Identifying mechanistic differences between co-fed CO₂ hydrogenation and reactive CO₂ capture using Ru and Pd dual function materials

Abstract

Dual function materials (DFMs) enable reactive carbon capture (RCkeC), an intensified approach to carbon dioxide capture and utilization for cost and energy input reductions. Yet, there is a fundamental lack of understanding of mechanisms around CO₂ adsorption and subsequent conversion on these materials, hindering further development. Herein, we investigated several supported alkaline metal oxides for their CO₂ adsorption characteristics to find that Na/Al₂O₃ had the highest CO₂ adsorption capacity, accompanied by a variety of CO₂ adsorption geometries as identified by in situ DRIFTS and computational modeling. The addition of catalytic metals (Ru, Pd) increased the adsorption capacity of Na/Al₂O₃ without altering binding modes. In the subsequent reactive desorption step, acetate and formate intermediates were observed. Notably, this mechanistic investigation identified that the formation of acetate species was unique to RCC on a DFM, as these species were not observed in co-fed hydrogenation over the DFM or RCC over a Na-free catalyst.

---

Broader context

Reactive carbon capture (RCC) seeks to address an overarching challenge in CO₂ utilization approaches: typical carbon capture and utilization (CCU) technologies require large amounts of energy for the multiple steps of purification, compression, and transportation of CO₂ after capture for conversion at a remote facility. To reduce the energy demand and break through current limitations of CCU, the RCC process instead employs dual function materials (DFMs) to enable capture and conversion in a single reactor. While recent literature has focused on the discovery of new DFM formulations and novel RCC processes, this study seeks to provide a comprehensive understanding of the key interactions between CO₂ and DFM components, and how reactive desorption mechanisms are affected. Our findings reveal that RCC catalytic mechanisms cannot be directly extrapolated from traditional co-fed hydrogenation mechanisms, providing an important learning on how the scientific community should approach the design of next generation DFMs with improved performance.

Introduction

The use of waste CO₂ as a carbon feedstock to produce fuels and chemicals through CO₂ recycling (i.e. capture and utilization, CCU)¹ holds great potential as a means of chemical storage to combat intermittency of renewables, a pathway commonly dubbed “power-to-X”.^2–4 However, typical CCU technologies require large amounts of energy for the multiple steps of purification, compression, and transportation of CO₂ after capture for conversion at a remote facility.⁵ To reduce the energy demand and break through current limitations of CCU, multi-functional materials that allow capture and conversion to occur in a single reactor through a process dubbed reactive carbon capture (RCC) have been the topic of recent research.

One such catalytic technology platform employs dual function materials (DFMs), which are composed of sorbents and catalysts co-dispersed on the same high surface area carrier. The sorbent enables selective capture of CO₂ from a gas stream (including dilute CO₂ concentrations) and the catalyst component subsequently converts the adsorbed CO₂ upon introduction of a reactive gas (typically H₂) in situ.^6–8 The product from the most established DFM, comprised of Ru and/or Ni with an alkaline sorbent, is methane via the CO₂ methanation reaction.^9,10 While renewable methane would be an excellent transition fuel that can leverage currently available infrastructure, it is economically noncompetitive relative to inexpensive fossil methane (averaging $2.19 per MMBTU in 2024 in the US¹¹). These challenges motivate the design and investigation of DFMs that enable CO₂ capture and conversion to more valuable and more useful C1 products, like CO or methanol (average price of methanol was $35.66 per MMBTU in 2024¹²). These products can be used in the chemical industry or upgraded to high energy density synthetic fuels for heavy duty vehicles, marine, and aviation.

