Fe/γ-Al₂O₃ and Fe–K/γ-Al₂O₃ as reverse water-gas shift catalysts

Abstract

The reverse water-gas shift (RWGS) reaction was investigated on Fe/γ-Al₂O₃ and Fe–K/γ-Al₂O₃ catalysts at temperatures between 723 K and 753 K and atmospheric pressure. Both materials exhibited fast catalytic CO formation rates and high CO selectivity (>99%). Reaction rates displayed a strong dependence on H₂ partial pressure (reaction orders of 0.58 and 0.54 on Fe/γ-Al₂O₃ and Fe–K/γ-Al₂O₃, respectively), and a weak dependence on CO₂ partial pressure (reaction orders of 0.37 and 0.21, respectively) under nearly equimolar CO₂ : H₂ composition. The catalysts were stable under excess H₂ but deactivated slowly (1–2% h⁻¹ of the overall reaction rate) under an equimolar mixture of CO₂ and H₂. Addition of potassium to the Fe/γ-Al₂O₃ material (Fe/K mass ratio = 1.24) led to a threefold increase in reaction rate, but also doubled the deactivation rate (CO₂ : H₂ = 1 : 1). Gas-switching experiments (CO₂ or H₂ only) and DRIFTS spectra collected in situ showed that stable intermediates formed on Fe–K/Al₂O₃ but not on Fe/Al₂O₃. This suggests, although does not prove, that a redox mechanism is the only reaction pathway on the Fe/Al₂O₃ catalyst, and is the predominant pathway on the Fe–K/Al₂O₃ catalyst. The potassium promoter activates a secondary pathway for CO formation, which may be the so-called associative pathway.

---

1. Introduction

The reverse-water gas shift (RWGS) reaction (eqn (1)) is the reaction of carbon dioxide (CO₂) and hydrogen (H₂) to form carbon monoxide (CO) and water (H₂O). The reaction is endothermic (ΔH^°₂₉₈ = 41.2 kJ mol⁻¹) and the chemical equilibrium favors CO and H₂O as the temperature increases. The water-gas shift (WGS) reaction should be mechanistically related to the RWGS reaction through the principle of microscopic reversibility, and has been investigated in great detail on many catalysts. With the growing importance of limiting anthropogenic CO₂ emissions, the RWGS reaction presents a straightforward alternative for the reduction of CO₂ to CO if an economically viable and carbon-neutral source of H₂ can be developed.

CO₂ + H₂ ⇌ CO + H₂O (1)

The WGS reaction is carried out on an industrial scale in two reactors connected in series; the first reactor is operated at “high temperature” (623–723 K) and uses Fe_xO_y-based catalysts with various promoters (Pt, Cu, Ag, Ba, K, Cr, etc.) and supports (Cr₂O₃, CeO₂–ZrO₂, MnO).¹ Magnetite is believed to be the active phase of iron under high temperature WGS conditions, and, when promoted with chromium, is the customary industrial catalyst for the high-temperature WGS reaction.² Chromium is a structural promoter that helps prevent the iron from sintering.^3,4 The second WGS reactor is operated at “low temperature” (453–523 K) and uses Cu–Zn/Al₂O₃ as the catalyst.⁵

In addition to WGS, supported iron is known to catalyze the RWGS reaction⁶ and a number of other industrially important reactions, including: i) Fischer–Tropsch synthesis,⁷ ii) ammonia synthesis,⁸ iii) ethylbenzene dehydrogenation to styrene,⁹ and iv) selective catalytic reduction of nitrogen oxides (NO_x) with ethanol (EtOH-SCR).¹⁰ The RWGS and WGS reactions are often carried out in conjunction with Fischer–Tropsch synthesis on iron catalysts,¹¹ in which case iron carbide is believed to be the active phase for hydrocarbon production,¹² and iron oxide is the active phase for WGS and RWGS.¹

Promoters are often used with iron catalysts to enhance Fischer–Tropsch or RWGS rates and tune the selectivity to the desired products.^13–17 One known effect of potassium on iron is an increased CO₂ adsorption capacity.¹² Alkali metals are considered electronic promoters, as they can facilitate electron transfer and enhance electrostatic interactions with reacting molecules.^18,19 Chen et al showed that for RWGS on Cu/SiO₂, promotion with potassium leads to the creation of new active sites at the copper–potassium interface.²⁰ Structural promoters such as Cr or Al are often used to prevent iron from sintering.^12,21

Two key issues remain unresolved in regard to the WGS and RWGS reaction mechanism(s): i) distinction between the ‘redox’ and ‘associative’ mechanisms, and ii) determination of the structure of the carbon-containing intermediate in the associative mechanism. The redox and associative models were proposed in 1920 by Armstrong and Hilditch,²² and provided the basis for many subsequent investigations with different catalysts.^9,23–26 Temkin and coworkers proposed that the redox mechanism is active for the WGS reaction on iron catalysts promoted with chromium.^27–29 In this mechanism, the catalyst is first reduced by adsorbed H₂ (or CO in WGS), and is subsequently oxidized by CO₂ (or H₂O in WGS) to complete the redox cycle. A distinguishing feature of the redox mechanism is that the products can be generated in the absence of either reactant (as in a reactant switching-type experiment). The associative mechanism is a Langmuir–Hinshelwood (LMHW) type mechanism, and was suggested by Oki and coworkers to be the dominant mechanism for the WGS reaction on iron oxide catalysts.^30,31 In this mechanism, both reactants must be adsorbed on the catalyst surface at the same time to create products. A number of carbon-containing intermediates have been proposed for the associative mechanisms, the most frequently suggested being a formate species.³² Other suggested intermediates include carbonate,³³ carbonyl,³⁴ and carboxyl³⁵ species.

In this report it is shown that iron supported on alumina (Fe/Al₂O₃) is a highly selective catalyst for the RWGS reaction at temperatures between 723 K and 753 K. It is also shown that the specific rate (per gram of material) is enhanced by addition of potassium. While it is not possible to establish the precise role(s) of potassium on a molecular level, it is shown that addition of potassium i) enhances reaction rates, ii) leads to the formation of stable carbon-containing surface species, and iii) changes the catalyst behavior during H₂/CO₂ gas switching experiments. It is suggested that the simple redox and associative mechanisms are insufficient to explain the observations from the gas-switching experiments, and a more complex reaction model is proposed.

2. Experimental

2.1 Materials

Fe/Al₂O₃ and Fe–K/Al₂O₃ were prepared using the wetness impregnation method. Gamma alumina (γ-Al₂O₃, Alfa Aesar, 99.97%) was added to an aqueous solution of 0.1 M iron nitrate (Fe(NO₃)₃·9H₂O, Aldrich, 99.99%), with or without potassium carbonate (K₂CO₃, Sigma Aldrich, ≥99.0%), and the mixture was stirred with a magnetic stir bar at room temperature for 1 h. The quantities of metal precursor and γ-Al₂O₃ added to each solution are given in Table S1;† the final elemental composition of the samples was determined by ICP-OES (Galbraith Laboratories, TN). The suspension was then heated to 353 K to evaporate water, and the resulting slurry was dried in a static oven at a temperature of 383 K for 24 h in air. The dried samples were subsequently calcined in a Thermolyne furnace in air by heating at a rate of 300 K h⁻¹ to 823 K and holding at that temperature for 4 h. Hematite (Fe₂O₃, Aldrich, 99.99%) and iron foil were used as standards for X-ray absorption near edge structure (XANES) experiments. Magnetite (Fe₃O₄, Aldrich, 99.99%) was used as a catalyst for comparison to the alumina-supported catalysts. Gases used were: CO₂ (Keen, Grade 5.0), H₂ (Matheson, UHP), helium (He, Keen, Grade 5.0), argon (Ar, Keen, Grade 5.0) and D₂ (Cambridge Isotopes, 99.6% gas purity, 99.8% isotope purity). Potassium bromide (KBr, Alfa Aesar, spectroscopy grade) was used as the background in DRIFTS experiments.

2.2 Analytical

X-Ray Diffraction (XRD) patterns of catalyst powders were collected at room temperature on a Philips X'pert diffractometer using Cu Kα radiation (λ = 1.5418 Å). Measurements were taken over the range of 5° < 2θ < 80° with a step size of 0.02° and a count time of 2 s at each step. Physisorption of N₂ was performed using a Micromeritics 3Flex instrument at a temperature of 77 K. The Brunauer–Emmett–Teller (BET) surface areas were calculated from data points at relative pressures (p/p₀) between 0.05 and 0.25. Before adsorption, samples were degassed under vacuum (P < 150 mTorr) for 8 h at a temperature of 573 K. An Auriga 60 high resolution focused ion beam and scanning electron microscope (SEM) was used to collect SEM micrographs and energy-dispersive X-ray (EDX) spectra to provide information regarding sample morphology and elemental composition, respectively. The microscope was operated at an accelerating voltage of 3–10 kV and a current of 10 μA.

