Investigation of the dynamic active center for CO₂ hydrogenation to light olefins over Na-modified iron catalysts

Abstract

Alkali-modified iron catalysts often undergo phase transitions during CO₂ hydrogenation, leading to mixed products and lower reactivity. Further elucidating their active sites, concurrently suppressing by-products, and maintaining stability remain significant challenges. This work initially explores the reactivity of pure Fe₅C₂ and Fe₃O₄ nano-catalysts, followed by the regeneration of deactivated catalysts by syngas-induced carburization to evaluate the phase-dependent catalytic behavior. Subsequently, Na₂CO₃ dopant was added via solid-phase mixing or wet impregnation to modify the single-phase iron catalysts. The resulting active sites responsible for olefin formation were further investigated, and changes in CO selectivity were explored. The results confirm a particle size–dependent dynamic phase transition between the active Fe₅C₂ phase and the inactive Fe₃O₄ phase. This transition leads to an initial decline in catalytic activity, followed by stabilization toward a pseudo-steady state. The physical mixing of Fe₅C₂ and Na₂CO₃ significantly enhances selectivity for light olefins among hydrocarbons. However, the Na₂CO₃ additive also further promotes the oxidation of Fe₅C₂ to Fe₃O₄, enhancing Reverse Water–Gas Shift activity. Additionally, Na₂CO₃ suppresses surface hydrogen dissociation, thereby preventing further CO conversion into olefins and long-chain products via the Fischer–Tropsch synthesis (FTS) reaction. Together, these effects result in higher CO selectivity during CO₂ hydrogenation over the Na₂CO₃-modified Fe₅C₂ catalyst. Operating at a low gas space velocity (GHSV) drastically reduces CO selectivity by promoting the CO-based FTS reactions; however, the associated high CO₂ conversion and elevated water partial pressure also promote and accelerate the oxidation of Fe₅C₂ to Fe₃O₄. These findings highlight that olefin formation originates at the iron carbide–alkali interface and that catalyst stability is governed by a complex interplay of phase transitions, alkali metal promoter effects, and reaction conditions. Moreover, the observed oxidation acceleration at low GHSV may aid in the rapid screening of stable iron-based catalysts in future studies.

---

1 Introduction

The use of captured CO₂ together with H₂ (obtained from sustainable electricity) to produce e-chemicals is a promising option for decarbonization of the chemical industry.¹ The production of light olefins (ethylene, propylene, and butene), which are important building blocks in the chemical and polymer industries, mostly relies on fossil resources in traditional processes such as naphtha steam cracking, alkane dehydrogenation, and fluid catalytic cracking (FCC).² These traditional processes not only consume large amounts of fossil resources but also emit significant amounts of CO₂.^2,3 If CO₂ can be used as a carbon source to produce olefins and then chemicals and plastics, both severe CO₂ emissions and fossil resource consumption can be avoided. Additionally, CO₂ can also be stored in many plastic commodities with long lifecycles.³ However, due to the high thermodynamic stability of CO₂ and the significant external energy input required for its conversion, the value-added use of CO₂ has historically been limited to specialized industrial processes such as those in the fertilizer and food industries.⁴ With the large-scale rise of renewable energy, especially the development of green hydrogen technologies that integrate solar and wind power with water electrolysis for green hydrogen production, the abundant supply of cheap hydrogen makes the valorization of CO₂ into value-added olefins via hydrogenation increasingly feasible.^1–4

Currently, there are two accessible pathways for the catalytic hydrogenation of CO₂ into olefins. One is the methanol-intermediated pathway, which employs a tandem catalyst system consisting of an oxide catalyst for converting CO₂ to methanol (CTM) and a zeolite component for the subsequent methanol-to-olefins (MTO) reaction.^1,4–8 However, the efficiency of CO₂ conversion remains low, with a significant amount of CO as a by-product. This is primarily due to the mismatch between the optimal reaction conditions for the CTM and MTO reactions.⁸ The iron-based Fischer–Tropsch synthesis (FTS) is another promising technology that can efficiently produce olefins from both CO₂ and CO. Over the iron catalyst containing various iron species (iron oxides/iron carbides), CO₂ hydrogenation is reported to proceed via two steps: the Reverse Water-Gas Shift (RWGS) reaction converting CO₂ to CO, followed by the FTS of CO to hydrocarbons.^{1,4–7,9–12} Specifically, the iron magnetite (Fe₃O₄) component is generally regarded as the active site for CO formation in the RWGS reaction, while Hägg iron carbide (Fe₅C₂) serves as the primary phase catalyzing the FTS reactions.^9,10 Notably, iron phase transformations and surface reconstruction occur during CO₂ hydrogenation. In general, CO₂ and H₂O act as oxidizers that can oxidize iron carbide into iron oxides, while CO, H₂, and hydrocarbon products act as reducing and carburizing agents. As a result, variations in product composition and reactant atmosphere under different reaction conditions can easily induce complex phase changes between iron oxides and iron carbides within the catalyst.^10,11

Alkaline metal dopants (e.g., Na, K), acting as electronic and structural promoters, are commonly added to the iron catalyst to enhance higher olefin yields during the FTS process.^9,12–19 These additives are believed to increase the surface basicity of the catalyst,^9,12,13 influence the formation of iron carbide,^12,13 and decrease the H : C ratio on the iron carbide surface.¹⁴ Consequently, they often modify the catalyst's activity, suppress CH₄ formation, and improve selectivity toward olefins. Numerous studies have investigated the effects of both single and mixed alkali metal promoters in iron catalysts for CO- or CO₂-based FTS reactions, with varying conclusions depending on the catalyst preparation method used.^9,13,15–18 For example, Na and K—introduced as cationic components of the precipitant (carbonate or hydroxide)—are typically incorporated into the precipitated iron catalyst during synthesis. Before CO₂ hydrogenation, the iron phase inside the catalyst was influenced by these alkaline elements during pre-carburizing by CO^17,18 or syngas,^11,15 or reduction by H₂ treatment.^{9,10,12,13,16} It has been reported that these alkaline metal dopants enhance CO conversion and dissociation on the iron surface, while also increasing carbon deposition via the Boudouard reaction, resulting in higher amounts of iron carbide and coke in Na- or K-modified iron catalysts, which leads to increased CO₂ activation.^9,10,13 However, the inhibition of hydrogen dissociation by alkaline dopants can simultaneously reduce the overall CO₂ conversion rate.^12,14 Additionally, some studies report that iron catalysts doped with alkaline metals enhance both the adsorption strength and adsorption capacity of CO₂.^13,19 Nevertheless, since iron carbide and metallic iron catalysts are readily oxidized by CO₂ to iron oxide, increasing CO₂ adsorption may accelerate the phase transformation from iron carbide to iron oxides, thereby also weakening CO₂ conversion.^10,17

So far, several fundamental questions regarding alkaline metal dopant-modified iron catalysts remain unresolved. In particular, there is a lack of detailed research on the active center responsible for CO₂ hydrogenation to light olefin while suppressing the formation of both CO and CH₄. A recent study reported that the active Fe₅C₂–K₂CO₃ interface, produced by the physical mixing of FeC₂O₄ and carbonaceous K with close proximity (mortar mixing with average particle size of 8 µm), can significantly enhance the olefins production during CO₂ hydrogenation, but other methods of integration resulted in either lower conversion of CO₂, or higher selectivity for CH₄ and CO, along with decreased selectivity for light olefins.⁹ However, an in-depth study is still lacking on how the proximity between iron catalysts and alkaline metal dopants affects the catalyst's surface properties and active phase composition, thus ultimately leading to correlated changes in product distribution and catalytic activity.

In this work, pure nano-sized iron carbide (FeC_x) or iron magnetite (FeO_y) catalysts were directly synthesized by thermally treating nano-iron oxide powder under different gas atmospheres. The Na₂CO₃-modified catalysts were obtained by mixing the Na₂CO₃ dopant with either FeC_x or FeO_y at varying degrees of intimacy. These pure iron catalysts (FeC_x and FeO_y) and their modified counterparts (Na-FeC_x and Na-FeO_y) were first tested during CO₂ hydrogenation. After subsequent regeneration using syngas (a mixture of CO and H₂), the reactivities of regenerated catalysts were checked and compared with their initial performance to reveal phase-dependent catalytic behaviors. Simultaneously, based on their different reactivity and selectivity toward light olefins, CO, and CH₄, the specific active phases resulting from varying degrees of proximity between iron carbide and Na₂CO₃ were further examined. To further explore the mechanism, the corresponding physicochemical properties of all catalysts were analyzed using ex situ and in situ techniques, such as X-ray diffraction (XRD), temperature-programmed oxidation (TPO), X-ray photoelectron spectroscopy (XPS), Mössbauer spectroscopy, and high-resolution transmission electron microscopy (STEM). Using this approach, the active sites of alkali metal-doped iron catalysts responsible for olefin production during CO₂ hydrogenation are systematically revealed and verified experimentally.

2 Materials and methods

2.1 Catalyst preparation

The pure iron carbide or iron magnetite catalysts were produced by thermal treatment of a commercial nano-Fe₂O₃ powder (NANOCAT, α-Fe₂O₃, 255 m² g⁻¹) under different gas atmospheres inside a quartz tubular flow reactor (16 mm diameter, 3 mm wall thickness). Typically, 0.3 g of nano-Fe₂O₃ powder was placed in a boat-shaped crucible and placed into the tube reactor. After purging with an Ar flow (500 Nml min⁻¹) for 20 min, the sample was carburized in a syngas atmosphere (3% H₂ and 1.5% CO in Ar, 500 Nml min⁻¹) for 12 h at 300 °C to obtain a batch of pure iron carbide, referred to as FeC_x. After carburization, all carburized samples were in situ cooled down under a flowing Ar (100% Ar, 500 Nml min⁻¹) atmosphere and then passivated with a diluted O₂ flow (2% O₂ in Ar, 500 Nml min⁻¹) at 25 °C for 2 h to prevent severe oxidation and ignition from direct exposure to air. The pure iron magnetite catalysts were prepared by oxidizing the obtained FeC_x with a steam atmosphere (1% H₂O in Ar, 500 Nml min⁻¹) for 12 h at 300 °C, denoted now as FeO_y. When the fresh FeO_y sample was re-carburized again by a syngas atmosphere, the obtained sample will be named Carb-FeO_y. To compare the effects of water and CO₂ on the oxidation of iron carbide, the same amount of FeC_x sample was treated with a CO₂ (20% CO₂ in Ar, 500 Nml min⁻¹) atmosphere at the same temperature and time, and the resulting sample will be labeled as FeO_z-CO₂. The overall iron-based catalysts preparation procedure is illustrated in Scheme 1. The resulting sample powders were then pressed into pellets, crushed, and sieved to obtain granules in the size range of 250–500 µm.