A survey of DFM formulations reveals a variety of sorbent-catalyst combinations that can perform RCC to methane, CO, and/or methanol.^8,13,14 In stark contrast to fundamental catalysis science and mechanistic studies of conventional co-fed CO₂ hydrogenation reactions (which include the Sabatier reaction, CO₂ + 4H₂ → CH₄; reverse water gas shift, CO₂ + H₂ → CO; and methanol synthesis, CO₂ + 3H₂ → CH₃OH),^15–20 there is a lack of depth in the fundamental understanding of how CO₂ binds to DFMs, and importantly, how the subsequent mechanisms during reactive desorption are affected. Furthermore, recent explorations of novel DFMs to achieve more valuable C1 products like MeOH have shown that co-fed CO₂ hydrogenation activity does not necessarily directly translate to RCC performance, suggesting that there are important mechanistic differences between the two modes of operation.²¹ Such differences were also suggested by Moreno Bravo et al., when studying Ru DFMs composed of rare earth oxides.²² While possible mechanisms over DFMs have been proposed,^23–25 a systematic study is needed to understand: (1) the role of the alkali sorbent site in CO₂ binding, (2) the subsequent effect of catalyst metal identity on the resulting transformation of the adsorbed CO₂, and (3) the differences in adsorbed intermediates between the RCC and co-fed hydrogenation mechanisms.

Herein, we first characterized several Al₂O₃-supported sorbents for their capacity and CO₂ binding geometries. We then characterized Ru and Pd DFMs with the same sorbent component (motivated by the breadth of RCC and co-fed CO₂ hydrogenation literature around these two catalytic metals) to determine the effect of the metal catalyst on CO₂ capacity and binding, followed by operando DRIFTS characterization of observable intermediates during simulated reactive desorption. Observed CO₂ binding geometries and resulting intermediates were further verified through molecular modeling. Comparisons between co-fed hydrogenation over the Ru DFM and RCC on Na-free Ru/Al₂O₃ identified that the formation of acetate surface species was unique to RCC on the DFM.

Experimental

Material synthesis

Supported sorbents were prepared by introducing alkali (Na, K) and alkaline earth (Mg, Ca) metal oxides via incipient wetness impregnation of γ-Al₂O₃ (Puralox TH100/150, Sasol). Aqueous solutions of Na₂CO₃, K₂CO₃, Ca(NO₃)₂·4H₂O, and Mg(NO₃)₂·6H₂O salts (Sigma Aldrich) were used for impregnation onto γ-Al₂O₃ powder to achieve a metal : Al₂O₃ molar ratio of 0.202. These salts were chosen to be representative of synthesis routes used by many previous RCC reports.^8,21,25–27 The impregnated powders were dried at 120 °C overnight (>12 h) in static air. Decomposition of the carbonate or nitrate precursors was performed using calcination at 400 °C for 4 h before characterization.

Ru and Pd DFMs were prepared by modifying Na/Al₂O₃ (prepared as described above) through incipient wetness impregnation using aqueous solutions of Ru(NO)(NO₃)₃ (Thermo Scientific Chemicals) and Pd(NO₃)₂·2H₂O (Sigma Aldrich) with a catalytic metal to sorbent metal molar ratio of 0.02 (equivalent to ca. 2 wt% Ru or Pd). The resulting DFMs were then dried at 50 °C overnight in static air. The catalytic metals were activated through a pre-reduction at 300 °C for 4 h in H₂ before characterization.

A non-DFM Ru/Al₂O₃ catalyst was prepared in the same manner by utilizing an aqueous solution of Ru(NO)(NO₃)₃ for impregnation of γ-Al₂O₃. The sample was dried at 120 °C overnight in static air and activated through a pre-reduction at 300 °C for 4 h in H₂.