2.3 Flow apparatus used for kinetics and gas-switching experiments

Reaction rate and selectivity were evaluated using a packed-bed microreactor operated in down-flow mode. Gas flows through the reactor were controlled by mass flow controllers (Brooks Instrument). Catalyst powders were pressed and sieved to obtain particle sizes within the range of 250–425 μm; the catalyst particles were supported on a quartz wool plug within a quartz tube reactor (7 mm I.D.). The quartz tube was placed inside a ceramic radiant full cylinder heater (Omega, CRFC-26/120-A), and the reaction temperature was controlled by an Omega CN/74000 temperature controller using the input from a thermocouple (Omega, K-type, 1/16 in diameter) placed around the outside of the quartz tube at the middle of the catalyst bed. Gas transfer lines for the effluent stream were heated to a temperature above 373 K and vented to atmospheric pressure. The composition of the effluent stream was analyzed online either by a gas chromatograph (GC, Agilent, 7890A) during continuous flow experiments or a mass spectrometer (MS, Pfeiffer, GSD320) during gas switching experiments. The GC was equipped with both a thermal conductivity detector (TCD) and a flame-ionization detector (FID). The TCD was used to quantify CO₂, CO, and H₂ concentrations, and the FID was used to quantify hydrocarbon concentrations. A Hayesep Q column (Agilent, 2 mm ID × 12 ft) was used in the GC to separate products quantified with the TCD, and a HP-Plot Q column (Agilent, 0.32 mm ID × 30 m) was used to separate products quantified with the FID.

2.4 Measurement of product formation rates

Catalyst samples were pretreated before all experiments in the microreactor by increasing the reactor temperature at a rate of 5 K min⁻¹ to 773 K in a gas flow containing 10 kPa H₂. After being held at 773 K for 2 h, the temperature was lowered to the initial reaction temperature. The total flow rate under all conditions, including pretreatment, was 75 sccm. Helium was used as the balance gas.

Rates of CO formation were calculated assuming differential reactor operation according to eqn (2):

In eqn (2), V is the total volumetric flow rate (L h⁻¹), ΔC_CO is the change in CO concentration (mmol L⁻¹), and m_cat. is the catalyst mass (g). Measured reaction rates are the net rate of the forward and reverse reactions; therefore, the observed rate must be transformed into the reaction rate for the forward reaction by using eqn (3)–(5). The equilibrium constant (K_C) is low (<1) for the RWGS at the temperatures investigated, although the reverse reaction had a negligible contribution to the observed rates because of the low conversion (<10%) under conditions at which the reactor was operated. Note that C_o (eqn (5)) represents the standard state (1 mol L⁻¹) and equals 1 since the reaction is equimolar.

r_obs. = r₊ − r₋ = r₊(1 − η) (3)

Experiments were designed to i) determine reaction rates in excess (i.e. non-equimolar) CO₂ or H₂, and ii) determine kinetic parameters. In the first case, CO₂ and H₂ were fed to the catalyst—4.5% Fe/Al₂O₃ (43 mg) or 4.2% Fe–3.4% K/Al₂O₃ (23 mg)—with the reactor temperature held at 753 K. The initial partial pressures of both CO₂ and H₂ were 15 kPa. After a break-in period of 15 h, the partial pressure of CO₂ was increased to 60 kPa, while the partial pressure of H₂ was held at 15 kPa. After another period of 3 h, the partial pressure of CO₂ was decreased to 15 kPa and the partial pressure of H₂ was increased to 60 kPa. Finally, both partial pressures were returned to 15 kPa. CO₂ conversion was quantified under the same conditions on a sample of magnetite (100 mg) and γ-Al₂O₃ (98 mg), but only with CO₂ and H₂ partial pressures of 15 kPa.

For the second case, kinetic parameters were determined with near equimolar concentrations of CO₂ and H₂ on both 4.5% Fe/Al₂O₃ (45 mg) and 4.2% Fe–3.4% K/Al₂O₃ (22 mg), and under large H₂ excess on both 4.5% Fe/Al₂O₃ (42 mg) and 4.2% Fe–3.4% K/Al₂O₃ (23 mg). With equimolar or near equimolar concentrations of CO₂ and H₂, the reaction was first performed for 15–16 h at a temperature of 753 K with reactant partial pressures of 15 kPa. The temperature was then lowered in 10 K increments to 723 K, with 5–6 GC injections (a period of about 60 min) taken at each temperature. After the period at 723 K, the CO₂ partial pressure was reduced to 10 kPa and increased in 2.5 kPa increments to a final partial pressure of 20 kPa. Finally, the CO₂ partial pressure was returned to 15 kPa and the H₂ partial pressure was lowered to 10 kPa and increased in 2.5 kPa increments. The basic outline of the experiments conducted with excess H₂ was the same as that used for near equimolar reactant concentrations. Reactant partial pressures during the initial period were 90 kPa H₂ and 10 kPa CO₂. During the variable CO₂ partial pressure period, the H₂ partial pressure was maintained at 85 kPa and the CO₂ partial pressure was varied between 5 kPa and 12.5 kPa in 2.5 kPa increments. To investigate the effect of H₂ partial pressure, the CO₂ partial pressure was kept at 10 kPa and the H₂ partial pressure was varied between 70–90 kPa in 5 kPa increments. At the end of the experiments, and several times throughout, reaction parameters were returned to a condition that had already been tested to determine if deactivation had occurred.

The kinetic isotope effect (KIE) of H₂/D₂ was investigated on 4.5% Fe/Al₂O₃ (42 mg) and 4.2% Fe–3.4% K/Al₂O₃ (24 mg). After pretreatment, the reaction began at a temperature of 753 K with CO₂ and H₂ partial pressures of 15 kPa. After 16 h, the temperature was lowered to 723 K and, after 1.5 h, H₂ in the feed was replaced by D₂.

2.5 Measurement of reaction rates with intermittent CO₂ and H₂ flows

CO formation rates were measured while alternating between CO₂ and H₂ gas flows. Catalysts were pretreated as described in section 2.4, with a minor difference being that the gas flow rates were 36 sccm He and 4 sccm H₂. After pretreatment, the H₂ flow was stopped and was replaced by 4 sccm of CO₂. After 20 min, CO₂ in the gas stream was replaced by H₂. This sequence was repeated three times. The reactor was then purged with He for 20 min before CO₂ was readmitted to the gas stream. After 20 min, the reactor was again purged with He before H₂ was readmitted to the gas stream. All sequences with a given gas composition lasted for 20 min, and the temperature of the reactor was held at 773 K throughout the duration of the gas switching portion of the experiment. All gas switches were performed by simultaneously turning off the mass flow controller of the reactant gas flowing into the reactor and turning on the mass flow controller of the other reactant. The hydrodynamic behavior of the gas flow during the transient experiments was analysed in a separate experiment by switching the flow from 10% H₂/He to 10% CO₂/1% Ar/He. The response time of the inert Ar (m/z = 40) relative to those of the products formed during the reaction was monitored to ensure that no artifacts were present.

Additional gas-switching experiments involving purge times of varying length with an inert gas (Ar) were carried out on Fe/Al₂O₃ at 753 K. Following a reduction of the catalyst in 10 kPa H₂ for 2 h at 773 K, 15 kPa CO₂ was admitted to the reactor. After 20 min, CO₂ was replaced by 15 kPa H₂ for 20 min. Then, the reactor was purged with Ar for 5 min. This sequence (CO₂ → H₂ → Ar) was repeated several times, but each time the length of the inert purge was increased by 5 min. After the inert purge reached 20 min, the cycle was repeated a final time with a 5 min inert purge to monitor any effects from catalyst deactivation.

2.6 Diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) measurements

Infrared (IR) spectra were collected with a Nicolet Nexus 470 spectrometer equipped with a mercury cadmium telluride (MCT) detector. A Praying Mantis accessory (Harrick Scientific) was used in conjunction with a HVC-VUV environmental chamber (Harrick Scientific) to collect diffuse reflectance spectra. Catalyst powders were held in the chamber on top of a wire mesh screen, and gases were delivered to the chamber by mass flow controllers. Potassium bromide was heated in the chamber to a temperature of 723 K under He flow and used to collect the background spectrum. Potassium bromide was also used to dilute catalyst samples in a mass ratio of 8 : 1. Catalysts were pretreated by heating from room temperature to 773 K at a rate of 5 K min⁻¹ and holding at that temperature for 2 h in a flow of 30 sccm He and 10 sccm H₂. The temperature of the chamber was then lowered to 723 K and the reactant gas in the feed was switched between H₂ and CO₂ two times, in 30 min intervals, for a total of two periods in CO₂ flow. After the final period in CO₂ flow, the chamber was purged with He. The IR spectra presented are the average of 128 scans collected with a resolution of 2 cm⁻¹.