Scheme 1 The pure iron-based catalysts were synthesized by thermally treating nano–iron oxide powder under different gas atmospheres.

The Na-modified iron catalysts with different proximities between Na₂CO₃ and pure iron-based catalysts but with the same composition (n_Fe : n_Na = 1 : 0.02 molar ratio) were prepared as illustrated in Scheme 2. The PM-Na-FeC_x or PM-Na-FeO_y were prepared by powder mixing (PM) by grinding Na₂CO₃ powder (Sigma-Aldrich >99%) with FeC_x or FeO_y powder in a mortar, respectively. The mixtures were then pressed into pellets, crushed, and sieved into granules of 250–500 µm. GM-Na-FeC_x was prepared by granular mixing (GM) of the obtained pure FeC_x granules with Na₂CO₃ particles (250–500 µm). IM-Na-FeC_x was prepared by a single-step incipient wetness impregnation (IWI) of granular FeC_x (250–500 µm) using an aqueous solution of Na₂CO₃ at room temperature. After impregnation, the sample was dried overnight at 110 °C before use.

Scheme 2 Na₂CO₃-modified iron catalysts were prepared by mixing the Na₂CO₃ dopant with either FeC_x or FeO_y at varying degrees of proximity.

2.2 Reactor experiments

Catalytic reaction tests were conducted in a stainless-steel fixed-bed reactor (VINCI Technologies) with an inner diameter of 8.3 mm and a length of 215 mm. Typically, 1 g of granular catalyst was packed into the reactor tube and flushed with an Ar flow (100 Nml min⁻¹) at 30 °C for 20 min at ambient pressure before the reaction. Then, a feed gas mixture with an H₂/CO₂ ratio of 3 was introduced into the reactor at a gas hourly space velocity (GHSV) of 6000 Nml g⁻¹ h⁻¹, and the reactor temperature was slowly increased at a ramping rate of 2 °C min⁻¹. Afterward, the reactor pressure started to build up until the temperature reached 325 °C. The performance of CO₂ hydrogenation was recorded when the reaction pressure was stable at 20 bar. To keep the reaction products completely in the vapor phase without condensation, the temperature of the pipeline between the reactor and cold trap was maintained at about 200 °C during the reaction. All effluent gas passed through a cold trap (25 °C) to remove the water vapor and any heavy hydrocarbons from the products. The selectivity for octane (C₈) and other heavier hydrocarbons (C₈₊) will not be discussed further in this study, because the selectivity for them was found to be less than 1% among all products after an 8 h reaction test. Note proper handling of H₂ is required due to its explosive properties.

The compositions of the gaseous effluents were continuously analyzed by an online gas chromatograph (GC-456, Bruker) equipped with a flame-ionization detector (FID) and a thermal conductivity detector (TCD). The CO₂ conversion and the selectivity for individual hydrocarbon products were calculated according to the equations below.⁸ All calculations were performed using data obtained from online GC analysis. Based on the experimental calculations, the carbon balance was maintained within 100% ± 5%. Any deviations beyond the acceptable experimental error range were addressed by repeating the corresponding reaction conditions. The errors of the datasets and the carbon balance calculation results can be found in the online original data repository.

CO conversion:

CO selectivity:

C_iH_j selectivity among all products:

where refers to the molar flow of CO₂ at the inlet, whereas represent the molar flow of the outlet CO₂, CO, and individual hydrocarbons, respectively. The symbol i is the number of carbon atoms in a given hydrocarbon.

Regarding the spent catalyst after reaction, the reaction process leads to partial oxidation of the catalyst and the formation of a surface oxide layer. After 8 h on stream, all spent catalysts were first flushed with an Ar flow (100 Nml min⁻¹), then cooled to room temperature before being removed from the steel reactor, and quickly transferred to sealed glass vials.

For catalyst regeneration, the spent catalysts were directly loaded into a quartz tubular reactor, purged with Ar flow again, and subsequently carburized under a syngas atmosphere (3% H₂ and 1.5% CO in Ar, 500 Nml min⁻¹) for 12 h at 300 °C; when regenerated catalysts were subjected to ex situ characterization, a surface passivation treatment was applied prior to air exposure, using 1% H₂O in Ar at 300 °C for 12 h, as described in the catalyst preparation section. All catalyst (fresh, spent and regenerated) samples were finally stored in a vacuum glass desiccator to minimize exposure to air and moisture and prevent further oxidation. Samples intended for Mössbauer analysis were transported in sealed vials and stored in containers with silica gel desiccant during shipment. Upon opening, all sample handling and storage were conducted in a glove box to minimize exposure to moisture and atmospheric oxygen prior to Mössbauer measurements.

The activity test of the regenerated (or re-carburized) catalyst was conducted using the same procedure as for the fresh catalysts. Here, the names of all spent, regenerated, and spent-regenerated catalyst samples are labeled and distinguished by adding a prefix such as ‘Spent-’, ‘Regen-’, or ‘Spent-Regen-’ to the names of the fresh catalyst samples, respectively.

Transient experiments with different gas hourly space velocities (GHSV) were conducted on catalysts whose phase compositions had reached dynamic equilibrium. Before a given run, the fresh catalyst was first operated at a GHSV of 6000 Nml g⁻¹ h⁻¹ for 8 h to allow the phase composition to reach equilibrium. Then, the catalyst was run sequentially at GHSVs of 1000, 6000, 10 000, 6000, and 2000 Nml g⁻¹ h⁻¹ for 1.5 h. Repeated operation at a GHSV of 6000 Nml g⁻¹ h⁻¹ was conducted to examine whether the catalytic performance, related to the phase state, changed after the catalyst had been exposed to low or high space velocities. Other reaction conditions, such as the reaction temperature of 325 °C and the reaction pressure of 20 bar, remained unchanged throughout the experiments.

2.3 Catalyst characterization

Ex situ X-ray diffraction patterns (XRD) of samples were measured under ambient air conditions using a Bruker D8 X-ray diffractometer with a CuKα (λ = 1.54 Å) radiation source. The diffraction angles (2θ) were scanned from 10 to 80° with a step size of 0.029° and a dwell time of 1 s.

In situ XRD experiments were conducted using a Rigaku D/max-2600 powder diffractometer (Rigaku, CuKα radiation) equipped with a high-temperature XRK-900 (Anton Paar) sample chamber and a gas supply system for thermal treatment under various atmospheres at an external laboratory. All in situ XRD patterns were recorded in the 2θ scanning range from 10° to 80° using a step size of 0.02° (2θ) and a time per step of 2 s. Typically, the sample powder (∼20 mg) was placed into a sample holder, inserted into the chamber, and purged with an Ar flow (20 Nml min⁻¹) for 10 min. Subsequently, carburizing gases (3 Nml min⁻¹ of H₂, 1.5 Nml min⁻¹ of CO, and 20 Nml min⁻¹ of Ar), or reactant gases (18 Nml min⁻¹ of H₂, 6 Nml min⁻¹ of CO₂) were introduced into the chamber, and XRD patterns were recorded at temperatures ranging from 25 °C to 500 °C with dwell periods of 20 min at each target temperature. The ramping rate between the two target temperatures was 5 °C min⁻¹, and all experiments were performed at atmospheric pressure.

The room temperature ⁵⁷Fe Mössbauer spectroscopic analysis was conducted on an MFD-500AV-02 transmission spectrometer (Topologic Systems) at an external laboratory. A ⁵⁷Co (Rh) source in constant acceleration mode was used as the radioactive source. The obtained spectra were analyzed and fitted using the MossWinn program, which modelled the spectra as appropriate superpositions of quadruple doublets and sextets of Lorentzian lines. The components of the iron phase were identified by parameters of the derived spectra, such as the isomer shift (IS), the quadruple splitting (QS), and the magnetic hyperfine field (Hhf). α-Iron foil was used to calibrate the doppler velocities and isomer shift at room temperature.

The X-ray photoelectron spectroscopy (XPS) measurements were performed on a PHI 5000 VersaProbe III scanning XPS microprobe system equipped with a monochromatic Al Kα (hν = 1486.6 eV) radiation source. The C 1s spectrum with the binding energy of 284.8 eV served as a reference. To prevent oxidation resulting from prolonged exposure to air, samples were prepared in a glove box and quickly transferred to the sample chamber. The pressure within the analysis chamber was maintained below 3 × 10⁻⁸ Pa. All samples used for XPS analysis were kept in an argon atmosphere or vacuum throughout the entire transfer process to avoid exposure to air.

The H₂O or CO₂ temperature-programmed oxidation (TPO) experiments were conducted using a flow apparatus comprised of a versatile gas feed system, a quartz tube reactor with an internal necking (L = 33 cm, Φ4 mm × 2 mm), and a mass spectrometer detector (Hiden HPR-20 QUI MS). Typically, the sample was placed in the isothermal zone of the reactor tube and secured by quartz wool plugs. A K-type thermocouple was inserted into the sample bed to monitor and control the temperature during heating. The reactor was surrounded by a metal coil for resistive heating and covered by a layer of quartz wool for thermal insulation. For the H₂O-TPO experiment, 30 mg of granular sample (250 µm–500 µm) was first flushed in an Ar flow for 30 min, and the tube reactor and gas pipelines were preheated to 110 °C. Afterward, a designed steam flow (2% H₂O in Ar, 100 Nml min⁻¹) produced by a water vapor generator was introduced into the reactor until saturation was achieved. The H₂O-TPO was then conducted in a continuous steam flow, with the sample being heated at a ramping rate of 10 °C min⁻¹ from 110 °C to 600 °C and held at 600 °C for 1 h to ensure complete oxidation. The CO₂-TPO experiment followed a similar temperature ramping procedure as the H₂O-TPO experiment but was conducted in a CO₂ flow (2% CO₂ in Ar, 100 Nml min⁻¹). All effluent streams were monitored by an online mass spectrometer, and the signals for H₂O (m/z = 18), CO₂ (m/z = 44), CO (m/z = 28), H₂ (m/z = 2), and CH₄ (m/z = 16) were calibrated using a gas mixture of known composition.