Characterization

CO₂ and CO chemisorption were measured using an Autosorb-1C gas sorption analyzer (Quantachrome Instruments). Approximately 100–150 mg of material was diluted with 1 g of quartz chips in a u-shaped analysis tube. For CO₂ chemisorption, sorbent materials were calcined at 400 °C in N₂ for 8 h, and DFMs were reduced at 300 °C in H₂ for 8 h. The sample was then evacuated for 8 h prior to analysis. CO₂ chemisorption was measured at 100, 200, and 300 °C between 80 and 760 Torr in 50 Torr increments. The site density of strong CO₂ chemisorption was determined by the difference of the total and weak isotherms, extrapolated to zero pressure. For CO chemisorption, DFMs were reduced at 300 °C in H₂ for 4 h, followed by evacuation for 8 h prior to analysis. CO adsorption isotherms were measured at 40 °C between 80 and 760 Torr in 50 Torr increments. The site density of total CO uptake from the combined isotherm, extrapolated to zero pressure, was used to determine the average particle size of Ru and Pd. A metal (M) to CO binding stoichiometry, M:CO, of 1.5 was used to represent the known distribution of atop (M–CO, 1 : 1) and bridged (M–CO–M, 2 : 1) CO binding on these metals. The chemisorption data has an estimated error of ±4.1% based on the reproducibility error of the calibration standard provided by Quantachrome.

DRIFTS analysis

In situ and operando diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) experiments were performed using a Thermo Scientific Nicolet iS50R FT-IR spectrometer equipped with a Harrick Praying Mantis high temperature reaction chamber attachment with CaF₂ windows. The DRIFTS spectra and backgrounds were collected at a resolution of 4 cm⁻¹ with 128 scans. The samples were loaded into the sample holder on top of a small bed of quartz chips and pre-treated in situ prior to analysis. Al₂O₃-supported sorbents were calcined under flowing He for 4 h at 400 °C. DFMs and Al₂O₃-supported catalysts were reduced under flowing 50% H₂/He for 4 h at 300 °C. Backgrounds were taken at 300 °C in He and 50% H₂/He to be subtracted from the adsorption and reactive desorption steps, respectively. Flowrates were kept at 40 sccm and ramp rates at 5 °C min⁻¹, unless otherwise specified.

CO₂ adsorption was conducted at 300 °C with 5% CO₂/He for 30 min, with spectra collected in 5 min intervals. Spectra were also collected in 5 min intervals during a 1 h purge step under flowing He following the adsorption step. For operando analysis, reactive desorption was performed on DFMs and Al₂O₃ by switching to 50% H₂/He for 1 h, with spectra again being collected in 5 min intervals. Effluent gas of the DRIFTS cell was analyzed using a downstream on-line FT-IR (600 HSC from California Analytical Instruments) to monitor product formation.

Operando co-fed hydrogenation was conducted in the DRIFTS cell at 300 °C with a gas mixture of 10% CO₂, 40% H₂, balance He, with a background spectrum collected in 40% H₂/He. Spectra were collected after a 2-h hold at reaction conditions, after which CO₂ flow was stopped. Spectra were then collected at 5 min intervals to monitor the evolution of surface species over 1 h. Effluent gas of the DRIFTS cell was again analyzed using a downstream on-line FT-IR to monitor product formation.

Computational methods

Density functional theory (DFT) calculations were implemented using the vienna ab initio simulation package (VASP).^28–30 The generalized gradient approximation (GGA) through the Perdew–Burke–Ernzerhof (PBE) functional was used to approximate the exchange–correlation energy.^31,32 Electron–ion interactions were captured by projector augmented-wave (PAW) potentials, with valence wavefunctions expanded in a plane-wave basis set with an energy cutoff of 500 eV.³³ Gaussian smearing with a smearing width of 0.01 eV was employed. Spin-polarized calculations were carried out for systems involving radical adsorbates. Electronic relaxation was converged to a tolerance of 1 × 10⁻⁶ eV and the norm of force convergence was set to 0.03 eV Å⁻¹. Atomistic construction of hydrated γ-Al₂O₃ surfaces denoted as Al₂O₃(110)_2H2O and Al₂O₃(100)_2H2O and Na-modified alumina are described in the SI.