2.7 X-ray absorption near-edge structure (XANES) spectroscopy

XANES spectra were collected at the National Synchrotron Light Source (NSLS) at Brookhaven National Laboratory on beamline X18A. Spectra of iron standards with known oxidation states (hematite and metallic iron) were collected after placing the materials on Kapton tape. In situ experiments were performed using the apparatus previously described by Paredis et al.³⁶ About 15 mg of catalyst was used in all experiments.

In situ experiments were performed after an initial reduction. During reaction, gas flows consisted of either i) an equimolar mixture of CO₂ and H₂, or ii) alternating flows of CO₂ and H₂. For continuous flow experiments, the temperature was increased from room temperature to 823 K under a gas flow consisting of 5 sccm H₂ and 5 sccm He. After a period of time in which the iron was almost completely reduced to Fe²⁺ (see below), He in the feed was replaced by CO₂. For experiments in which the flow was alternated between CO₂ and H₂, the temperature was increased to 773 K under a gas flow of 2 sccm H₂ and 8 sccm He and held at that temperature until the iron was almost completely reduced to Fe²⁺. Then, the catalyst was purged for 15 min with He, and 2 sccm of CO₂ was added to the feed. The reactor was then purged with He for another 15 min before 2 sccm of H₂ was added to the feed.

XANES data were analysed using the Athena extension of IFEFFIT software. All spectra were normalized by adjusting the pre- and post-edge line parameters in Athena so the regression lines passed through the middle of the data in their respective regions. Iron oxidation states and Fe³⁺/Fe_tot ratios were estimated by linear-combination fitting (LCF) analysis, assuming that the collected data are linear combinations of Fe²⁺ and Fe³⁺. The edge energies (E_o) for Fe⁰ and Fe³⁺ were determined from the standard materials to be 7111.9 eV and 7123.5 eV, respectively. These values correspond to the energies with the maximum first derivatives, and provide a linear relation that relates the oxidation state of iron to the edge energy. Then, the edge energies of spectra collected during the in situ measurements were determined and fit to the linear relation to quantify the amounts of Fe²⁺ and Fe³⁺ present in the samples.

3. Results

3.1 Elemental composition and dispersion of iron in supported catalysts

Table 1 summarizes the elemental compositions and surface areas of the catalyst samples. Iron loadings were between 0.9 and 9.1%, and were within 10% of their nominal values (see Table S1†). The surface areas of all supported iron samples were 60–80 m² g⁻¹, with most samples having a surface area slightly below that of γ-Al₂O₃ (a representative adsorption isotherm is shown in Fig. S1†). In general, increased potassium loading led to a reduction in surface area by 10–20%. Powder XRD patterns (see Fig. S2†) showed no reflections indicating the presence of bulk iron or iron oxides in the samples. Elemental mapping images (see Fig. S3 and S4†) revealed a uniform distribution of iron and potassium, consistent with the absence of bulk iron diffraction peaks in the XRD patterns.

Table 1 Iron and potassium compositions, and surface areas, of catalyst samples

Material	Fe^a (wt%)	K^a (wt%)	Surface area^b (m^2 g^−1)
a Determined from elemental analysis (ICP-OES, Galbraith Laboratories). b Brauner–Emmett–Teller (BET) surface area from N2 adsorption isotherm.
1Fe 3K	0.9	3.5	69.7
4Fe	4.5	—	79.0
10Fe	9.1	—	76.1
8Fe 3K	7.7	3.4	65.1
5Fe 1K	6.7	2.3	68.3
4Fe 4K	4.2	3.4	60.5
γ-Al2O3	—	—	77.4

---

3.2 Reaction rates and stability of Fe/Al₂O₃, Fe–K/Al₂O₃, and bulk iron oxide

Fig. 1 shows catalytic CO formation rates on Fe/Al₂O₃, Fe–K/Al₂O₃, and bulk iron oxide. Bare γ-Al₂O₃ catalyzed CO formation with a rate of 3 mmol h⁻¹ g⁻¹, which was lower than the rates on all other samples by at least an order of magnitude (data not shown). Bulk iron oxide was loaded as magnetite (Fe₃O₄), and, after a short induction period, produced CO at a steady rate of 30 mmol h⁻¹ g⁻¹, which was the lowest specific rate on any of the iron-containing materials. The selectivity to CO on all of the materials under equimolar CO₂ and H₂ was always greater than 99%, methane (CH₄) being the only minor side product. 4.5% Fe/Al₂O₃ catalyzed CO formation with an initial rate of 50 mmol h⁻¹ g⁻¹, higher than the rate on bulk iron oxide, but deactivated at a nearly linear rate of 0.48 mmol h⁻¹ g⁻¹ per h. The initial rate of CO formation on 4.2% Fe–3.4% K/Al₂O₃ was 140 mmol h⁻¹ g⁻¹, which is much higher than the rate on both bulk iron oxide and 4.5% Fe/Al₂O₃, but deactivation occurred at an initial rate of 3.96 mmol h⁻¹ g⁻¹ per h before gradually decreasing to 1.2 mmol h⁻¹ g⁻¹ per h.

Fig. 1 CO formation rates on bulk Fe oxide (loaded as magnetite), 4.5% Fe/Al₂O₃, and 4.2% Fe–3.4% K/Al₂O₃ at partial pressures of CO₂ and H₂ indicated in the legend. Other reaction conditions: T = 753 K, F_tot. = 75 sccm.

After an initial 800 min break-in period at a temperature of 753 K, both supported catalysts operated without further deactivation when the temperature was lowered by 30 K under equimolar CO₂ and H₂ partial pressures (see Fig. S5†). This was determined by measuring the CO formation rate at the very end of the experiment under the same conditions used during the break-in period. Rates collected in the middle of the experiment, during which gas flow rates and temperatures were changed (lowered), were used to determine reaction orders and activation energies. Because the CO formation rate measured after the series of gas flow rate and temperature changes was the same as before the changes, the kinetic parameters extracted from the rate measurements were not corrupted by deactivation.

As shown in Fig. 1, rates of CO formation increased on both catalysts when the gas composition was changed from an equimolar CO₂ : H₂ mixture to 60 kPa CO₂ and 15 kPa H₂. The rate of deactivation on 4.5% Fe/Al₂O₃ increased to 2.6 mmol h⁻¹ g⁻¹ per h but remained linear. The 4.2% Fe–3.4% K/Al₂O₃ catalyst also continued to deactivate under excess CO₂, but showed an exponential deactivation profile. When the gas composition was changed to 15 kPa CO₂ and 60 kPa H₂, the CO formation rate increased on both catalysts. The rate on 4.5% Fe/Al₂O₃ increased steadily with time on stream, whereas the rate on 4.2% Fe–3.4% K/Al₂O₃ first increased rapidly to a maximum of 222 mmol g⁻¹ h⁻¹, exhibited stable activity for approximately 80 min, and then began to slowly deactivate at a rate of 1.5 mmol h⁻¹ g⁻¹ per h. The selectivity to CO decreased to 95% under excess H₂ on the 4.5% Fe/Al₂O₃ catalyst (CH₄ is the main side product). On 4.2% Fe–3.4% K/Al₂O₃, the selectivity to CO remained greater than 99% under excess H₂. Returning the gas composition to 15 kPa CO₂ and 15 kPa H₂ resulted in an initial rate of 116.8 mmol g⁻¹ h⁻¹ on Fe–K/Al₂O₃ (compared to 103.6 mmol g⁻¹ h⁻¹ observed at the end of the first period of flowing this gas composition). The rate decreased to 102.5 mmol g⁻¹ h⁻¹ over the course of an additional hour. On Fe/Al₂O₃, returning the gas composition to 15 kPa of CO₂ and 15 kPa of H₂ resulted in a rate of 45.2 mmol g⁻¹ h⁻¹ (compared to 43.7 mmol g⁻¹ h⁻¹ observed at the end of the first period of flowing this gas composition), and the catalyst showed no deactivation over the course of an additional hour.