The H₂ (or CO)- TPD and H₂-TPR experiments were performed in the same system used in TPO experiments. Before TPD or TPR experiments, 30 mg of power sample (<500 µm) was first carburized at 350 °C under a syngas flow (3% H₂ and 1.5% CO in Ar) for 60 min, followed by flushing with Ar for 30 min while cooling the sample to 50 °C. For the subsequent TPD experiments, H₂ or CO adsorption (2% CO in Ar, or 2% H₂ in Ar) was carried out at 50 °C for 1 h, Finally, desorption was performed under a pure Ar flow with a heating rate of 10 °C from 50 °C to 800 °C, followed by holding at 800 °C for 30 min. The H₂-TPR experiments were conducted using a gas mixture of 5% H₂ in Ar, the reduction temperature was increased from 50 °C to 800 °C at a heating rate of 10 °C, and maintained at 800 °C for 30 min. Note proper handling of H₂ and CO is required due to their respective explosive and toxic properties.

Scanning Transmission Electron Microscopy (STEM) images and HAADF/BF (High Angle Annular Dark Field/Bright Field) STEM images were acquired by using a JEOL JEM ARM200F with a Cs-corrected probe and 200 kV high tension. The incident probe with a nominal 8.7 pA current at a convergence semi-angle of about 20 mrad was used for HAADF and BF images. The acquisition time was set to 10.8 s for a full-size image (4096 × 4096) to reduce the sample drift for minimizing the image distortion. The elemental mapping was acquired using energy dispersive spectroscopy (EDS, Oxford X-maxN 100TLE) that scanned over the specimen collaborating with the automatic drift correction during the spectrum collection. Before characterization, all samples were dispersed in anhydrous ethanol and then transferred onto a lacey-carbon thin film for the STEM analysis.

3 Results and discussion

3.1 The physicochemical properties of FeC_x and FeO_y catalysts

The starting material, fresh nano-sized iron oxide powder (NANOCAT, MACH-1 Inc), has been widely reported in previous works.^20–23 As a nanosized commercial iron oxide, it consists of poorly crystalline iron oxide (α-Fe₂O₃) with an average particle size of 6.2 nm (Fig. S1) and a surface area of 255 m² g⁻¹ after drying.²³ Its good availability and high stability ensure its consistency for material pretreatment and modifications in subsequent research. When the nano-sized Fe₂O₃ (Nano-Fe₂O₃) was exposed to the CO₂ hydrogenation atmosphere, it underwent various phase transitions induced by the reactants (CO₂ and H₂) and the products (CO and H₂O). The ex situ XRD diffraction patterns of Nano-Fe₂O₃ oxide treated with different atmospheres are listed in Fig. 1A. After being treated with a mixture of CO and H₂ (1.5% CO + 3% H₂ in Ar) at a low temperature of 300 °C for 12 h, the obtained sample FeC_x showed only obvious diffractions at 2θ of 39.4°, 44.2°, 47.3°, and 50.2°, indicating that the nano-sized Fe₂O₃ was converted into Fe₅C₂ (JCPDS 36-1248),^10,11,21 and the purity of the obtained iron carbide was greater than 88.7%. The further detailed quantitative analysis is discussed in Section 3.3. Since the FeC_x sample was passivated with a low oxygen concentration (2% O₂ in Ar) at 25 °C after carburization treatment, an oxide layer formed on its surface. The corresponding STEM image (Fig. 1B) revealed that the sample FeC_x primarily consisted of nanorod-shaped particles (particle width ≈ 30 nm) with a core–shell structure. The selected area electron diffraction pattern (SAED, inset image in Fig. 1B) of the core further indicated a lattice spacing of 0.21 ± 0.05 nm, corresponding to the (510) plane of crystalline Fe₅C₂.²⁴ The thin amorphous shell layer (width ≈ 4 ± 1 nm) was identified as iron oxide rather than carbon. A clear interface between the oxide layer and iron carbide can be clearly distinguished by high-angle annular dark-field imaging (HAADF) STEM, as shown in Fig. S2. The shell consists of small amorphous iron oxide species (magnetite, blue box), while the core is composed of crystallites of Fe₅C₂ viewed along the [010] zone axis.

Fig. 1 (A) XRD patterns of starting Nano-Fe₂O₃, as-prepared FeC_x, FeO_y catalyst, comparison sample FeO_z-CO₂, and Carb-FeO_y obtained after different gas atmosphere treatments. STEM images of as-prepared FeC_x (B) and FeO_y (C) catalysts. Inset images are selected area electron diffraction patterns (SAED) of the corresponding catalysts.

It has been reported that both CO₂ and H₂O play an important role in oxidizing Fe catalysts during the CO₂ hydrogenation reaction.¹⁰ Here, the ex situ XRD results further prove that iron carbide is easily oxidized by CO₂ and H₂O, with the oxidation effect of H₂O being significantly stronger than that of CO₂. After being treated with CO₂ (20% CO₂ in Ar) at 300 °C for 12 h, the Fe₅C₂ catalyst was partially oxidized to Fe₃O₄, while a portion of Fe₅C₂ was preserved (Fig. 1A). The XRD patterns of the sample FeO_z-CO₂ exhibited both crystalline phases of Fe₃O₄ (JCPDS 19-0629) and Fe₅C₂ (JCPDS 36-1248). In contrast, when the Fe₅C₂ was treated with steam (1% H₂O in Ar) at the same temperature and duration, the iron carbide was completely converted into Fe₃O₄, as indicated by characteristic diffractions at 2θ of 30.1°, 35.4°, 43.2°, 57.1°, and 62.6° (sample FeO_y in Fig. 1A), illustrated in Fig. 1C. The SAED image (inset in Fig. 1C) indicates that the core–shell structure of Fe₅C₂ disappeared and was replaced by even-textured particles. Specifically, clear lattices of crystalline Fe₃O₄ aligned along the zone axis [110] are observed in the FeO_y sample. The core–shell Fe₅C₂ particles were oxidized into even-textured Fe₃O₄ particles during the steam treatment, providing additional insight into the identification of the shell structure of the iron carbide. In early literature, it was generally speculated that the surface shell of nano-iron carbide particles was decorated by a small amount of Fe₃O₄ or amorphous graphitic carbon.²⁴ Although it is difficult to completely rule out the presence of graphitic carbon, it is evident that carbon or graphite remains unchanged during the steam treatment. This indirectly confirms that Fe₃O₄, rather than graphitic carbon, is the primary component of the shell structure of the FeC_x sample.

In the following sections, the core–shell structure will be used as a key characteristic to identify the morphology of iron carbide particles. In terms of particle sizes before and after steam treatment, it is worth noting that there was a slight increase in the size of the Fe₃O₄ particles obtained via water treatment of Fe₅C₂. The particles transformed from long, rod-like structures with a rod average diameter of 29 ± 3 nm to irregularly shaped particles with an average diameter of 35 ± 10 nm. When the resulting Fe₃O₄ was subsequently carburized again using CO and H₂ treatment, it was not completely converted back to Fe₅C₂. This is evident in the XRD pattern of the comparison sample Carb-FeO_y (Fig. 1A). The partial carburization may be attributed to the growth of the Fe₃O₄ crystals (FeO_y), which likely require higher temperatures for complete carburization than the initial nanocrystals (Nano-Fe₂O₃).

To study the details of the carburization of iron oxides under CO and H₂ atmospheres and explore the differences in the temperatures required for carburizing the iron oxide with different particle sizes, in situ XRD experiments were carried out, with the results shown in Fig. 2A and B. From the XRD pattern in Fig. 2A, the carburization process of Nano-Fe₂O₃ proceeded through the following stages: Fe₂O₃ → Fe₃O₄ → a mixture of FeO and Fe₅C₂ → Fe₅C₂ → a mixture of Fe₅C₂ and Fe₃C. For carburizing at 200 °C and below, the nanosized Fe₂O₃ was only reduced to Fe₃O₄. The particle size was approximately 10.8 nm (Fig. S3, Table S1), estimated from the half-maximum width of the characteristic diffraction peak. When the temperature reached 250 °C, Fe₃O₄ began to partially reduce to FeO (JCPDS 77-2355). Simultaneously, the carburization process commenced, and a weak diffraction peak at 44.2°, attributed to Fe₅C₂, appeared in the XRD pattern of the obtained sample. At 300 °C, the reduction of Fe₃O₄ to FeO weakened, and the carburization of Fe₃O₄ to Fe₅C₂ became dominant. When the temperature increased to 400 °C, nearly all iron oxides were carburized into Fe₅C₂. However, high-temperature carburization also led to carbon deposition due to the Boudouard reaction, as indicated by a broadened diffraction peak of amorphous carbon at 2θ of approximately 26° in the XRD pattern. It should be noted that, in the experiments to prepare large batches of pure Fe₅C₂ sample (FeC_x), the carburization treatment was carried out at a low temperature of 300 °C for 12 h, with small amounts of sample treated per batch. This approach ensured complete carburization while avoiding carbon deposition associated with higher temperatures.

Fig. 2 The in situ XRD patterns of Nano-Fe₂O₃ (A) and FeO_y (B) samples carburized under a syngas atmosphere (3 Nml min⁻¹ of H₂, 1.5 Nml min⁻¹ of CO, and 20 Nml min⁻¹ of Ar), and the in situ XRD patterns of FeC_x (C) and FeO_y (D) samples treated under a reactant atmosphere (18 Nml min⁻¹ of H₂, 6 Nml min⁻¹ of CO₂). The patterns of all samples were collected after stabilization at the corresponding temperatures (25 °C, 200 °C, 250 °C, 300 °C, 350 °C, 400 °C, 450 °C, and 500 °C) for 20 min.