The adsorption energy (ΔE_ads,i) for adsorbate (i) is defined as

ΔE_ads,i = E_i/slab − E_slab,r − E_i(g) (1)

where E_i/slab is the total energy of the slab with adsorbate, E_slab,r is the energy of the relaxed slab without adsorbate, and E_i(g) is the energy of the free, gas-phase molecule. E_i(g) was computed by placing one “i” molecule in a 16 × 17 × 18 ų box. Gas-phase molecules used as reference states in this work include neutral species (CO₂, H₂O) and radical species (HCOO, CH₃COO). The corresponding optimized structures and energies are reported in Table S7 and were created using VESTA.³⁴

Vibrational modes and harmonic frequencies of adsorbed species were obtained using perturbation theory without symmetry.³⁵ Scaling factors were not applied because the calculated scaling value for CO₂ was near unity (Table S8). Atomic simulations environment (ASE) was used to visualize and confirm the vibrational modes.³⁶ Frequencies for carbonate (CO₃*) and bicarbonate (HCO₃*) species were computed by relaxing CO₂O_s and CO₂O_sH_s atoms, where ‘s’ denotes surface atoms. Only adsorbate atoms were relaxed in vibrational calculations of CH₃COO and HCOO; all substrate atoms were fixed. Dispersion corrections to energies were not applied as this analysis focuses on strongly chemisorbed species likely to persist at 300 °C for which binding is dictated by electrostatic interactions with the surface.

Born effective charges and the displacement vectors are used to determine the intensities (I) of infrared active vibrational modes. Intensities are calculated as

where e_β(l) is the normalized vibrational eigenvector of the ω-th mode calculated for different atoms (l) of the system, α and β indicate the Cartesian polarizations, and is the Born effective charge tensor of the l-th atom.³⁷

Results and discussion

Adsorption characteristics of Ru and Pd-based DFMs

Most DFMs employ group 1 alkali (G1-Alk) or group 2 alkaline earth (G2-Alk) metal oxides as their main sorbent component, which when dispersed on a high surface area oxide support (like Al₂O₃)^38,39 offers an intermediate CO₂ binding strength that retains CO₂ on the surface during purge and heating steps of RCC processes (e.g., 200–400 °C) and is also reactive upon the introduction of H₂. Prior literature teaches that adsorption capacity, strength, and the subsequent product, can vary dramatically depending on the identity of the sorbent metal.^21,40 We began our investigation by comparing the adsorption characteristics of Al₂O₃-supported G1-Alk and G2-Alk metals, standardized by a Alk : Al₂O₃ molar ratio of 0.202, equivalent to ca. 10% CaO/Al₂O₃, a standard weight loading used in recent DFM reports (elemental analysis reported in Table S1). Chemisorption analysis with the in situ DRIFTS spectra for this series of sorbents (Fig. S1) revealed Na/Al₂O₃ has the highest gravimetric adsorption capacity, the greatest efficiency in terms of molar adsorption capacity, and a diversity of CO₂ binding modes (Fig. S2). As previous RCC reports also commonly chose Na as the sorbent material,^27,41–44 Na/Al₂O₃ was selected in this work as the most interesting candidate for a deeper spectroscopic study.