3.3 Rate orders, kinetic isotope effect, and CO₂/H₂ switching experiments

After the initial break-in period, reaction rates on both catalysts were stable at a temperature of 723 K. This allowed for the determination of kinetic parameters without having to model deactivation profiles (Arrhenius plots are shown in Fig. S6†). Table 2 summarizes these data at near equimolar CO₂ and H₂, and in large H₂ excess. Under near equimolar CO₂ and H₂ composition, the reaction order with respect to H₂ was nearly the same on both catalysts (0.58 and 0.54 on Fe/Al₂O₃ and Fe–K/Al₂O₃, respectively). In contrast, the reaction order with respect to CO₂ on 4.2% Fe–3.4% K/Al₂O₃ was nearly half the order of that on 4.5% Fe/Al₂O₃ (0.37 and 0.21 on Fe/Al₂O₃ and Fe–K/Al₂O₃, respectively). Under excess H₂, the reaction rate on Fe/Al₂O₃ was nearly first order with respect to CO₂ and was independent of H₂ pressure. The activation energy increased by 6 kJ mol⁻¹ with respect to the value determined at equimolar concentrations of CO₂ and H₂. In contrast, the rate on Fe–K/Al₂O₃ under excess H₂ depended on the concentrations of both reactants (CO₂ reaction order = 0.65, H₂ reaction order = 0.53), and the activation energy of 23 kJ mol⁻¹ was significantly lower than the value of 69 kJ mol⁻¹ determined under equimolar CO₂ and H₂.

Table 2 Activation energies (E_meas.) and reaction orders with respect to CO₂ and H₂

Catalyst	E meas. (kJ mol^−1)	Order in CO2	Order in H2
Equimolar CO2 and H2
4.5% Fe/Al2O3	46	0.37	0.58
4.2% Fe–3.4% K/Al2O3	69	0.21	0.54
Excess H2
4.5% Fe/Al2O3	52	0.87	0.01
4.2% Fe–3.4% K/Al2O3	23	0.65	0.53

---

The reaction rate on 4.5% Fe/Al₂O₃ was considerably higher under CO₂/D₂ flow compared to CO₂/H₂ flow, indicating the occurrence of an inverse KIE (the average value of r_H/r_D was ∼0.65, see Fig. S7†). In contrast, the reaction rates on 4.2% Fe–3.4% K/Al₂O₃ under CO₂/D₂ flow and CO₂/H₂ flow were nearly identical, with an average r_H/r_D value of 1.03.

CO and H₂O were the main products formed during gas-switching experiments (top panel in Fig. 2). On 4.5% Fe/Al₂O₃, CO was formed only when switching from H₂ to CO₂, whereas H₂O was formed when switching from H₂ to CO₂ and when switching from CO₂ to H₂. However, when the catalyst was purged with He before switching from H₂ to CO₂, water was not formed, even though CO was produced. The potassium promoted catalyst showed different properties (bottom panel in Fig. 2). First, CO was produced when switching from CO₂ to H₂. Additionally, after flowing H₂ and purging the reactor with He, water was produced upon admission of CO₂. Concurrent with water formation, H₂ was observed as determined from the mass spectrometer signal at m/z = 2 (see Fig. S8†).

Fig. 2 Ion currents at m/z = 18 (H₂O) and 28 (CO) during H₂/CO₂ switching experiments on 4.5% Fe/Al₂O₃ (top) and 4.2% Fe–3.4% K/Al₂O₃ (bottom). Arrows with a label indicate a change in gas composition to the indicated gas. The catalysts were reduced in flowing H₂ for 2 h before the first admission of CO₂. Reaction conditions: T = 773 K, F_He = 36 sccm, F_H2 or F_CO2 = 4 sccm.

The hydrodynamic behavior of the system was monitored by switching the gas flow from 10% H₂/He to 10% CO₂/1% Ar/He (see Fig. S9†). The transient response curve of Ar (m/z = 40) appeared much faster compared to response curves of CO₂ (m/z = 44, 28) and CO (m/z = 28), indicating that the hydrodynamic behavior of the gas flow in the system did not obscure our ability to accurately detect a kinetic response upon the gas switch.³⁷ It should also be noted that GC data collected during the gas-switching experiments verified the observations seen with the MS, although only the MS data are presented because of the higher time resolution. The CO₂ contribution to the m/z = 28 signal was accounted for in order to identify the production of CO.

3.4 DRIFTS

IR spectra of the 9.1% Fe/Al₂O₃ and 7.7% Fe–3.4% K/Al₂O₃ catalysts after pretreatment in H₂ showed only weak absorption bands (Fig. 3). In the spectra of both materials, a small band at 3550 cm⁻¹ was visible, and on 7.7% Fe–3.4% K/Al₂O₃, additional minor bands at 1379 cm⁻¹ and 1538 cm⁻¹ were also observed. The overall reflectance of the reduced materials increased dramatically after admission of CO₂, which is consistent with the change in color of the materials from dark grey to orange. The spectrum of 9.1% Fe/Al₂O₃ in Fig. 3 shows bands at 3550–3750 cm⁻¹ from gas phase CO₂, and minor inflections at 1308 cm⁻¹ and 1595 cm⁻¹. On 7.7% Fe–3.4% K/Al₂O₃, several intense bands formed at 1343 cm⁻¹, 1568 cm⁻¹, 2613 cm⁻¹, and 2904 cm⁻¹ after admission of CO₂ (the band at 1343 cm⁻¹ has a shoulder on the high energy side). The changes in intensity of the bands during the course of a 30 min purge with inert gas were small (see Fig. S10†). The IR spectra were collected at 723 K, since results from the packed-bed reactor experiments indicated that the materials are stable at this temperature (see Fig. S5†). Additionally, note that the actual bed temperature in commercial in situ spectroscopic environmental chambers, such as the one used in this work, is below the set point.^38,39 Thus, it is highly unlikely that there were any changes to the materials caused by deactivation during the course of the DRIFTS experiments.

Fig. 3 Diffuse reflectance IR spectra collected in situ of Fe/Al₂O₃ (top) and Fe–K/Al₂O₃ (bottom). The catalysts were pretreated at a temperature of 723 K for 2 h in flowing H₂, exposed to flowing CO₂ for 30 min, and reduced again in flowing H₂.

3.5 XANES spectroscopy

XANES spectra were acquired during H₂/CO₂ gas switching experiments on 4.5% Fe/Al₂O₃ and 4.2% Fe–3.4% K/Al₂O₃ (see Fig. S11 and S12†). Before pretreatment, the positions of the absorption edge energy and pre-edge energy were 7123 eV and 7114.5 eV, respectively. During heating and flow of H₂, the absorption edge energy shifted to 7119.5 eV by the time a temperature of ∼773 K was reached, and remained at that energy for the duration of the period in H₂ flow. When H₂ was replaced by CO₂, the position of the absorption edge energy shifted immediately to 7122.2 eV. The fraction of Fe³⁺ in both catalysts as determined from LCF analysis was ∼0.85 during H₂ pretreatment (up to a temperature of nearly 773 K). At 773 K, no Fe³⁺ was observed within the detection limits of the technique. When CO₂ was admitted, the fraction of Fe³⁺ increased rapidly to ∼0.65.

XANES spectra were also collected during a continuous equimolar flow of CO₂/H₂ (see Fig. S13 and S14†). For both catalysts, the fraction of Fe³⁺ was ∼0 after the pretreatment. During equimolar CO₂/H₂ flow on 4.2% Fe–3.4% K/Al₂O₃, the fraction of Fe³⁺ increased with time on stream for the first 25 minutes until a steady value of 0.08 was reached. A different behavior was noted with 4.5% Fe/Al₂O₃; with this catalyst, the position of the absorption edge energy did not change upon admission of CO₂, and the fraction of Fe³⁺ remained ∼0 during the period of feeding CO₂ and H₂.

Table 3 summarizes the average fraction of Fe³⁺ present in both catalysts during different gas flows as calculated from the XANES data. The values indicate that the iron is present as mostly Fe²⁺ under flow of H₂ and mostly Fe³⁺ under flow of CO₂. Under an equimolar flow of CO₂ and H₂, the majority of Fe was in the 2+ oxidation state on both catalysts.

Table 3 Fraction of Fe³⁺ in 4.5% Fe/Al₂O₃ and 4.2% Fe–3.4% K/Al₂O₃ catalysts during in situ XANES measurements

Gas flow	Fe^3+/Fetot. for Fe/Al2O3	Fe^3+/Fetot. for Fe–K/Al2O3
50% CO2 + 50% H2	0.01	0.08
20% H2/He	0	0.02
20% CO2/He	0.61	0.65
He	0.59	0.65

---

3.6 Summary of experimental results

Table 4 compares key properties and results from experiments carried out with Fe/Al₂O₃ and Fe–K/Al₂O₃, and clearly shows that major differences exist between the two materials for nearly every experiment. Most significant are the inverse KIE observed on Fe/Al₂O₃ but no KIE observed on Fe–K/Al₂O₃, the observation of stable surface intermediates on Fe–K/Al₂O₃ but not on Fe/Al₂O₃, differences in kinetic parameters, and the production of H₂O on Fe/Al₂O₃ during the gas switch from H₂ to CO₂, but the production of CO and H₂O on Fe–K/Al₂O₃ during the same gas switch.