As the reaction temperature continued to increase, the diffraction intensity of both carbon deposits and Fe₅C₂ gradually increased, suggesting that their degree of crystallization also increased, or their particle size grew larger. Furthermore, the diffraction attributed to Fe₃C appeared in the XRD pattern collected at 500 °C, consistent with previous reports indicating that high temperatures favor the phase transformation from χ-Fe₅C₂ to θ-Fe₃C.¹⁴

Unlike the nanosized Fe₃O₄ (Fig. 2A, the sample after 200 °C carburization of the Nano-Fe₂O₃), which consisted of small particles (DP ≈ 10.8 nm), the carburization of the FeO_y sample containing larger Fe₃O₄ particles (DP ≈ 21.7 nm, obtained from Fig. S3, Table S1) occurred at higher temperatures and followed a different phase transformation path. The XRD patterns of the carburized FeO_y samples, collected at various temperatures, are shown in Fig. 2B. As the carburization temperature increased, the phase transformation proceeded in the sequence: Fe₃O₄ → a mixture of Fe₅C₂ and Fe₃O₄ → a mixture of Fe₅C₂ and Fe₃C → a mixture of FeO, Fe₅C₂, and Fe₃C. When the temperature was below 300 °C, the FeO_y retained the Fe₃O₄ crystalline phase without any noticeable reduction or carburization. At 350 °C, carburization of Fe₃O₄ to Fe₅C₂ began, as evident by the weakening of Fe₃O₄ diffractions and the emergence of characteristic Fe₅C₂ diffraction peaks. At 400 °C and above, diffraction peaks corresponding to Fe₃C appeared, indicating the phase transformation to Fe₃C. At 500 °C, the Fe₃O₄ in the FeO_y sample was not fully converted into iron carbide. Instead, the remaining uncarburized iron oxide underwent further reduction to form FeO, and coke formation also began. As a result, the XRD pattern at 500 °C displayed mixed diffraction signals from ferrous oxide (FeO), iron carbide (Fe₅C₂/Fe₃C), and amorphous carbon (2θ of 26°).

Fig. 2C and D show the in situ XRD diffraction patterns of FeC_x and FeO_y samples, respectively, collected in a reaction atmosphere at atmospheric pressure but at different temperatures. These patterns reveal the stability of the two catalysts before the CO₂ hydrogenation reaction begins. The iron carbide catalyst FeC_x remained stable at temperatures below 400 °C when exposed to a reactant atmosphere of 25% CO₂ and 75% H₂. Only distinct diffraction peaks attributed to Fe₅C₂ were present in the XRD pattern (Fig. 2C). When the temperature exceeded 400 °C, a broadened diffraction peak attributed to amorphous carbon (2θ of 26°) also appeared, and its intensity increased as the temperature rose. In addition, higher temperatures also promoted the phase transition from Fe₅C₂ to Fe₃C, as indicated by the trace diffraction of Fe₃C in the XRD pattern of FeC_x after treatment at 500 °C. The FeO_y catalyst also exhibited high stability below 400 °C. Under the CO₂ and H₂ atmosphere, the phase transformation of FeO_y occurred only above 500 °C. Unlike the FeC_x catalyst, the FeO_y catalyst underwent reduction and carburization simultaneously during high-temperature treatment, leading to the phase transformation of Fe₃O₄ into iron carbide and ferrous oxide, as illustrated in Fig. 2D.

During CO₂ hydrogenation reaction test in the fixed-bed reactor, the temperature was first increased to 350 °C, followed by a pressure increase to 10 bar to initiate the reaction. The above in situ XRD results clearly showed that FeC_x and FeO_y catalyst samples were phase-stable in a reactant atmosphere at atmospheric pressure and temperatures below 400 °C. This clears up doubts that the catalyst might undergo phase changes during prolonged exposure to the reactants. Subsequent phase changes of the catalyst are mainly caused by the reaction itself when the reaction pressure increases. Therefore, the bulk phase composition of the fresh catalyst, collected under ex situ conditions (at room pressure and temperature), can reliably represent the composition of the catalyst just before the reaction begins. It is also reasonable to use ex situ measurements to study the phase changes of the catalyst during the reaction process.

Up to now, it has been confirmed that the prepared FeC_x catalyst sample is mostly composed of Fe₅C₂ (≥88.7%), while FeO_y catalyst sample consists entirely of Fe₃O₄ (comparison in Fig. 2A and B). FeC_x is synthesized by carburizing nano-Fe₂O₃ at a relatively low temperature of 300 °C and can be oxidized to Fe₃O₄ through water vapor treatment (FeO_y sample). However, even at higher carburization temperatures, FeO_y cannot be fully converted back to Fe₅C₂. This is mainly because the carburization of Fe₃O₄ with different particle sizes follows distinct phase transformation pathways and rates. Since the fresh FeC_x and FeO_y catalyst samples are almost pure phases, similar in particle size, stable before reaction, and capable of undergoing reversible phase transformations under both oxidizing and carburizing conditions, they are used in thus study as model catalysts. All subsequent experiments, results, comparisons, and discussions are based on these model systems.

3.2 Catalytic performance of pure or Na₂CO₃-modified FeC_x and FeO_y catalysts

The catalytic performances of pure and Na₂CO₃-modified FeC_x and FeO_y catalysts are shown in Fig. 3 (from A to D). Compared with pure FeO_y (Fig. 3B), the pure FeC_x catalyst (Fig. 3A) exhibited higher Fischer–Tropsch synthesis (FTS) activity with higher CO₂ conversion (X_CO2 ≥ 40%) and lower CO selectivity (S_CO ≤ 10%). In contrast, FeO_y showed higher Reverse Water-Gas Shift (RWGS) activity (S_CO ≥ 80%, X_CO2 = 15%). During CO₂ hydrogenation, the fresh FeC_x catalyst was likely oxidized and partially transformed into FeO_y catalyst (Fig. 3A), leading to a gradual decrease in the conversion of CO₂ from 51.2% to 38.9% over 4 h of time on stream (TOS). Concurrently, the selectivity for CO increased from 3.9% to 7.4%, and then reached an approximate steady state. After re-carburizing the spent FeC_x catalyst (TOS = 8 h), the regenerated catalyst (TOS = 8.5 h) exhibited enhanced activity and reduced CO selectivity compared to the spent state (TOS = 8 h). Notably, the steady-state activity of the regenerated catalyst was very similar to that of the fresh catalyst. This indicates that the deactivated FeC_x catalyst can be efficiently restored by re-carburization and confirms that Fe₅C₂, rather than Fe₃O₄, serves as the effective catalytic phase for CO₂ conversion and hydrocarbon formation. This conclusion is further supported by the high activity of the carburized (or regenerated) FeO_y catalyst. After carburization with syngas (as shown in Fig. 3B at TOS = 8.5 h), the nearly FTS-inert Fe₃O₄ can be partially converted into active Fe₅C₂, as demonstrated in the previous sections (Fig. 1 A, Carb-FeO_y sample). Consequently, the carburized FeO_y catalyst exhibited high FTS activity, comparable to that of the FeC_x catalysts.

Fig. 3 CO₂ hydrogenation performance over fresh & regenerated FeC_x (A), FeO_y (B), PM-Na-FeC_x (C), and PM-Na-FeO_y (D) catalysts. Reaction conditions: Catalyst weight = 1 g, GHSV = 6000 Nml g⁻¹ h⁻¹, T = 325 °C, P = 20 bar, H₂/CO₂ = 3 mol mol⁻¹.

High selectivity for light olefins was achieved over a fine physical mixture of FeC_x catalyst and Na₂CO₃ dopant (Fig. 3C). In contrast, a similar mixture of FeO_y and Na₂CO₃ was not active for producing light olefins or other hydrocarbons. The performance of PM-Na-FeO_y was almost the same as that of the fresh FeO_y catalyst (Fig. 3B and D, TOS ≤ 8 h). However, once the Fe₃O₄ within FeO_y was carburized into Fe₅C₂, the mixed catalyst quickly established high activity for CO₂ conversion and high selectivity for light olefins at TOS of 8.5 h (Fig. 3D). This fully confirms that the active center for olefin production originates from the combination of Fe₅C₂ and Na₂CO₃. Nevertheless, regardless of Na modification, all catalysts containing Fe₅C₂ experienced initial deactivation at the beginning of the reaction. As the reaction time progressed, CO₂ conversion decreased while CO selectivity increased until a steady-state equilibrium was reached. Notably, the CO selectivity was higher and increased more significantly in the sodium-containing samples than in the sodium-free counterparts. For FeC_x or carburized FeO_y, the CO selectivity remained below 10%, whereas for PM-Na-FeC_x or carburized PM-Na-FeO_y, the CO selectivity reached as high as 25% at steady-state.

3.3 The phase transition of the pure FeC_x and FeO_y catalysts during CO₂ hydrogenation

Since the combination of iron carbide and sodium is a key active center for olefin production, both sodium-free and sodium-containing catalysts exhibited a similar activity decay trend of first decreasing and then reaching a steady state. However, they were restored to short-term activity after carburization treatment, so it is essential to gain a deeper understanding of phase transitions occurring related to the diverse catalytic performances. First, the phase change-induced deactivation and regeneration (carburization) of the sodium-free catalysts were explored using ex situ XRD, Mössbauer spectroscopy, XPS, and STEM.

Fig. 4A and B show the XRD patterns of FeC_x (A) and FeO_y (B) catalysts in their fresh, spent, regenerated (carburized), and final spent-regenerated states, respectively. For the FeC_x catalyst, the fresh state contained only pure Fe₅C₂ iron carbide species. After 8 h of the catalytic reaction, most of the Fe₅C₂ species were oxidized to Fe₃O₄ species. Compared with stable Fe₅C₂ under the reactants' atmosphere (CO₂ + H₂) at 400 °C and atmospheric pressure (Fig. 2C), the reactant CO₂ and product H₂O were proven to be direct oxidizing agents that oxidize Fe₅C₂ under high temperature and high-pressure conditions. However, when the spent catalyst was treated with syngas (CO + H₂), part of the Fe₃O₄ species was carburized back into the Fe₅C₂ species. Consequently, the regenerated FeC_x catalyst (Regen-FeC_x) had restored FTS activity, evident by higher CO₂ conversion and increased selectivity for hydrocarbons. After another reaction cycle, nearly all the Fe₅C₂ phase in the regenerated catalyst was oxidized back to the Fe₃O₄ phase. See the XRD pattern of the spent-regenerated FeC_x catalyst (Spent-Regen-FeC_x).