It is noted that there is currently no clear agreement in DFM literature on the exact structure and phase of alumina-supported Na species; however, we are able to speculate that the majority of Na sites are well dispersed based on our chemisorption and DRIFTS data. First, following analysis by Keturakis et al., we believe to be at just below a monolayer of Na. Keturakis et al. ascribed 5% “Na₂O” as the weight loading for monolayer formation on 200 m² g⁻¹ γ-Al₂O₃, resulting in a Na density of 5.11 Na atoms per nm².³⁹ Our DFMs contain 2.7 wt% Na (3.6% “Na₂O”) on 142 m² g⁻¹ γ-Al₂O₃, resulting in 4.8 Na atoms per nm². These materials are below the expected monolayer coverage, which would favor high dispersion of Na species after activation. We consider that Na is likely to exist in a combination of oxide or hydroxide forms (NaO_x, NaAlO_x, NaOH) and form surface carbonates and/or bicarbonates upon CO₂ adsorption. Stoichiometrically, this would result in molar capacities (mol_CO2 mol_Na⁻¹) ranging from 1 (1 : 1 Na : CO₂ binding, as in NaHCO₃ or NaAlCO₃(OH)₂ formation) to 0.5 (2 : 1 Na : CO₂, as in Na₂CO₃ formation) in a perfectly dispersed material. Comparatively, Na/Al₂O₃ exhibited a strong chemisorption capacity of 0.22 mol_CO2 mol_Na⁻¹ at 100 °C which we equate to between 22 and 44% dispersion, indicating that there may a portion of clustered, inaccessible Na sites or the existence of lower binding strength sites that do not strongly bind CO₂ at 100 °C. Realistically, our DRIFTS data (Fig. S2c) at 300 °C showed a mixture of polydentate (Na:CO₂ > 2) carbonates, conceptually favored by larger Na (or Na-O-Al) clusters, as well as bridged bidentate (Na:CO₂ = 2) and chelating bidentate (Na:CO₂ = 1) carbonates, favored by smaller particles. Thus, Na speciation is likely a mixture of well-dispersed oxides and hydroxides with some portion that is either inaccessible in clusters or binds CO₂ too weakly.

Na/Al₂O₃ was modified with two representative CO₂ hydrogenation catalysts, Ru and Pd. Ru is commonly used in DFM formulations to achieve methanation, showing excellent redox behavior (i.e., rapid reduction following oxidation) and long-term stability.^26,27 Pd is a versatile metal that catalyzes reverse water gas shift (RWGS) and methanol formation from CO₂, and we employ it here to study CO formation over a DFM.^45–47 The catalysts were introduced via serial incipient wetness impregnation, targeting a molar ratio of catalyst : sorbent metal of 0.1, resulting in ca. 2 wt% target loading (analyzed to be 2.3% Ru and 2.5% Pd). After reduction at 300 °C, CO chemisorption at 40 °C was employed to determine the average particle sizes. As expected, the low wt loadings resulted in small average particle sizes of 0.70 and 1.1 nm for Ru+Na/Al₂O₃ and Pd+Na/Al₂O₃, respectively. We note that DFMs for CO and MeOH synthesis based on Cu are exciting new additions to the RCC field,²¹ but these are less well understood compared to Ru- and Pd-based DFMs, and thus, are not the most suitable catalysts for the goals of this study.

The strongly chemisorbed CO₂ on the resulting DFMs (termed, Ru+Na/Al₂O₃ and Pd+Na/Al₂O₃) were quantified over a range of temperatures (Fig. 1A). At all adsorption temperatures, the DFMs displayed enhanced adsorption capacity relative to the metal-free Na/Al₂O₃ (Fig. 1B). The Ru-DFM exhibited a greater increase in strong CO₂ chemisorption which was attributed to two factors: (i) the reported behavior of Ru to promote the decomposition of Na₂CO₃ precursor salts during DFM pretreatment, activating more Na adsorption sites,^25,48 and (ii) additional CO₂ adsorption on Ru/RuO_x during the capture step. Considering these two factors for the Pd-DFM, the more modest increase in CO₂ adsorption capacity versus Na/Al₂O₃ may be caused by a combination of less effective participation of Pd to activate Na sites during pretreatmentrelative to Ru and a limited CO₂ adsorption on Pd/PdO_x surface species, which has been described to be nondissociative and weak.^49,50

Fig. 1 (A) Strong CO₂ chemisorption for Na/Al₂O₃ (yellow), Ru+Na/Al₂O₃ (brown), and Pd+Na/Al₂O₃ (purple) over a range of adsorption temperatures (100–300 °C). (B) Percent increase in strong CO₂ chemisorption compared to Na/Al₂O₃.