Table 4 Summary of experimental results on 4.5% Fe/Al₂O₃ and 4.2% Fe–3.4% K/Al₂O₃ catalysts

Experiment	Fe/Al2O3	Fe–K/Al2O3
KIE (rH/rD)	0.65	1.03
Gas-switching (CO2 → H2)	CO and H2O produced	CO and H2O produced
Gas-switching (H2 → CO2)	H2O produced	CO and H2O produced
DRIFTS	No intermediate	Stable intermediate(s)
CO2 reaction order (∼equimolar inlet)	0.37	0.21
H2 reaction order (∼equimolar inlet)	0.58	0.54
E meas. (kJ mol^−1)	46	69

---

4. Discussion

4.1 Stable surface species on Fe–K/Al₂O₃

Experiments in which gas flows were alternated between H₂ and CO₂ on 4.2% Fe–3.4% K/Al₂O₃ (Fig. 2) indicate that stable surface species form when the reduced catalyst is contacted with CO₂. After reduction in H₂ and upon admission of CO₂, CO and H₂O were produced. However, CO was also produced when H₂ was readmitted, even after an intermediate purge with He. Consequently, a stable carbon-containing intermediate must form during the period in CO₂ flow; the intermediate does not desorb or react in He, and H₂ is required to decompose it and form products. It is possible that the CO released during the switch from CO₂ to H₂ on Fe–K/Al₂O₃ could result from the preferential adsorption of H₂ on iron, which causes desorption of CO from iron. However, no evidence was observed of adsorbed CO on Fe–K/Al₂O₃ or Fe–K/Al₂O₃ in the DRIFTS spectra. CO typically shows strong IR absorption bands around 2100 cm⁻¹ and 1830–1880 cm⁻¹,⁴⁰ but none of these bands were observed. The bond dissociation energies for Fe–CO complexes are in the range of 145–274 kJ mol^−1,41 while the bond dissociation energy for Fe–H is 162 kJ mol⁻¹.⁴² The difference between these bond dissociation energies indicates that the Fe–CO bond tends to be stronger than the Fe–H bond, and therefore CO will not likely desorb because of the introduction of H₂. In contrast to the potassium promoted sample, the 4.5% Fe/Al₂O₃ catalyst did not release CO upon switching gas flows from CO₂ to H₂, suggesting that stable intermediates do not form on this material.

DRIFTS spectra verified that stable intermediates form only on the catalyst containing potassium. After pretreatment in H₂, the maximum intensity across the IR spectrum decreased by ∼50% on both materials (Fig. 3), a change consistent with the change in sample color from orange to grey and the reduction of Fe³⁺ to Fe²⁺ (evidence for the reduction of Fe³⁺ was also observed using XANES, section 4.4). Although the Fe/Al₂O₃ catalyst was oxidized by CO₂, as indicated by the overall increase in reflectance of the material and its change in color from grey to orange, only weak absorption bands were observed around 1308 cm⁻¹ and 1595 cm⁻¹. In contrast, when Fe–K/Al₂O₃ was contacted with CO₂, intense absorption bands formed that disappeared very slowly when CO₂ was removed from the gas stream (see Fig. S10†). The formation of intense absorption bands on Fe–K/Al₂O₃ (but not on Fe/Al₂O₃) is in agreement with the results from the gas-switching experiments and confirms the presence of a stable, carbon-containing surface species on Fe–K/Al₂O₃.

The most intense IR bands on Fe–K/Al₂O₃ were observed at 1343 cm⁻¹ and 1568 cm⁻¹. A smaller band at 2904 cm⁻¹ is likely from the νCH vibration of a formate. The intense bands at 1300–1600 cm⁻¹ appear to be the superposition of bands from several species. When the catalyst was purged with He, two bands at 1651 cm⁻¹ and 1292 cm⁻¹ (identified from a difference plot of the spectra, see Fig. S10†) slowly disappear, while the majority of the absorbing species remain unchanged. A shoulder at ∼1380 cm⁻¹ is also visible, and might be paired with the band at 1550 cm⁻¹, both of which did not completely disappear after re-reduction in H₂. It is likely that the remainder of the band intensity comes from another species, with bands at 1343 cm⁻¹ and 1568 cm⁻¹.

This analysis indicates that there are at least three surface species that form on Fe–K/Al₂O₃, one of which is a formate. The formate ion exhibits characteristic IR vibrations at 1555 cm⁻¹ (asymm ν_CO), 1376 cm⁻¹ (δ_CH), and 1348 cm⁻¹ (symm ν_CO) on α-Fe₂O₃.⁴³ IR bands were observed at 1568 cm⁻¹ and 1380 cm⁻¹ here and, together with the band at 2904 cm⁻¹, strongly suggest that formate is present on the catalyst surface. Other possible surface species that can form during exposure to CO₂/H₂ include bicarbonate, carboxylate, and carbonate. Carboxylates show IR vibrations at 1560 cm⁻¹ and 1340 cm⁻¹ on iron-oxide materials.⁴³ The bands we observed at 1550 cm⁻¹ and 1343 cm⁻¹ agree well with these previous assignments and indicate that carboxylates may also form on the Fe–K/Al₂O₃ material. This is also expected since this sample is basic. The bands at 1651 cm⁻¹ and 1292 cm⁻¹, which disappear during the He purge, could potentially be assigned to bicarbonate (1655–1615 cm⁻¹ [asym ν_CO], 1400–1370 cm⁻¹ [symm ν_CO], 1300 cm⁻¹ [δ_OH]), carboxylate (1660–1560 cm⁻¹[ν_CO]), or bidentate carbonate (1730–1660 cm⁻¹ and 1270–1230 cm⁻¹ on Al₂O₃) species, based on assignments from the literature.⁴³

IR bands in the 3600–3800 cm⁻¹ region of the spectra, which are associated with surface hydroxyl (–OH) groups, were not detected for the catalysts (see Fig. 3). This is peculiar, as typically these bands appear very strongly in IR spectra. For instance, IR spectra of bare Al₂O₃ did reveal IR bands associated with –OH groups (spectra not shown). Thus, the absence of detectable –OH groups on the catalyst samples is not an artifact, and suggests that the incorporation of iron and potassium into the sample has an effect on these groups.

4.2 Redox and associative reaction pathways on Fe/Al₂O₃ and Fe–K/Al₂O₃

The associative pathway has often been proposed as the dominant mechanism for WGS.^44–48 However, in a recent review Burch et al.⁴⁹ concluded that the associative pathway typically accounts for less than 10–15% of the overall WGS reaction rate^50–52 and that formates are often only spectator species.^38,53 The authors found that the associative pathway⁴⁹ can be dominant on low-activity materials such as MgO (ref. 44) and 0.2% Rh/CeO₂,⁴⁵ but that the vast majority of investigations in which an associative pathway was said to be dominant lacked the quantitative data necessary to validate the claim.^46–48 Steady-state isotopic transient kinetic analysis (SSITKA) is often coupled with DRIFTS and MS to elucidate reaction mechanisms and differentiate between active reaction intermediates and spectator species, and has been applied several times to the WGS and RWGS reactions. In a study of the RWGS reaction on 10% Cu/SiO₂, Yang et al used ¹²C/¹³C isotopic transient analysis with MS and IR to simultaneously measure the site coverage and residence time of adsorbed formate species.⁵⁴ Their results showed that the formate removal rate was two orders of magnitude greater than the catalytic RWGS reaction rate, and thus the reaction has no influence on the formate surface coverage. Using SSITKA–DRIFTS–MS, Burch and co-workers showed that surface carbonates were exchanged significantly faster than formates on 2% Pt/CeO₂, and thus formates observed by IR were not actually a major reaction intermediate.^35,55 Finally, in a study of the WGS reaction on 2% Pt/CeO₂, Kalamaras et al. used SSITKA–DRIFTS and SSITKA–MS to show that formates present on the catalyst could not be considered important reaction intermediates, and proposed that a redox mechanism is dominant relative to the associative mechanism.⁵⁶

Gas-switching experiments in which flows of H₂ and CO₂ were alternated (Fig. 2) were used here to distinguish and quantify contributions from redox and associative pathways.⁵⁷ CO formation, when the reduced forms of both Fe/Al₂O₃ and Fe–K/Al₂O₃ catalysts are contacted with CO₂, even after the reduced catalysts were purged with He to ensure the absence of H₂, is evidence of a redox pathway. In the simplest form of the redox mechanism, gas-phase CO₂ adsorbs on a reduced site to form CO and an oxidized site (eqn (6)), which can then be re-reduced by gas phase H₂ to reform the reduced site (eqn (7)).