Fig. 4 The ex situ XRD patterns of FeC_x (A) and FeO_y (B) catalysts in fresh, spent, regenerated, and final spent-regenerated states. The corresponding ex situ Mössbauer spectra of FeC_x (C) and FeO_y (D) catalysts in their corresponding states.

For the FeO_y catalyst, after the CO₂ hydrogenation reaction, the bulk phase composition of the spent catalyst remained nearly unchanged compared with the fresh one: both exhibited pure Fe₃O₄ phases, which also explained their lower activity during CO₂ hydrogenation. After carburization treatment with syngas, the carburized FeO_y catalyst showed a mixture of crystalline Fe₃O₄ and Fe₅C₂ phases and exhibited good activity similar to the regenerated FeC_x (Regen-FeC_x) catalyst. This suggests that H₂ and CO can be used to form the active Fe₅C₂ phase. Like the regenerated FeC_x catalyst, the carburized FeO_y catalyst was also unstable. After an additional 8 h of CO₂ hydrogenation, the most of Fe₅C₂ phase were oxidized, and the XRD patterns of the spent-regenerated FeO_y catalyst (Spent-Regen-FeO_y) showed predominantly Fe₃O₄ crystal phases.

Further quantitative and qualitative analysis of the bulk phase composition of the catalysts at different states (fresh, spent, regenerated, and final spent-regenerated) was performed using Mössbauer spectrometry (MES), and the results are shown in Fig. 4C and D. The detailed parameters are listed in Tables S2 and S3. The sextets with Hhf of around 491 ± 1 and 460 ± 5 kOe are commonly attributed to the tetrahedral (A) and octahedral (B) sites of Fe₃O₄, respectively.^13,25–27 The sextets with Hhf of 116 ± 3, 213 ± 2, and 190 ± 5 kOe are assigned to the three different nonequivalent sites of Fe₅C₂.^13,26,28 Another observed doublet is likely ascribed to the superparamagnetic phase (spm) of Fe³⁺¹³ or FeO.²⁹ The quantitative analysis from MES was consistent with the XRD results. The fresh FeC_x catalyst exhibited an almost entirely Fe₅C₂ (≥88.7%) bulk composition, while the fresh FeO_y catalyst consisted purely of Fe₃O₄ species (100%). After the CO₂ hydrogenation, 74.8% of the Fe₅C₂ species in the FeC_x catalyst were oxidized to Fe₃O₄ species, while the pure FeO_y catalyst (formed via H₂O treatment of FeC_x) had a stable phase composition, with 100% Fe₃O₄ content both before and after reaction. Upon carburization of the spent catalysts with syngas, more than 50% of the iron carbide components were restored in the regenerated catalysts. The Fe₅C₂ contents in the regenerated FeC_x (Regen-FeC_x) catalyst and carburized FeO_y catalyst (Regen-FeO_y) are 52.8% and 56.6%, respectively. However, continued reaction once again oxidized iron carbide to iron oxide. The residual Fe₅C₂ content in the spent-regenerated FeC_x (Spent-Regen-FeC_x) and spent-regenerated FeO_y (Spent-Regen-FeO_y) catalysts decreased to 14.7% and 7.1%, respectively.

The bulk composition of FeC_x and FeO_y catalysts was correlated with their corresponding FTS activity (represented by the yield of C₂₊) at different stages, as shown in Fig. 5A. The results indicated that Fe₅C₂ was the active phase responsible for hydrocarbon production during CO₂ hydrogenation. Notably, regardless of whether the Fe₅C₂ phase originated from carburized nano-Fe₂O₃ (in FeC_x series) or was formed by carburizing bulk Fe₃O₄ (in FeO_y series), only a small amount (≤30%) of active Fe₅C₂ phase could be preserved, eventually reaching a dynamically stable state during the CO₂ hydrogenation reaction.

Fig. 5 (A) Relationship between the bulk composition from Mössbauer spectroscopy of two iron catalysts and their corresponding C₂₊ yields. (B) STEM image of Spent-FeO_x catalyst, (C) STEM image of Spent-Regen-FeO_y catalyst. The positions of Fe₅C₂ particles are marked with red dashed lines in images, as Fe₅C₂ has typical core–shell structural characteristics. For the definitions in Figure A, Fresh: the fresh FeC_x and FeO_y catalysts respectively; Spent: the catalyst after 8 h of CO₂ hydrogenation conditions (GHSV = 6000 Nml g⁻¹ h⁻¹, T = 325 °C, P = 20 bar, H₂/CO₂ = 3 mol mol⁻¹); Regen: the catalysts after re-carburizing under a syngas atmosphere (3% H₂ and 1.5% CO in Ar, 500 Nml min⁻¹) for 12 h at 300 °C; Spent-Regen: the regenerated catalyst after a second 8 h run under CO₂ hydrogenation conditions (GHSV = 6000 Nml g⁻¹ h⁻¹, T = 325 °C, P = 20 bar, H₂/CO₂ = 3 mol mol⁻¹).

For example, the fresh FeC_x catalyst contained 89% Fe₅C₂ and yielded 32% C₂₊ products. In contrast, the Spent-FeC_x catalyst, with only 25% Fe₅C₂ remaining, showed a reduced C₂₊ yield of 23%. Interestingly, the Spent-Regen-FeO_y catalyst, despite having just 7.1% Fe₅C₂, still achieved a C₂₊ yield of 20%. This raises the question: why were these catalysts, containing relatively small amounts of Fe₅C₂ phase, still highly FTS-reactive even though their XRD diffractions looked indistinguishable from pure Fe₃O₄? To ascertain the reason for this apparent discrepancy, the detailed morphologies of the inactive Spent-FeO_y and active Spent-Regen-FeO_y catalysts were studied and compared, as shown in Fig. 5B and C, respectively. The Spent-FeO_y catalyst contained only even-textured particles with an average particle size of 64 nm. Combining the results of XRD and Mössbauer spectroscopy analysis, these particles were identified as Fe₃O₄. In contrast, the Spent-Regen-FeO_y catalyst contained both even-textured particles (∼89 nm in average) and smaller particles (∼40 nm) exhibiting a core–shell structure. In particular, the number of core–shell particles was less than that of the even-textured particles in the mapping area. As discussed previously, the core–shell structure was the special characteristic of iron carbide species, whose surface has always been oxidized into Fe₃O₄ during the passivation treatment or exposed to air during sampling.

In the Spent-Regen-FeO_y catalyst, further EDX analysis of selected regions around the core–shell particles clearly distinguished the boundary between iron carbide and iron oxide, indicating that these core–shell particles consist of iron carbide in cores surrounded by iron oxide surface layers. The other even-textured particles were identified as iron oxides. Combining these observations with XRD and Mössbauer spectra results, it can be further concluded that these iron carbides were Fe₅C₂, and iron oxides were Fe₃O₄. Additionally, the formation of larger Fe₃O₄ particles in the Spent-Regen-FeO_y catalyst may be attributed to particle aggregation during the reaction. XRD results also confirmed this sintering behavior, as evidenced by the larger size of Fe₃O₄ crystals in the Spent-Regen-FeO_y catalyst compared to those in the Spent-FeO_y catalyst (Fig. 4B and Table S1).

XPS analysis further revealed differences in the surface iron species between the Spent-FeO_y and Spent-Regen-FeO_y catalysts, helping to explain the observed differences in FTS catalytic activity. As shown in Fig. S5, the Fe 2p_3/2 peaks centered around 710.7 eV and the Fe 2p_1/2 peaks located around 723.8 eV are attributed to the overlapping signals of Fe²⁺/Fe³⁺ oxide species; The binding energy of Fe 2p peaks of Fe atoms in Fe–C structures are located at 706.7 eV (Fe 2p_3/2) and 719.9 eV (Fe 2p_1/2).²⁶ It is evident that the surface species of the fresh and reacted FeO_y catalysts were predominantly iron oxide species. However, after the carburization treatment with syngas, iron carbide species began to form on the surface of Regen-FeO_y. The Spent-Regen-FeO_y catalyst still retained iron carbide on its surface after CO₂ hydrogenation, although its content had decreased significantly, as indicated by the weakening of the Fe–C related Fe 2p peaks in Fig. S5. Interestingly, although the Fe₅C₂ content in the bulk phase of the Spent-Regen-FeO_y catalyst was six times lower than that of the Regen-FeO_y catalyst according to quantitative Mössbauer spectroscopy analysis, the difference in surface iron carbide content is small. This may suggest that Fe₅C₂ was enriched on the surface of the Spent-Regen-FeO_y catalyst, thereby contributing to its observed reactivity. Overall, the small size of the iron carbide particles and their higher surface concentration likely explain why the Spent-Regen-FeO_y catalyst maintains considerable activity despite the low iron carbide content in its bulk phase.

Notably, the particle size of surface iron carbide is also closely related to the catalytic performance.^10,17 From the discussion in Section 3.1, it was shown that small particles of Fe₃O₄ and Fe₅C₂ can transform into each other under different CO + H₂/CO₂ (H₂O) atmospheres. However, once the particles grew larger, the phase transformation became less reversible at low temperatures (≤350 °C). In the early stages of the reaction, catalysts containing a large number of smaller Fe₅C₂ particles exhibited higher CO₂ conversion activity. However, the water produced by the reaction accelerated the oxidation of Fe₅C₂ into less-active Fe₃O₄, and the hydrothermal atmosphere also promoted particle growth, leading to a gradual decline in activity. At the same time, the increasing iron oxide content and decreasing iron carbide content resulted in enhanced CO selectivity. As the reaction proceeded, the increased CO production carburized only the small Fe₃O₄ particles, while the larger Fe₃O₄ particles remained unconverted. Therefore, in Fe₅C₂-containing catalysts, after the initial decline in activity, the reaction tended to reach a steady-state activity. This behavior was mainly attributed to the establishment of a dynamic phase equilibrium of Fe₅C₂ within the catalyst, characterized by stable content and particle size.

3.4 Influence of contact intimacy between Na₂CO₃ and Fe₅C₂ on olefin yield

In previous literature, Fe-based catalysts are typically impregnated with Na or K dopants and then calcined, reduced at high temperatures, or carburized to form effective catalysts for the hydrogenation of CO₂ to produce olefins.^9,12–19 As shown in Fig. 3D, the olefin selectivity of the catalyst can be significantly improved by simple physical mixing of Na₂CO₃ and Fe₅C₂ components. To further investigate the interaction between Fe₅C₂ and Na₂CO₃ and to identify the active center responsible for olefin production, three catalysts with identical Na₂CO₃ loadings but different contact distances between FeC_x and Na₂CO₃ were prepared using three different methods. The preparation methods and corresponding catalytic performance are illustrated in Fig. 6.