Characterization of CO₂ binding geometries on Ru- and Pd-DFMs was performed through in situ DRIFTS. The resulting spectra after pre-treatment, CO₂ exposure, and He purge are displayed in Fig. 2 with the spectrum after CO₂ adsorption on Na/Al₂O₃ included for comparison. The broad absorbances observed were deconvoluted to make precise peak assignments, associate them to adsorption geometries (Table 1), and detect any perturbations resulting from the presence of a catalytic metal. We note that exact assignment between chelating and bridged bidentate carbonate geometries to the peaks fit at 1618/1311 cm⁻¹ and 1669/1378 cm⁻¹ on the Na/Al₂O₃ – which we will refer to as CO₃*(B-I) and CO₃*(B-II), respectively – is difficult as metrics for determining the strength of binding were found to be contradictory. For instance, CO₃*(B-II), located at higher wavenumbers and showing the greatest decrease in intensity after an inert purge (Fig. S2), would be expected to correspond to a weaker bound bidentate geometry. Yet, the Δν₃ splitting of the CO₃* (B-II) species – a heuristic that can be used to determine the strength of surface carbonate binding – is lower at 291 cm⁻¹ compared to 307 cm⁻¹ of the CO₃*(B-I) species, indicating a marginally stronger bound species. It is more conservative to conclude that these bidentate geometries are quite similar in binding strength and refrain from making exact assignments based on this dataset. Peaks at 1554 and 1340 cm⁻¹ correlated to a polydentate geometry (CO₃*(P)), and a small shoulder at 1755 cm⁻¹ corresponded to Na₂CO₃ nanoparticles.³⁹

Fig. 2 In situ DRIFTS spectra of (A) Na/Al₂O₃, (B) Ru+Na/Al₂O₃, and (C) Pd+Na/Al₂O₃ after 30 min adsorption with 5% CO₂/He and 30 min purge with He at 300 °C. Catalyst-containing samples were reduced for 4 h in 50% H₂/He at 300 °C before analysis; Na/Al₂O₃ was pre-treated for 4 h in He at 400 °C. Larger formatting is provided in SI (Fig. S3).

Table 1 Assignments of observed peaks in DRIFTS spectra of Na/Al₂O₃, Ru+Na/Al₂O₃, and Pd+Na/Al₂O₃ after 30 min of adsorption (5% CO₂/He at 300 °C), followed by an inert purge in He (displayed in Fig. 2). Asymmetric and symmetric vibrational splitting is also reported (Δν₃)

Surface species	Na/Al2O3	Ru+Na/Al2O3	Pd+Na/Al2O3
CO3* (B-I)	1618 cm^−1	1614 cm^−1	1614 cm^−1
	1311 cm^−1	1320 cm^−1	1305 cm^−1
	(Δν3 = 307 cm^−1)	(Δν3 = 294 cm^−1)	(Δν3 = 309 cm^−1)
CO3* (B-II)	1669 cm^−1	1659 cm^−1	1657 cm^−1
	1378 cm^−1	1371 cm^−1	1367 cm^−1
	(Δν3 = 291 cm^−1)	(Δν3 = 288 cm^−1)	(Δν3 = 290 cm^−1)
CO3* (P)	1554 cm^−1	1559 cm^−1	1549 cm^−1
	1340 cm^−1	1349 cm^−1	1330 cm^−1
	(Δν3 = 214 cm^−1)	(Δν3 = 211 cm^−1)	(Δν3 = 219 cm^−1)
Carbonyls	—	1990 cm^−1	2045 cm^−1
		1859 cm^−1	1957 cm^−1
		1799 cm^−1

---

Further understanding of potential CO₂ binding geometries on Na/Al₂O₃ was obtained through DFT calculations. Na/Al₂O₃ models were constructed by adding a Na₂O molecule to γ-Al₂O₃ surfaces (Al₂O₃(110)_2H2O and Al₂O₃(100)_2H2O, Fig. S4–S6, Table S2). We note that simulations of CO₂ bound to Al₂O₃(110)_2H2O and Al₂O₃(100)_2H2O showed reasonable agreement between the computed IR spectra of various CO₂ binding geometries and the observed spectra on γ-Al₂O₃ (Fig. S7 and Table S3). In the most stable configurations, Na₂O on Al₂O₃(110)_2H2O and Al₂O₃(100)_2H2O transformed to Na₂O(H) through migration of a H* from a surface Al-hydroxyl group (Fig. S8). Simulated IR spectra of unique CO₂ adsorption geometries, varying in CO₂ coordination to surface Na, Al, or to both Na and Al surface atoms on Na/Al₂O₃ models (Table S4), showed that various bicarbonate, bidentate, and polydentate species were well-aligned with the experimental spectrum (Fig. S9).