CO₂ (g) + s_red. → CO (g) + O·s (6)

H₂ + O·s → H₂O (g) + s_red. (7)

During the gas-switching experiments, H₂O was produced during periods of only CO₂ or only H₂ flow. This differs from what is expected in the traditional redox cycle (eqn (6) and (7)), in which H₂O is only produced during the H₂ feeding period. Table 5 presents estimated initial rates of CO production on Fe–K/Al₂O₃ during each segment of the gas-switching experiment. The rate after the switch from H₂ to CO₂ can tentatively be attributed to the rate from a redox reaction pathway, while the rate after the switch from CO₂ to H₂ can tentatively be attributed to the rate from an associative reaction pathway. The rates were calculated from the initial slopes of the concentration vs. time data in Fig. 2, essentially modeling the system as a batch reactor (eqn (8)). It is observed from Table 5 that the rate after the switch from CO₂ to H₂ decreased on Fe/Al₂O₃ with each cycle and increased on Fe–K/Al₂O₃ with each cycle (even following the He purge).

Table 5 Estimated initial rates of CO production after gas switches from H₂ to CO₂ and from CO₂ to H₂ during gas-switching experiments on 4.5% Fe/Al₂O₃ and 4.2% Fe–3.4% K/Al₂O₃

Material	Period	Rate after H2 → CO2 gas switch (μmol L^−1 s^−1 gcat.^−1)	Rate after CO2 → H2 gas switch (μmol L^−1 s^−1 gcat.^−1)	(H2 → CO2 rate)/CO2 → H2 rate) ratio
4.5% Fe/Al2O3	1st CO2	1.48	—	—
	2nd CO2	1.24	—	—
	3rd CO2	1.15	—	—
	4th CO2 (after He purge)	0.47	—	—
4.2% Fe–3.4% K/Al2O3	1st CO2	0.43	0.28	1.53
	2nd CO2	1.07	0.26	4.19
	3rd CO2	1.48	0.26	5.63
	4th CO2 (after He purge)	2.31	0.11	21.5

---

The presence of a stable, carbon-containing surface intermediate on Fe–K/Al₂O₃, (see section 4.1) may be evidence of a concurrent associative pathway. Apparently, potassium allows for a new reaction pathway to CO that involves a stable intermediate. In the associative pathway, described generally by eqn (9), CO₂ and H₂ adsorb on the catalyst surface to form a carbon-containing intermediate (i.e. formate, carbonate, or bicarbonate), which then decomposes in the presence of H₂ to form CO and H₂O.

CO₂ (g) + H₂ (g) → CO₂·s + 2H·s → CO·s + H₂O·s (9)

In summary, CO formed upon switching from H₂ to CO₂ is evidence supporting the redox mechanism, while the CO formed upon switching from CO₂ to H₂ is evidence in support of the associative mechanism. Both the redox and associative reaction pathways then appear to contribute to the overall RWGS rate on Fe–K/Al₂O₃, while only the redox pathway is active on Fe/Al₂O₃. The ratios of the rates during CO₂ flow to H₂ flow for Fe–K/Al₂O₃ (Table 5) indicate that the redox pathway is the dominant contributor to the overall reaction rate.

4.3 Kinetics parameters and formulation of a kinetic model on Fe/Al₂O₃

There are few detailed kinetic investigations of the RWGS reaction on Fe-based catalysts. Osaki et al reported an activation energy of 78.2 kJ mol⁻¹ on 20% Fe/Al₂O₃, determined in the temperature range of 500–800 K.⁵⁸ This value is significantly higher than the 46 kJ mol⁻¹ observed here on 4.5% Fe/Al₂O₃. The reaction orders of 0.37 for CO₂ and 0.58 for H₂ obtained on Fe/Al₂O₃ under 15 kPa CO₂ and 15 kPa H₂ (Table 2) appear to follow the power-law rate relation described by eqn (10); that is, reaction orders are not independent. Additionally, under excess H₂ conditions the reaction orders of 0.87 for CO₂ and 0.01 for H₂ also follow this relation.

r = k_app.[CO₂]ⁿ[H₂]¹⁻ⁿ (10)

A similar relation between the reaction orders was also observed by Ginés et al.⁵⁹ on a CuO/ZnO/Al₂O₃ catalyst for P_H2/P_CO2 < 3 (CO₂ order ≈ 0.3, H₂ order ≈ 0.8), and by Kim et al.⁵⁷ on Pt/TiO₂ (CO₂ order = 0.831, H₂ order = 0.201) and Pt/Al₂O₃ catalysts (CO₂ order = 0.323, H₂ order = 0.702). The reaction orders of 0.21 for CO₂ and 0.54 for H₂ observed on the Fe–K/Al₂O₃ catalyst (15 kPa CO₂ and 15 kPa H₂, Table 2) do not follow eqn (10). Osaki et al.⁵⁸ observed reaction orders of 1.10 for CO₂ and 0.37 for H₂ on a 20% Fe/Al₂O₃ catalyst, indicating a stronger dependence on CO₂ partial pressure compared to H₂ pressure; this was the opposite from what we observed with our samples. Perhaps the secondary pathway on Fe–K/Al₂O₃ has a lower or zero reaction order for CO₂ that leads to a lower observed reaction order compared to Fe/Al₂O₃ (or the amount of iron affects the reaction orders). Potassium is known to increase the adsorption capacity of CO₂ because of its basicity. Evidence for the higher CO₂ adsorption capacity of Fe–K/Al₂O₃ relative to Fe/Al₂O₃ was seen by its ability to form carbon-containing intermediates during the DRIFTS and gas-switching experiments. The higher CO₂ coverage that results could lower the dependence of the reaction on the gas-phase concentration of CO₂.

As shown in Table 2, the activation energy (E_meas.) for the reaction on Fe–K/Al₂O₃ (69 kJ mol⁻¹) is significantly greater than that on Fe/Al₂O₃ (46 kJ mol⁻¹), even though Fe–K/Al₂O₃ showed higher catalytic rates. This is surprising, as normally reaction rates increase with a decrease in the activation energy. The higher rate on Fe–K/Al₂O₃, despite its greater E_meas., indicates that the pre-exponential factor of the rate determining step is large. According to Transition State Theory, a large pre-exponential factor indicates a small negative (or even positive) entropy of formation of the transition state (ΔS^‡), and is characteristic of monomolecular reactions.⁶⁰ A monomolecular RDS typically involves bond dissociation; in this reaction it would likely involve C–O bond breaking. The lower reaction rates observed on Fe/Al₂O₃, in spite of a lower E_meas., indicate that the RDS is likely bimolecular and possibly involves C–H bond formation. As such, ΔS^‡ will be much more negative for this reaction pathway, resulting in a smaller pre-exponential factor that reduces the rate constant.⁶⁰

We observed an inverse KIE on Fe/Al₂O₃, with a r_H/r_D ratio of ∼0.65; this value is nearly the inverse of the typical r_H/r_D ratio of ∼1.4 for a normal H/D KIE.⁶¹ In general, isotope effects are such that k_H/k_D > 1 and K_H/K_D < 1, and therefore a switch from H₂ to D₂ can affect both of these ratios in opposite directions. This leads to the possibility of observing a normal, inverse, or negligible isotope effect, depending on the relative magnitudes of change for these ratios.⁶² The inverse KIE observed on Fe/Al₂O₃ suggests, then, that the equilibrium isotope effects have a significant influence on the measured overall isotope effect and prevail over normal kinetic isotope effects associated with C–H bond formation. The isotopic substitution from H₂ to D₂ is known to change the equilibrium of certain elementary steps, such as hydrogen adsorption.⁶³ Therefore, the thermodynamic terms of the apparent rate constant, which depend on the relative chemisorption enthalpies of D₂ and H₂ on the catalyst surface, are affected.⁶³ Previous studies have indicated that chemisorption of D₂ is preferred over that of H₂ on Fe catalysts, and thus K_H/K_D is less than 1.⁶⁴ In contrast, switching from H₂ to D₂ on Fe–K/Al₂O₃ had very little effect on the CO formation rate, as the r_H/r_D ratio was observed to be 1.03. The absence of a significant KIE on Fe–K/Al₂O₃ indicates that the RDS does not involve bond breaking or forming with H, which is consistent with the occurrence of a redox mechanism in which CO₂ dissociation is the RDS. The difference between the observed KIEs on the two catalysts implies that the incorporation of potassium alters the RDS of the reaction mechanism.

