₂O and CO during the reaction period. The pre-reduced catalysts were shown to be active for WGS even at 80 °C. Complete conversion was achieved by 140-CO and 140-H₂ during the majority of the reaction period, and 200-CO produced conversion above 0.9, corresponding to the reaction rate of 2.3 × 10⁻⁵ mol-CO g-Au⁻¹ s⁻¹ or more. Although the conversions using the other samples were above 0.6 at the initial stage of the reaction, they gradually decreased. From the characterization of Au/Fe₃O₄ catalysts, the former and the latter groups were assigned to the reduced catalysts having surface areas greater and less than 80 m² g⁻¹, respectively. The Au/Fe₂O₃ catalyst calcined at 400 °C and preliminarily reduced by H₂ at 350 °C for 2 h was also examined for the 80 °C WGS reaction to compare the catalytic activity; however, it showed zero capability for CO conversion. These results elucidate our suggestion that the low-temperature pre-reduction of Au/Fe₂O₃ catalyst having amorphous iron oxide to prepare Au/Fe₃O₄ catalysts is very effective for low-temperature WGS reaction. Au/Fe₃O₄ catalysts that were appropriately reduced before WGS showed a potential to completely remove CO even at the operating temperature of PEFC, 80 °C, which indicated that a compact on-site H₂ production system, where the CO oxidation section is not necessary, can be developed using the catalyst developed in this study.

Figure 8 shows the result of 80 °C WGS using non-reduced Au/Fe₂O₃ catalyst. Although high conversion was obtained at the initial stage of the reaction, appropriate amounts of H₂ were not produced, which is similar to the reaction behavior observed during CO pre-reduction (Figure 2(c)). This indicates that the WGS cycle, which is promoted by redox transfer between Fe²⁺ and Fe³⁺, did not occur despite the presence of H₂O in the reactant. CO conversion and CO₂ yield show that the reduction of iron oxide progressed, but the low yield of H₂ shows that the catalyst was inactive for dissociative adsorption of H₂O. The reduction of iron oxide by CO was caused by high activity of Au nano-particles for adsorption of CO. In fact, Silberova et al. [32] found with DRIFTS study that the conversion of CO to CO₂ on Au/Fe₂O₃ catalyst occurs even at a room temperature through the decomposition of the carbonate species. Meanwhile, low activity for the dissociative adsorption of H₂O observed in this non-reduced catalyst shows that the site on iron oxide to activate Au for the reaction, “□”, was not formed. This is possibly due to a low crystallinity of the iron oxide support resulted from the low reaction temperature. In the case of Fe₂O₃-ZnO and Fe₂O₃-ZrO₂ catalysts, Au on the crystallized metal oxide was more active compared to the one on amorphous metal oxide for WGS [33]; therefore, it is believed that the site for dissociative adsorption of H₂O cannot be formed on amorphous iron oxides. Similarly to CO oxidation using gold catalysts, where the rate-determining step is the activation of oxygen [34], it is important to activate the catalyst for the dissociative adsorption of H₂O. The Au/iron oxide catalyst employed here, thus indicated, can have high catalytic activity for low-temperature WGS only when appropriately reduced and crystallized before the reaction.

The catalytic activity of the samples, which showed high conversion during the 80 min reaction period, gradually decreased after further reaction time. The conversion of 140-CO after 24 h of 80 °C WGS was 0.21, with a decrease in surface area to 48 m² g⁻¹. Possible factors for the catalytic deactivation are the formation of carbon species on the catalyst surface and degradation of the Au particles and of Fe₃O₄. Although there is small number of studies on the deactivation of the gold catalyst during WGS, Kim and Thompson [35] specified the deposition of formate and carbonate species on the catalyst surface to cause the deactivation in Au/CeO₂ catalyst. In our study, deconvolution of C 1s spectra in XPS analysis showed that the area proportion of –COOR:CO₃²⁻ among R-OH (286.0 eV), –COOR:CO₃²⁻ (288.5 eV), adventitious carbon (284.6 eV), and adsorbed CO₂ (undetected) species was 17.9/17.8% before/after 24 h of 80 °C WGS reaction, respectively, for the 140-CO sample. Additionally, the intensity of the –COOR:CO₃²⁻ peak did not significantly change before and after the 24 h reaction. Thus, since there was little difference during the long-term reaction, the formation of carbon species cannot explain the catalyst deactivation. On the other hand, the increase in Au 4f spectrum intensity in XPS analysis was observed after 24 h of reaction (Figure 6(g)), which represents the transformation of cationic Au to the metallic state. As reported in literatures that the cationic Au is responsible for high catalytic activity of the gold catalyst [36,37], the degradation of Au can lead to a decrease of catalytic activity. Considering that Au is responsible for the catalytic activity at low-temperature WGS, which was shown in Figure 1, it is deemed that the deactivation was caused mainly by the loss of cationic Au. Crystallization of Fe₃O₄ and the resulting decrease in surface area can also affect the catalytic activity. However, because the crystallization accompanied the transformation of cationic Au to the metallic state, maintaining lower crystallinity and/or higher surface area would prevent the transformation of cationic Au, leading to the maintenance of high catalytic activity. Hereafter the most important issue for this catalyst is to develop a method for the prevention of structural change during WGS reaction.