Fig. 6 CO₂ hydrogenation over Na₂CO₃-modified FeC_x catalyst with varying initial proximities. (A) Mixing pure FeC_x granules with Na₂CO₃ particles (250 µm −500 µm) in a vial. (B) Mixing FeC_x powder with Na₂CO₃ powder using mortar grinding. (C) Wetness impregnation (WI) of granular FeC_x (250–500 µm) with Na₂CO₃ aqueous solution. The schematic diagram is shown above, and the images below depict the respective catalytic activities. The pink balls represent Na₂CO_3, and the black ones represent FeC_x. Reaction conditions: Catalyst weight = 1 g, GHSV = 6000 Nml g⁻¹ h⁻¹, T = 325 °C, P = 20 bar, H₂/CO₂ = 3 mol mol⁻¹.

For the GM-Na-FeC_x catalyst (Fig. 6A), where the Na₂CO₃ and FeC_x were integrated by granule mixing with a particle size ranging from 250 to 500 µm, its catalytic performance was nearly the same as that of pure FeC_x. Under steady-state conditions, the CO₂ conversion was approximately 38%, and the selectivities for CH₄, light olefins, and CO were 34%, 28%, and 11%, respectively. When the contact distance between Fe₅C₂ and Na₂CO₃ was reduced to 100–5 µm, the active centers for olefin production were generated, but the activity was unstable, as demonstrated by the PM-Na-FeC_x catalyst produced by mortar grinding and mixing (Fig. 6B). The selectivity for light olefins exceeded 40%, but the initial activity (CO₂ conversion) dropped sharply from 46% to 25% and required approximately 8 h to reach a steady state. Further increasing the proximity between Na₂CO₃ and Fe₅C₂ can rapidly promote the formation of stable active centers for olefin production. The IM-Na-FeC_x catalyst, prepared via incipient wetness impregnation to achieve close contact between the components, exhibited stable performance after 3 h (Fig. 6C). Interestingly, when steady-state conditions were reached, the performance of the IM-Na-FeC_x and PM-Na-FeC_x catalysts was very similar. Their CO₂ conversion decreased to 26%, and the selectivity for CH₄ dropped to 11%, while the selectivity for light olefins and CO increased to 40% and 26%, respectively. These results demonstrate that a highly intimate contact between Na₂CO₃ and Fe₅C₂ forms highly active centers to produce light olefins. Similar phenomena have been reported for Fe/C catalysts modified with carbonaceous potassium promoters, in which an appropriate proximity between the potassium species and Fe/C was found to promote the formation of iron carbide, leading to enhanced olefin selectivity and yield.⁹ Our findings further confirm that the interface between iron carbide and Na or K carbonate is the key factor in generating highly active sites for olefin production (Fig. S6).

The formation and evolution of the active Na₂CO₃–Fe₅C₂ interface within the PM-Na-FeO_y catalysts were visualized using STEM and EDX, as shown in Fig. 7, S7, and S8. Before carburization, both the fresh and spent PM-Na-FeO_y catalysts contained only Fe₃O₄ particles with an average particle size of approximately 30 nm, on which the Na dopants were well dispersed, as seen in Fig. 7A and B, and the corresponding elemental mapping images (Fig. 7E and F). After syngas carburization and subsequent oxygen passivation, some Fe₃O₄ particles in the spent PM-Na-FeO_y catalyst were converted into Fe₅C₂ particles, these Fe₅C₂ particles were intermixed with the remaining Fe₃O₄ particles, as shown in the images of the regenerated PM-Na-FeO_y catalyst (Fig. 7C and G). After 8 hours of the CO₂ hydrogenation reaction again, the morphology of spent-regenerated PM-Na-FeO_y was examined. It was observed that most of the Fe₅C₂ in the catalyst had been re-oxidized, with only a small amount of Fe₅C₂ remaining. In addition, the particle size increased, as shown in Fig. 7D and H.

Fig. 7 STEM images and corresponding EDX elemental mapping of PM-Na-FeO_y catalyst in fresh (A and E), spent (B and F), regenerated (C and G), and final spent-regenerated (D and H) states.

The difference between iron carbide (Fe₅C₂) and iron oxide (Fe₃O₄) can be clearly distinguished by the distribution of carbon (C) and oxygen (O) elements in the EDX mapping images (Fig. S7 and S8). The C element was more concentrated in the iron carbide regions, while the O element was enriched in the iron oxide regions. Accordingly, in Fig. 7, the EDX images show that the Fe₃O₄ particles appeared yellow-green, reflecting enrichment with oxygen (green) and iron (red), whereas the areas containing Fe₅C₂ particles appeared red-pink, indicating enrichment with carbon (blue) and iron (red). In Fig. S7 and S8, The Na element is evenly dispersed over both single particles of Fe₃O₄ (Fig. S7) and Fe₅C₂ (Fig. S8), showing no preferential distribution. During sample preparation for the TEM experiments, the fresh bulk catalyst was ground again and dispersed onto a carbon mesh using anhydrous ethanol as the solvent. As a result, the Na was also evenly distributed at the nanometer scale on the catalysts, which represents a stable state of Na distribution.

Notably, while the fresh and spent PM-Na-FeO_y catalysts contained only interfaces between Fe₃O₄ and Na, the Regen-PM-Na-FeO_y catalyst generated new interfaces between Fe₅C₂ and Na. This is the primary reason for the high yield of light olefins observed in the Regen-PM-Na-FeO_y catalyst. After an additional 8 hours of CO₂ hydrogenation, Fe₃O₄ once again became the dominant phase in the Spent-Regen-PM-Na-FeO_y catalyst (final spent state, see Fig. 7). However, a small amount of residual Fe₅C₂ and the corresponding Na–Fe₅C₂ interfaces persisted, as confirmed by the TEM and EDX images in Fig. S8. This retention helps explain the Spent-Regen-PM-Na-FeO_y catalyst's ability to maintain a certain level of reactivity. Although its reactivity was lower than that of the Regen-PM-Na-FeO_y catalyst, it remained higher than that of the fresh and spent PM-Na-FeO_y catalysts, which contained only Na–Fe₃O₄ interfaces.

3.5 Effect of Na₂CO₃ on the phase transition of Fe₅C₂ during CO₂ hydrogenation

As established earlier, the interface between Fe₅C₂ and Na₂CO₃ serves as the active center to produce light olefins. However, Fe₅C₂ was easily oxidized during the CO₂ hydrogenation reaction, resulting in a decrease in Fe₅C₂ and an increase in Fe₃O₄ content, which in turn results in reduced catalyst activity. To investigate the effect of Na₂CO₃ on the oxidation of Fe₅C₂ species and to explore the reasons behind the lower activity (or CO₂ conversion) and higher CO selectivity observed in sodium-modified catalysts, comparative experiments were conducted between pure FeC_x and Na-modified PM-Na-FeC_x catalysts.

Fig. 8A displays the XRD patterns of these two catalysts before and after the reaction. In the fresh PM-Na-FeC_x catalyst, the Na₂CO₃ dopant was evenly dispersed, and no diffraction peaks attributable to sodium carbonate were observed; only diffraction peaks of pure Fe₅C₂ appeared. Moreover, the addition of Na₂CO₃ did not change the hyperfine interactions of the iron nuclei with their surroundings, as evident from the Mössbauer spectrum of the fresh PM-Na-FeC_x catalyst. These spectra were identical to those of the fresh FeC_x, indicating a high Fe₅C₂ composition (88.8%) as shown in Fig. 8B.

Fig. 8 The ex situ XRD patterns (A) and the corresponding Mössbauer spectra (B) of FeC_x and PM-Na-FeC_x catalysts in fresh and spent states.

After the CO₂ hydrogenation reaction, both the Na-containing and non-Na-containing catalysts underwent oxidation, with most of the Fe₅C₂ transforming into Fe₃O₄. The XRD pattern of Spent-FeC_x exhibited a mixture of diffraction patterns of the remaining Fe₅C₂ and oxidized Fe₃O₄ species, while the XRD pattern of the spent PM-Na-FeC_x catalyst showed only Fe₃O₄ diffraction peaks. Based on the integration of their Mössbauer spectra (Fig. 8B), the quantification result indicated that the Fe₅C₂ content in spent PM-Na-FeC_x (11.75%) was lower than that in the spent FeC_x (24.57%), as shown in Table S4. This suggests that Na did not prevent the oxidation of the nanosized Fe₅C_2; rather, the Na-modified nanosized Fe₅C₂ catalyst appeared to oxidize more extensively than the pure Fe₅C₂ catalyst.

The promoting effect of Na₂CO₃ on the oxidation of Fe₅C₂ can be further verified by H₂O-TPO and CO₂-TPO experiments. During the CO₂ hydrogenation, both the reactant CO₂ and product H₂O can oxidize the iron carbide as follows:³⁰

3FeC_x + (4 + 3x)H₂O → 3xCO + Fe₃O₄ + (4 + 3x)H₂

3FeC_x + (4 + 3x)CO₂ → (4 + 6x)CO + Fe₃O₄

To simulate the oxidation of catalysts during the reaction and investigate the effect of Na₂CO₃ on the oxidation of iron carbide, temperature programmed oxidation (TPO) experiments with FeC_x and PM-Na-FeC_x catalysts were conducted in water vapor and CO₂ atmospheres, respectively. The consumed oxidant and released products were recorded and are shown in Fig. 9 and S9. In the H₂O-TPO process (Fig. 9A), the oxidation of FeC_x occurred between 250 °C and 460 °C, with a maximum at 405 °C, while the oxidation temperature range of PM-Na-FeC_x was broader, extending from 250 °C to 600 °C. By examining the amount of consumed water and produced hydrogen, it is evident that the PM-Na-FeC_x catalyst underwent deeper oxidation than the FeC_x catalyst, consuming more H₂O and producing more H₂ effluent. Similar results were observed in the CO₂-TPO experiment (Fig. 9B). The PM-Na-FeC_x catalyst exhibited greater CO₂ consumption and lower CO production than the FeC_x catalyst. Additionally, the CO₂ oxidation of FeC_x primarily occurred around 554 °C, which is higher than the oxidation temperature of FeC_x in H₂O. This explains why H₂O oxidized iron carbide more completely than CO₂ at lower temperatures (such as 300 °C), as discussed in Section 3.1. Although high-pressure TPO experiments were not conducted, it is reasonable to speculate that the oxidation of iron carbide by H₂O and CO₂ may be more serious under high pressure than under normal pressure. However, our results can be used to explain the different degrees of oxidation for the two catalysts before and after the CO₂ hydrogenation. Compared with the pure iron carbide (FeC_x) catalyst, the sodium-modified iron carbide catalyst (PM-Na-FeC_x) retained less iron carbide after the CO₂ hydrogenation reaction. This is mainly because the Na additive promoted the oxidation of nanosized iron carbide by the product water and the reactant CO₂ during the CO₂ hydrogenation reaction, which is also consistent with the results from XRD and Mössbauer spectroscopy.