The gas-switching experiments with CO₂ and H₂ led us to conclude that a redox pathway is active on both Fe/Al₂O₃ and Fe–K/Al₂O₃ (see above). Based on this result, an initial model for the reaction pathway for both catalysts is given in Scheme 1. The mechanism shown is a classical redox pathway that includes steps for CO₂ adsorption, CO desorption, and H₂O desorption.

Scheme 1 Redox reaction pathway for CO formation.

Using CO formation on the surface (step 2) as the rate-determining step (RDS), a rate expression can be derived for CO formation in the gas phase at differential conversion (eqn (11)).

The rate expression in eqn (11) does not show a dependence on the partial pressure of H₂, and thus it is inconsistent with the experimentally determined reaction orders. Additionally the expression incorrectly predicts a negative reaction order with respect to CO₂ at high coverage. To incorporate H₂ into the rate expression, a 6th step can be included to allow for competitive adsorption by H₂, but the resulting rate expression (eqn (12)) still does not show a positive reaction order with respect to H₂. The reaction pathway proposed in Scheme 1 (with and without the inclusion of competitive adsorption by H₂) is thus insufficient to properly model CO formation rates on either catalyst, as it does not agree qualitatively with the kinetics measurements.

Considering the inverse KIE that was observed on Fe/Al₂O₃, a different reaction mechanism can be proposed in which C–H bond formation is the RDS (Scheme 2). The model depicted in Scheme 2 is an associative mechanism involving the reaction of adsorbed CO₂ with dissociated H₂ to form a surface intermediate (step 4), which subsequently desorbs as CO (g) and H₂O (g) (step 5). This reaction scheme also incorporates competitive adsorption by H₂ (step 1) and H₂ dissociation on the surface (step 2).

Scheme 2 Associative reaction pathway for CO formation based on observation of inverse KIE.

Evidence for H₂ dissociation (step 2 in Scheme 2) was observed when H₂/D₂ mixtures were fed to the catalyst in the presence or absence of CO₂ (see Fig. S15†). HD formation occurs quickly (on the same time scale as the chemical conversion of CO₂ to CO), indicating that H₂ dissociation is reversible and not rate limiting. The intermediate formed in step 4 of Scheme 2 should not be observable by in situ IR spectroscopy because, by definition, its formation is rate limiting and it rapidly decomposes (recall that such intermediates were not observed on Fe/Al₂O₃ catalysts). The rate equation for CO formation according to Scheme 2 is presented in eqn (13), assuming that step 4 is the RDS and that conversion levels are low.

Eqn (13) has both CO₂ and H₂ terms in the numerator, agreeing with the experimental results in which positive reaction orders were observed for both of these reactants. The fit of this equation is good for both catalysts (Table 6), with the only exception being that the fitted reaction orders for H₂ differ from the experimentally determined values. For instance, Table 6 shows that the H₂ reaction orders on Fe/Al₂O₃ and Fe–K/Al₂O₃ are 0.58 and 0.54, respectively, while the H₂ reaction orders fitted from the model are 0.50 for both catalysts. Note that the apparent rate constant in the rate expression (eqn (13)) is a product of the elementary rate constant for step 4 (k₄) and the equilibrium constants of steps 1–3 (K₁, K₂, K₃). Equilibrium isotope effects are typically less than 1, and therefore these terms could be the reason for the inverse KIE observed experimentally.⁶² Only K₁ or K₂ can be the cause of the inverse KIE, though, since only these steps involve H₂.

Table 6 Measured and fitted kinetic parameters on 4.5% Fe/Al₂O₃ and 4.2% Fe–3.4% K/Al₂O₃ using the associative reaction pathway illustrated in Scheme 2

Material		E a (kJ mol^−1)	Order in CO2	Order in H2
Fe/Al2O3	Experimental	46	0.37	0.58
	Fitted (step 4 RDS)	46	0.37	0.50
	Fitted (step 5 RDS)	45	0.38	0.59
Fe–K/Al2O3	Experimental	69	0.21	0.54
	Fitted (step 4 RDS)	70	0.22	0.50
	Fitted (step 5 RDS)	71	0.22	0.98

---

Finally, it is also possible that step 5 of Scheme 2 (decomposition of the surface intermediate) is the RDS on Fe–K/Al₂O₃, since stable carbon-containing intermediates were observed on this catalyst during the DRIFTS experiments. In this case, the rate expression shown in eqn (14) is obtained.

It is unlikely that step 5 is the RDS on Fe/Al₂O₃, since no carbon-containing intermediates were observed during the IR experiments; nonetheless, an attempt was made to fit the experimental data to the rate expression in eqn (14) for this scenario. As Table 6 shows, the assumption that step 5 is the RDS results in a better fit with the experimental data for Fe/Al₂O₃, as opposed to the case when step 4 was assumed to be the RDS and the H₂ reaction orders did not match well. The assumption of step 5 as the RDS for the reaction on Fe–K/Al₂O₃ results in poor agreement between the fitted (0.98) and experimentally determined (0.54) reaction orders for H₂. It is clear from the results that the simple redox and associative mechanisms proposed are not sufficient to describe all of the experimental results. This suggests that a more complex mechanism or combination of competing pathways may be occurring, as is often the case in the WGS and RWGS reactions.

5. Conclusions

Packed-bed microreactor studies indicated that incorporation of potassium into Fe/Al₂O₃ results in a significant increase (∼3 times) in CO formation rates. Fe/Al₂O₃ and Fe–K/Al₂O₃ slowly deactivated under excess CO₂, but excess H₂ resulted in stable catalytic rates. Reaction rates depend more strongly on H₂ (orders 0.58 and 0.54 for Fe/Al₂O₃ and Fe–K/Al₂O₃, respectively) compared to CO₂ (orders 0.37 and 0.21 for Fe/Al₂O₃ and Fe–K/Al₂O₃, respectively) under equimolar CO₂ : H₂ composition. Gas-switching experiments on Fe/Al₂O₃ revealed that CO was formed only when switching from H₂ to CO₂, whereas H₂O was formed when switching from H₂ to CO₂ and when switching from CO₂ to H₂. The results of the gas-switching experiments on Fe/Al₂O₃ suggest that a redox mechanism is active, since the order in which the reactants are adsorbed on the surface affects the products that are observed. On Fe–K/Al₂O₃, both CO and H₂O were produced when switching from H₂ to CO₂ and from CO₂ to H₂, which suggests that an associative reaction mechanism may also be occurring. Evidence of carbon-containing surface intermediates on Fe–K/Al₂O₃—supporting the occurrence of an associative mechanism—was obtained using DRIFTS. IR bands associated with formate, and possibly carbonate, bicarbonate, and carboxylate, were observed in the presence of CO₂ and H₂. No such IR bands were observed on the Fe/Al₂O₃ material, in agreement with the results from the gas-switching experiments, in which CO was not produced when switching from CO₂ to H₂. Reaction rates measured under H₂ or D₂ revealed an inverse KIE on Fe/Al₂O₃ (r_H/r_D = 0.65), but no KIE on Fe–K/Al₂O₃ (r_H/r_D = 1.03). The observed inverse KIE suggests that the RDS for the mechanism on Fe/Al₂O₃ involves hydrogen addition, while that on Fe–K/Al₂O₃ does not. Although the experimental evidence suggested that a redox mechanism is the only (on Fe/Al₂O₃) or dominant (on Fe–K/Al₂O₃) mechanism occurring, the rate expression obtained from a proposed redox mechanism (Scheme 1) did not show a dependence on H₂ pressure, that is, it is inconsistent with the experimental results in which a clear dependence on H₂ pressure was observed. The associative model provided an excellent fit to the experimental reaction data, but did not explain other experimental results, which strongly suggested a redox mechanism.

Author contributions

J. A. L. and M. J. W. contributed equally to this report by conducting experiments, analyzing the results, and co-writing the manuscript. N. S. M. assisted in conducting the XANES experiments at Brookhaven National Laboratory and helped with the data analysis. R. F. L. supervised the research and co-wrote the manuscript.

Acknowledgements

The authors thank E. Schreiner and H. Sheng from the University of Delaware for their help in collecting XANES data at Brookhaven National Laboratory. This research was funded by the U.S. Army under grant GTS-S-14-392. The authors would also like to acknowledge the Synchrotron Catalysis Consortium funded by the U.S. Department of Energy under grant DE-SC0012335.