The spectra of both DFMs show broad absorbances at 1700–1500 cm⁻¹ and 1400–1300 cm⁻¹, similar to Na/Al₂O₃, with minor perturbations to the exact peak locations. On Ru+Na/Al₂O₃, the symmetric vibration of the CO₃*(B-I) species (lower wavenumber peak) shifted, resulting in a Δν₃ splitting value of 294 cm⁻¹, compared to 307 cm⁻¹ on the Na/Al₂O₃, indicating a marginally stronger carbonate-surface binding strength. CO₃*(B-II) peaks both shifted to lower wavenumbers, with a Δν₃ value of 288 cm⁻¹ (291 cm⁻¹ on Na/Al₂O₃). The CO₃*(P) shifted to higher wavenumbers, yielding a similar Δν₃ value of 211 cm⁻¹ for Ru-DFM compared to 214 cm⁻¹ for Na/Al₂O₃. The Ru-DFM also displayed strong interactions between Ru and carbonyl species, presumably formed from CO₂ dissociation during the adsorption step, as evidenced by the complex feature between 2000 and 1750 cm⁻¹. We attributed peaks observed at 1990 and 1799 cm⁻¹ to a mixture of linear, di-, and bridged carbonyls on Ru.⁵¹ The additional peak at 1859 cm⁻¹ could be attributed to a Ru carbonyl that is further stabilized by the presence of Na species. Such phenomenon has been reported with other noble metals (e.g., Au, Pd), where the presence of Na⁺ (and other alkali metal cations) induced red-shifted IR peaks of carbonyls associated with the noble metal site.⁵² We further confirmed this assignment using CO DRIFTS at 300 °C on Ru+Na/Al₂O₃ and Na-free Ru/Al₂O₃ (Fig. S10), which revealed a peak near 1850 cm⁻¹ only on the Ru+Na/Al₂O₃ material.

For Pd+Na/Al₂O₃, no major perturbations were detected for the CO₃*(B-I) peak locations (1614/1305 cm⁻¹ versus 1618/1311 cm⁻¹ on Na/Al₂O₃), preserving a Δν₃ value of 309 cm⁻¹. A greater shift was observed on the CO₃*(B-II) species to lower wavenumbers as also observed on the Ru-DFM. The CO₃*(P) peaks shifted to lower wavenumbers as well; however, neither shifts associated with the two geometries resulted in significant differences to the splitting values. The Pd component of the DFM also exhibited interactions with in situ generated CO, as detected by peaks at 2045 and 1957 cm⁻¹.⁵³ Based on the low peak intensities, however, these Pd-(CO)_x species are much less abundant than for Ru+Na/Al₂O₃, consistent with prior reports of low CO₂ and CO binding energy on Pd.^49,50