Fig. 9 The H₂O-TPO (A) and CO₂-TPO (B) profiles of FeC_x and PM-Na-FeC_x catalysts.

3.6 Steady-state reactivity of phase equilibrated catalysts with and without Na₂CO₃ dopants

In addition to affecting the phase transition of iron catalysts, sodium and other alkali metal dopants are often considered to influence the adsorption and dissociation of hydrogen on the surface of iron catalysts,^31,32 and Our H₂-TPD, CO-TPD and H₂-TPR experiments further support this conclusion (Fig. S10 and S11). To gain a deeper understanding of the effect of the Na₂CO₃ promoter on product distribution and related kinetics, a series of transient experiments with varying gas hourly space velocities (GHSV) were conducted over the catalysts. The phase transitions of the iron species were assumed to reach a pseudo-dynamic equilibrium once stable catalytic performance was achieved at each GHSV. The resulting reactivity and selectivity for the main products at steady state are presented in Fig. 10A and B.

Fig. 10 The reactivity of the steady-state FeC_x (A) and PM-Na-FeC_x (B) catalysts with varying gas hourly space velocities (GHSV), and the proposed schemes for product formation during CO₂ hydrogenation (C). Other reaction conditions: catalyst weight = 1 g, T = 325 °C, P = 20 bar, H₂/CO₂ = 3 mol mol⁻¹. The steady-state catalyst refers to a catalyst whose phase content reaches dynamic equilibrium. Usually, the catalyst was first operated at a GHSV of 6000 Nml g⁻¹ h⁻¹ for 8 h to allow the phase equilibrium state to be reached.

Over the FeC_x catalyst, as the GHSV increased, the conversion of CO₂ decreased, while the selectivity for CH₄ and CO increased linearly (Fig. 10A), even under extremely high GHSV conditions. This suggests that both the Reverse Water–Gas Shift (RWGS) reaction and the CO₂ to methane reaction were initial reactions over FeC_x during CO₂ hydrogenation. Unlike FeC_x catalysts, when the FeC_x catalyst was modified by Na₂CO₃ dopant, its selectivity for CH₄ decreased significantly. As the GHSV increased over the PM-Na-FeC_x catalyst, the selectivity for CH₄ further reduced, while the selectivity for CO increased rapidly, making CO the dominant product (Fig. 10B). This indicated that the Na₂CO₃ promoter inhibits the direct hydrogenation of CO₂ to CH₄ over the Na-modified iron catalyst. Moreover, it has been generally accepted that the formation of C₂₊ compounds occurs via the CO-FTS pathway with CO as an intermediate.^10,14,16 This was also confirmed by low space velocity experiments on the PM-Na-FeC_x catalyst. As the residence time increased (GHSV decreased), the CO was further converted to C₂₊ hydrocarbons. Therefore, the selectivity of CO decreased, and the selectivity for light olefins (C₂⁼–C₄⁼) increased. At a low GHSV of 1000 Nml g⁻¹ h⁻¹, over the PM-Na-FeC_x catalyst with a steady-state phase composition, CO₂ conversion increased to 30%, both the selectivity for methane and the selectivity for CO can be controlled below 20%, and the selectivity for light olefins can be boosted to more than 40%. However, low GHSV also led to more significant oxidation of the catalyst. When the operation returned to a GHSV of 6000 Nml g⁻¹ h⁻¹, the CO₂ conversion decreased, and the CO selectivity increased, as shown in Table S5. The deactivation caused by deep oxidation here may result from increased water vapor generation at low GHSV, which disrupted the dynamic balance of iron oxide and iron carbide inside the iron catalyst.

Considering the different product selectivity and CO₂ conversion on the iron carbide catalyst with and without sodium modification, along with the relationship between product formation discussed above, a simple schematic diagram of product formation during CO₂ hydrogenation over iron carbide catalysts was proposed, as shown in Fig. 10C. Over the Fe₅C₂ catalyst, both CH₄ and CO were primary products, produced from initial reaction paths (a) and (b). CO could be further converted into C₂₊ hydrocarbons (c) and CH₄ (d) through Fischer–Tropsch synthesis (FTS) reaction over the active FeC_x component. A higher Fe₅C₂ composition results in a higher CO₂ conversion and higher selectivity for CH₄ and C₂₊ hydrocarbons. Once the iron carbide (Fe₅C₂) was oxidized to iron oxide (Fe₃O₄), both the activities for the FTS (path c) and methanation (path a, d) were weakened, while the reactivity of RWGS became dominant (b). This led to a decrease in the conversion of CO₂ and CO, causing CO to accumulate and resulting in decreased selectivity for both CH₄ and all other hydrocarbons. As a result, the pure Fe₃O₄ catalyst exhibited very high selectivity for CO. When Na₂CO₃ and iron carbide formed a sufficiently intimate interface, H₂ molecular dissociation was supposed to be suppressed on the surface of the sodium-modified iron carbide catalyst (PM-Na-FeC_x),^12,14 resulting in a decreased CO₂ conversion. At the same time, limited hydrogen dissociation inhibited the rapid formation of CH₄ from direct hydrogenation of CO₂ (path a) and hydrogenation of CO (path d), making CO the primary product. The accumulated CO was further converted via the classic CO-FTS reaction, especially under low GHSV, thereby improving the selectivity for C₂₊ hydrocarbons. Moreover, Na₂CO₃ dopants also inhibited the deep hydrogenation of olefins on the surface of the catalysts, resulting in a higher selectivity for light olefins on the sodium-modified iron carbide catalyst.

In summary, two factors can explain the high CO selectivity over the sodium-modified iron carbide catalyst at high GHSV. Firstly, the dopant Na₂CO₃ accelerated the oxidation of iron carbide into iron oxide during CO₂ hydrogenation, which resulted in a decrease in the effective catalyst component responsible for olefin production. Secondly, Na⁺ was supposed to alter the dissociation of H₂ on the surface of iron carbide,^12,14 slowing down the reaction rate of the CO-based FTS reaction, which led to the accumulation of intermediate CO in CO₂ hydrogenation and an increased selectivity for CO in the final product. Decreasing the space velocity (GHSV) of the reactants can increase the conversion of CO₂ and CO, thereby reducing the selectivity for CO and maintaining a high olefin yield. However, the increased CO₂ conversion resulting from the lower GHSV also produced more water, disrupting the dynamic phase balance between iron carbide and iron oxide. This further led to oxidative deactivation of the iron carbide catalyst, especially the Na-modified iron carbide catalyst under a low GHSV condition.

In the future development of efficient iron-based catalysts for the hydrogenation of CO₂ to light olefins, studying the phase stability of the catalyst will be crucial. Avoiding oxidation and maintaining its carbonized state will undoubtedly be a key focus of further research. Our work revealed that the active center for olefin production originates from the interface between the iron carbide and the alkali metal additive. Therefore, it may be simpler to exclude the alkali metal additive and focus on the stability of the iron-based catalyst alone. Additionally, this work also confirmed that low space velocity accelerates the oxidation of the iron catalyst. This condition can be used to design accelerated oxidation experiments to speed up the screening of more suitable iron-based catalysts in future studies.

4 Conclusions

In this study, two pure-phase nanostructured Fe₅C₂ and Fe₃O₄ catalysts, with and without Na₂CO₃ dopant, were prepared and evaluated for CO₂ hydrogenation to light olefins. The results indicated a dynamic phase transition between the Fe₅C₂ and Fe₃O₄ phases in the iron-based catalysts, with varying activities during CO₂ hydrogenation. The efficiency of this transformation, between Fe₃O₄ and Fe₅C₂, is strongly influenced by particle size and reaction conditions. Nanosized Fe₅C₂ was easily oxidized to Fe₃O₄ by H₂O and CO₂, causing particle sintering, reduced CO₂ conversion, and increased CO selectivity. However, only small Fe₃O₄ particles were re-carburized to iron carbide by CO and H₂, partially restoring catalytic activity. Therefore, during the CO₂ hydrogenation reaction, with both oxidizing and reducing atmospheres, the Fe₅C₂ and Fe₃O₄ components eventually reach a dynamic phase equilibrium. This equilibrium results in catalytic behavior characterized by an initial decline in activity, followed by stabilization towards a pseudo-steady state.

By physically mixing Fe₅C₂ and Na₂CO₃ and controlling their contact distance, the results showed that closer proximity of Na₂CO₃ and Fe₅C₂ contributed to a higher yield of light olefins during CO₂ hydrogenation. However, Na₂CO₃ also promoted oxidation of the Fe₅C₂ and influenced the phase equilibrium inside iron-based catalysts. Besides, the Na₂CO₃ dopant was supposed to inhibit the dissociation and adsorption of H₂ on the surface of the iron carbide catalyst. This suppression reduced the rapid methanation of both CO₂ and CO, it also inhibited the further hydrogenation of olefins. Nevertheless, the inhibition of hydrogen dissociation also slowed down the CO-mediated FTS synthesis. Combined with the phase oxidation from Fe₅C₂ to Fe₃O₄, this effect ultimately increased CO selectivity over the Na₂CO₃-modified Fe₅C₂ catalyst. Operating at low gas space velocity (GHSV) can be used to reduce CO selectivity while maintaining high selectivity for light olefins by enhancing the extent of the CO-FTS reaction. However, the low GHSV also accelerated the oxidation of Fe₅C₂ to Fe₃O₄, resulting in oxidative deactivation of the iron carbide catalyst.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data herein includes the data collected and analyzed in this study, including the data underlying the main text and the supplementary information (SI). All data have been packaged as a single dataset and deposited in the Zenodo repository. The dataset is publicly available at the following DOI: https://doi.org/10.5281/zenodo.18310645. Supplementary Information: additional STEM, XRD, XPS, H₂O-TPO, CO₂-TPO, H₂-TPD, CO-TPD, H₂-TPR, and Mössbauer spectra of the catalysts, together with detailed analyses and interpretations of these characterization results, as well as catalyst reactivity data obtained from transient experiments. See DOI: https://doi.org/10.1039/d6ta00959j.