References

1. C. Ratnasamy and J. Wagner, Catal. Rev.: Sci. Eng., 2009, 51, 325–440.
2. A. Hakeem, R. Vasquez, J. Rajendran, M. Li, R. Berger, J. Delgado, F. Kapteijn and M. Makkee, J. Catal., 2014, 313, 34–35.
3. D. Newsome, Catal. Rev.: Sci. Eng., 1980, 21, 275–318.
4. H. Topsoe and M. Boudart, J. Catal., 1973, 31, 346–359.
5. D. Jeong, W. Jang, J. Shim, W. Han, H. Roh, U. H. Jung and W. L. Yoon, Renewable Energy, 2014, 65, 102–107.
6. A. Sun, Z. Qin, S. Chen and J. Wang, J. Mol. Catal. A: Chem., 2004, 210, 189–195.
7. F. Farias, R. Rabelo Neto, M. Baldanza, M. Schmal and F. Fernandes, Braz. J. Chem. Eng., 2011, 28, 495–504.
8. H. Hinrichs and R. Thoma, Process for the production of iron oxide catalysts, US Pat., US 3417031A, 17 Dec. 1968.
9. C. Nederlof, G. Talay, F. Kapteijn and M. Makkee, Appl. Catal., A, 2012, 423–424, 59–68.
10. D. Worch, W. Suprun and R. Glaser, Chem. Pap., 2014, 68, 1228–1239.
11. M. Kumaran Gnanamani, W. Shafer, D. Sparks and B. Davis, Catal. Commun., 2011, 12, 936–939.
12. P. Sai Prasad, J. Bae, K. Jun and K. Lee, Catal. Surv. Asia, 2008, 12, 170–183.
13. D. Bukur, D. Mukesh and S. Patel, Ind. Eng. Chem. Res., 1990, 29, 194–204.
14. J. Dun, E. Gulari and K. Ng, Appl. Catal., 1985, 15, 247–263.
15. G. Maiti, R. Malessa and M. Baerns, Appl. Catal., 1983, 5, 151–170.
16. A. Pour, M. Zare and Y. Zamani, J. Nat. Gas Chem., 2010, 19, 31–34.
17. Y. Yang, H. Ziang, Y. Xu, L. Bai and Y. Li, Appl. Catal., A, 2004, 266, 181–194.
18. D. Miller and M. Moskovits, J. Phys. Chem., 1988, 92, 6081–6085.
19. A. Kotarba, A. Baranski, S. Hodorowicz, J. Sokolowski, A. Szytula and L. Holmlid, Catal. Lett., 2000, 67, 129–134.
20. C. Chen, W. Cheng and S. Lin, Appl. Catal., A, 2003, 238, 55–67.
21. H. Wan, B. Wu, C. Zhang, H. Xiang, Y. Li, B. Xu and F. Yi, Catal. Commun., 2007, 8, 1538–1545.
22. E. Armstrong and T. Hilditch, Proc. R. Soc. London, Ser. A, 1920, 97, 265.
23. C. Chen, W. Cheng and S. Lin, Catal. Lett., 2000, 68, 45–48.
24. C. Chen and W. Cheng, Catal. Lett., 2002, 83, 121–126.
25. L. Wang, M. Khazaneh, D. Widmann and R. Behm, J. Catal., 2013, 302, 20–30.
26. S. Fujita, M. Usui and N. Takezawa, J. Catal., 1992, 134, 220–225.
27. N. Kulkova and M. Temkin, Zh. Fiz. Khim., 1949, 23, 695–698.
28. M. Temkin, in Advances in Catalysis, ed. D. Eley, H. Pines and P. Weisz, Academic Press, New York, 1979, vol. 28.
29. G. Shchibrya, N. Morozov and M. Temkin, Kinet. Catal., 1965, 6, 1057–1059.
30. R. Mezaki and S. Oki, J. Catal., 1973, 30, 488–489.
31. S. Oki, J. Happel, M. Hinatow and Y. Kancko, Catal., Proc. Int. Congr., 5th, 1973, 1, 173–183.
32. J. Ladebeck and J. Wagner, in Handbook of Fuel Cells- Fundamentals, Technology and Applications, ed. W. Vielstich, A. Lamm and H. A. Gasteiger, John Wiley & Sons, Ltd, Chichester, 2003, pp. 190–201.
33. D. Tibiletti, A. Goguet, F. Meunier, J. Breen and R. Burch, Chem. Commun., 2004, 1636–1637.
34. A. Goguet, F. Meunier, D. Tibiletti, J. Breen and R. Burch, J. Phys. Chem. B, 2004, 108, 20240–20246.
35. A. Gokhale, J. Dumesic and M. Mavrikakis, J. Am. Chem. Soc., 2008, 130, 1402–1414.
36. K. Paredis, L. Ono, F. Behafarid, Z. Zhang, J. Yang, A. Frenkel and B. Cuenya, J. Am. Chem. Soc., 2011, 133, 13455–13464.
37. A. M. Efstathiou and X. E. Verykios, Appl. Catal., A, 1997, 151, 109–166.
38. F. C. Jentoft, Adv. Catal., 2009, 52, 129–211.
39. H. Li, M. Rivallan, F. Thibault-Starzyk, A. Travert and F. C. Meunier, Phys. Chem. Chem. Phys., 2013, 15, 7321–7327.
40. E. Kock, M. Kogler, T. Bielz, B. Klotzer and S. Penner, J. Phys. Chem. C, 2013, 117, 17666–17673.
41. C. P. McNary and P. B. Armentrout, Phys. Chem. Chem. Phys., 2014, 16, 26467–26477.
42. R. G. Pearson, Chem. Rev., 1985, 85, 41–49.
43. G. Busca and V. Lorenzelli, J. Mater. Chem., 1982, 7, 89–126.
44. A. Ueno, T. Onishi and K. Tamaru, Trans. Faraday Soc., 1970, 66, 756–763.
45. T. Shido and Y. Iwasawa, J. Catal., 1993, 141, 71–81.
46. G. Jacobs and B. Davis, Appl. Catal., A, 2007, 333, 192–201.
47. R. Leppelt, B. Schumacher, V. Plzak, M. Kinne and R. Behm, J. Catal., 2006, 244, 137–152.
48. C. Kalamaras, G. Olympiou and A. Efstathiou, Catal. Today, 2008, 138, 228–234.
49. R. Burch, A. Goguet and F. Meunier, Appl. Catal., A, 2011, 409–410, 3–12.
50. D. Tibiletti, F. Meunier, A. Goguet, D. Reid, R. Burch, M. Boaro, M. Vicario and A. Trovarelli, J. Catal., 2006, 244, 183–191.
51. F. Meunier, A. Goguet, C. Hardacre, R. Burch and D. Thompsett, J. Catal., 2007, 252, 18–22.
52. F. Meunier, D. Reid, A. Goguet, S. Shekhtman, C. Hardacre, R. Burch, W. Deng and M. Flytzani-Stephanopoulos, J. Catal., 2007, 247, 277–287.
53. C. Kalamaras, P. Panagiotopoulou, D. Kondarides and A. Efstathiou, J. Catal., 2009, 264, 117–129.
54. Y. Yang, C. A. Mims, R. S. Disselkamp, C. H. F. Peden and C. T. Campbel, Top. Catal., 2009, 52, 1440–1447.
55. F. C. Meunier, D. Tibiletti, A. Goguet and R. Burch, Oil Gas Sci. Technol., 2006, 61, 497–502.
56. C. M. Kalamaras, S. Americanou and A. M. Efstathiou, J. Catal., 2011, 279, 287–300.
57. S. Kim, H. Lee and S. Hong, Appl. Catal., A, 2012, 423–424, 100–107.
58. T. Osaki, N. Narita, T. Horiuchi, T. Sugiyama, H. Masuda and K. Suzuki, J. Mol. Catal. A: Chem., 1997, 125, 63–71.
59. M. Ginés, A. Marchi and C. Apesteguia, Appl. Catal., A, 1997, 154, 155–171.
60. I. Chorkendorff and J. Niemantsverdriet, in Concepts of Modern Catalysis and Kinetics, Wiley-VCH, Darmstadt, 2007, 2nd edn.
61. M. Kumaran Gnanamani, G. Jacobs, W. Shafer, D. Sparks and B. Davis, Catal. Lett., 2011, 141, 1420–1428.
62. M. Ojeda, A. Li, R. Nabar, A. Nilekar, M. Mavrikakis and E. Iglesia, J. Phys. Chem., 2010, 114, 19761–19770.
63. S. Krishnamoorthy, M. Tu, M. P. Ojeda, D. Pinna and E. Iglesia, J. Catal., 2002, 211, 422–433.
64. P. Biloen, H. Helle and W. Sachtler, J. Catal., 1979, 44, 439.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5cy02111a

---