Acknowledgements

This project was conducted at the Chemical Engineering division, the Competence Centre for Catalysis (KCK), and the Centre for Process Chemical Engineering (CPE) at Chalmers University of Technology, and was performed in collaboration with Perstorp AB and Josefsson Sustainable Chemistry AB. We gratefully acknowledge the funding from Swedish Energy Agency (P49617-1) and NextGenerationEU (The Recovery and Resilience Facility, RRF).

References

1. O. A. Ojelade and S. F. Zaman, A review on CO₂ hydrogenation to lower olefins: Understanding the structure-property relationships in heterogeneous catalytic systems, J.CO₂ Util., 2021, 47, 101506.
2. D. Goud, R. Gupta, R. Maligal-Ganesh and S. C. Peter, Review of catalyst design and mechanistic studies for the production of olefins from anthropogenic CO₂, ACS Catal., 2020, 10, 14258–14282.
3. G. Centi, G. Laquaniello and S. Perathoner, Can We Afford to Waste Carbon Dioxide? Carbon Dioxide as a Valuable Source of Carbon for the Production of Light Olefins, ChemSusChem, 2011, 4, 1265–1273.
4. W. Liu, S. Cheng, H. S. Malhi, X. Gao, Z. Zhang and W. Tu, Hydrogenation of CO₂ to Olefins over Iron-Based Catalysts: A Review, Catalysts, 2022, 12, 1432.
5. B. Pawelec, N. Mota, R. Guil-López, J. L. G. Fierro and R. M. N. Yerga, Catalysts for the Conversion of CO₂ to Low Molecular Weight Olefins—A Review, Materials, 2021, 14, 6952.
6. R. Ye, J. Ding, W. Gong, M. D. Argyle, Q. Zhong, Y. Wang, C. K. Russell, Z. Xu, A. G. Russell, Q. Li, M. Fan and Y. G. Yao, CO₂ Hydrogenation to High-Value Products via Heterogeneous Catalysis, Nat. Commun., 2019, 10, 5698.
7. T. Numpilai, C. K. Cheng, J. Limtrakul and T. Witoon, Recent advances in light olefins production from catalytic hydrogenation of carbon dioxide, Process Saf. Environ. Prot., 2021, 151, 401–427.
8. W. Di, A. Achour, P. H. Ho, S. Ghosh, O. Pajalic, L. Josefsson, L. Olsson and D. Creaser, Modulating the Formation of Coke to Improve the Production of Light Olefins from CO₂ Hydrogenation over In₂O₃ and SSZ-13 Catalysts, Energy Fuel., 2023, 37, 17382–17398.
9. Y. Han, C. Y. Fang, X. W. Ji, J. Wei, Q. J. Ge and J. Sun, Interfacing with carbonaceous potassium promoters boosts catalytic CO₂ hydrogenation of iron, ACS Catal., 2020, 10, 12098–12108.
10. J. Zhu, P. Wang, X. Zhang, G. Zhang, R. Li, W. Li, T. P. Senftle, W. Liu, J. Wang, Y. Wang and A. Zhang, Dynamic structural evolution of iron catalysts involving competitive oxidation and carburization during CO₂ hydrogenation, Sci. Adv., 2022, 8, 3629.
11. L. Tang, L. He, Y. Wang, B. Chen, W. Xu, X. Duan and A. H. Lu, Selective fabrication of χ-Fe₅C₂ by interfering surface reactions as a highly efficient and stable Fischer-Tropsch synthesis catalyst, Appl. Catal., B, 2021, 284, 119753.
12. Q. Yang, V. A. Kondratenko, S. A. Petrov, D. E. Doronkin, E. Saraçi, H. Lund, A. Arinchtein, R. Kraehnert, A. S. Skrypnik, A. A. Matvienko and E. V. Kondratenko, Identifying Performance Descriptors in CO₂ Hydrogenation over Iron-Based Catalysts Promoted with Alkali Metals, Angew. Chem., Int. Ed., 2022, 61, 202116517.
13. B. Liang, H. Duan, T. Sun, J. Ma, X. Liu, J. Xu, X. Su, Y. Huang and T. Zhang, Effect of Na promoter on Fe-based catalyst for CO₂ hydrogenation to alkenes, ACS Sustain. Chem. Eng., 2018, 7, 925–932.
14. P. P. Paalanen and B. M. Weckhuysen, Carbon pathways, sodium-sulfur promotion and identification of iron carbides in iron-based Fischer-Tropsch synthesis, ChemCatChem, 2020, 12, 4202–4223.
15. C. G. Visconti, M. Martinelli, L. Falbo, A. Infantes-Molina, L. Lietti, P. Forzatti, G. Palo, E. Iaquaniello, B. Picutti and F. Brignoli, CO₂ hydrogenation to lower olefins on a high surface area K-promoted bulk Fe-catalyst, Appl. Catal., B, 2017, 200, 530–542.
16. M. Albrecht, U. Rodemerck, M. Schneider, M. Bröring, D. Baabe and E. V. Kondratenko, Unexpectedly efficient CO₂ hydrogenation to higher hydrocarbons over non-doped Fe₂O₃, Appl. Catal., B, 2017, 204, 119–126.
17. Y. Zhang, C. Cao, C. Zhang, Z. Zhang, X. Liu, Z. Yang, M. Zhu, B. Meng, J. Xu and Y. F. Han, The study of structure-performance relationship of iron catalyst during a full life cycle for CO₂ hydrogenation, J. Catal., 2019, 378, 51–62.
18. B. Y. Chen, G. Dobele, A. Plavniece, A. Volperts, L. Tamasauskaite-Tamasiunaite, E. Norkus, C. L. Chen and Y. C. Lin, Catalytic hydrogenation of CO₂ to light olefins by using K-doped FeC_x catalysts derived from the Fe-chitosan complex, Int. J. Hydrogen Energy, 2023, 48, 4276–4286.
19. J. I. Orege, N. Liu, C. C. Amoo, J. Wei, Q. Ge and J. Sun, Boosting CO₂ hydrogenation to high-value olefins with highly stable performance over Ba and Na co-modified Fe catalyst, J. Energy Chem., 2023, 80, 614–624.
20. Z. Feng, J. Zhao, F. E. Huggins and G. P. Huffman, Agglomeration and phase transition of a nanophase iron oxide catalyst, J. Catal., 1993, 143, 510–519.
21. D. E. Miser, E. J. Shin, M. R. Hajaligol and F. Rasouli, STEM characterization of phase changes and the occurrence of maghemite during catalysis by an iron oxide, Appl. Catal., A, 2004, 258, 7–16.
22. E. J. Shin, D. E. Miser, W. G. Chan and M. R. Hajaligol, Catalytic cracking of catechols and hydroquinones in the presence of nano-particle iron oxide, Appl. Catal., B, 2005, 61, 79–89.
23. D. Mahajan, P. Gütlich, J. Ensling, K. Pandya, U. Stumm and P. Vijayaraghavan, Evaluation of nanosized iron in slurry-phase Fischer − Tropsch synthesis, Energy Fuel., 2003, 17, 1210–1221.
24. C. Yang, H. Zhao, Y. Hou and D. Ma, Fe₅C₂ nanoparticles: A facile bromide-induced synthesis and as an active phase for Fischer–Tropsch synthesis, J. Am. Chem. Soc., 2012, 134, 15814–15821.
25. F. J. Berry, S. Skinner and M. F. Thomas, Mössbauer spectroscopic examination of a single crystal of Fe₃O₄, J. Phys.: Condens. Matter, 1998, 10, 215–220.
26. H. Yang, Y. Dang, X. Cui, X. Bu, J. Li, S. Li, Y. Sun and P. Gao, Selective synthesis of olefins via CO₂ hydrogenation over transition-metal-doped iron-based catalysts, Appl. Catal., B, 2023, 321, 122050.
27. M. K. Khan, P. Butolia, H. Jo, M. Irshad, D. Han, K. W. Nam and J. Kim, Selective conversion of carbon dioxide into liquid hydrocarbons and long-chain α-olefins over Fe-Amorphous AlO_x bifunctional catalysts, ACS Catal., 2020, 10, 10325–10338.
28. T. Riedel, H. Schulz, G. Schaub, K. W. Jun, J. S. Hwang and K. W. Lee, Fischer–Tropsch on iron with H₂/CO and H₂/CO₂ as synthesis gases: the episodes of formation of the Fischer–Tropsch regime and construction of the catalyst, Top. Catal., 2003, 26, 41–54.
29. M. Hermanek, R. Zboril, M. Mashlan, L. Machala and O. Schneeweiss, Thermal behavior of iron (II) oxalate dihydrate in the atmosphere of its conversion gases, J. Mater. Chem., 2006, 16, 1273–1280.
30. J. Zhu, M. Mu, Y. Liu, M. Zhang, G. Zhang, Z. Cheng, B. H. Yin, A. C. Yip, C. Song and X. Guo, Unveiling the promoting effect of potassium on the structural evolution of iron catalysts during CO₂ hydrogenation, Chem. Eng. Sci., 2023, 282, 119228.
31. J. B. Li, H. F. Ma, H. T. Zhang, Q. W. Sun, W. Y. Ying and D. Y Fang, Direct production of light olefins from syngas over potassium modified Fe–Mn catalyst, React. Kinet. Mech. Catal., 2014, 112, 409–423.
32. H. Wan, B. Wu, C. Zhang, H. Xiang and Y. Li, Promotional effects of Cu and K on precipitated iron-based catalysts for Fischer–Tropsch synthesis, J. Mol. Catal. A:Chem., 2008, 283, 33–42.

---