Reactive desorption over DFMs and sorbent-free catalysts

The subsequent transformation or decomposition of the surface carbonates during a simulated reactive desorption step (i.e., hydrogenation) of an RCC cycle were monitored using operando DRIFTS. Following the adsorption and purge steps as outlined in the previous section, 50% H₂/He was introduced to the DRIFTS reactor at ambient pressure. Upon the introduction of H₂ to the CO₂-laden Ru+Na DFM (Fig. 3A), there was rapid consumption of the Ru-carbonyls, CO₃*(B-II), and CO₃*(B-I) species, correlated with a sharp rise in the signal for gas-phase CH₄ in the reactor effluent (Fig. 3B). The carbonyl and carbonate species were almost fully consumed by the end of H₂ exposure (1 h), and the Ru+Na DFM maintained selective production of CH₄ without any CO observed. A peak at 1460 cm⁻¹ also rapidly formed upon the introduction of H₂, presumably associated with an intermediate species. While this species was largely consumed, some small fraction persisted at the end of H₂ exposure, along with peaks at ∼1575 and ∼1350 cm⁻¹. Based on literature reports for traditional CO₂ methanation over a Ru/Al₂O₃ catalyst (i.e., not a DFM and not operated with a cyclic CO₂–H₂ feed), an Al₂O₃-associated formate (HCOO*) intermediate would be expected, exhibiting peaks at 1593, 1392, and 1375 cm⁻¹.¹⁵ The observed peaks here do not align with these expected wavenumber positions, and with CO₂ binding dominated at Na sites as indicated by the data in Fig. S2, we do not expect an HCOO*-Al intermediate to form under these reaction conditions. Instead, it may be possible that a formate is associated with the Na sites. Previous reports assign Na-formate with peaks at 1620–1567 and 1366–1377 cm⁻¹, which would lie under the large peaks observed in our spectra, and are also in the range for polydentate surface carbonates noted above.⁵⁴ Thus, an exact assignment for a Na-formate species is difficult. Further identification of the observed intermediate(s), including those associated with 1460 cm⁻¹, was explored with computational modeling (vide infra). As mentioned, the peaks at 1575, 1350, and 1460 cm⁻¹ persist at low intensities even after 1 h of reactive desorption, indicating that the Na sites that are occupied by the assocciated intermediates are not available for CO₂ adsorption in subsequent cycles. This is in line with the larger DFM literature, which often show a decrease in CO₂ adsorption capacity in the first few cycles. It is worth noting, however, that extended aging studies of Ru+Na DFMs have been well reported, showing no decrease in methanation activity over 50+ cycles under “clean CO₂” adsorption conditions (i.e., no contaminants such as H₂O and O₂ present).²⁷ We thus believe that the residual build up of intermediates observed here would neither cause cycle-to-cycle deactivation in the studied conditions nor hinder the mechanistic characterization presented in this work.

Fig. 3 DRIFTS spectra of reactive desorption following CO₂ adsorption and inert purge (as seen in Fig. 2) on (A) Ru+Na/Al₂O₃ and (C) Pd+Na/Al₂O₃ with accompanying downstream product analysis, (B) and (D), respectively. Dark gray spectra were taken at the end of a 30 min He purge following the CO₂ adsorption step; light gray spectra were taken following exposure to H₂ in 5 min intervals; and blue spectra were taken after 1 h of continuous H₂ exposure. All steps were carried out at 300 °C.

The simulated reactive desorption step progressed similarly on Pd+Na/Al₂O₃, with the introduction of H₂, there was a sharp and selective release of CO (Fig. 3C and D), correlated with the initial rapid consumption of each surface carbonate species. However, even after 1 h of H₂ exposure, a large portion of the observed peak intensity for surface carbonates remained. It is expected that the kinetics for CO production are slow at this low temperature of 300 °C, since thermodynamics dictate that the RWGS reaction is generally operated above 400 °C. With the persistence of the carbonate species, it is difficult to clearly identify specific intermediate structures that may lie under those peaks. However, a broad peak around 1460 cm⁻¹ appeared rapidly with the introduction of H₂, indicative of a surface intermediate similar to that formed on the Ru+Na DFM. It has been proposed that the mechanism of CO₂ hydrogenation on Pd/Al₂O₃ progresses through a CO* intermediate (derived from other intermediates such as formate^47,53 or carboxyl⁴⁵). Here, we observed that adsorption of CO on Pd is not favored (as evidenced by the low intensity of peaks between 2100 and 1900 cm⁻¹), likely resulting in desorption of gaseous CO (the major product) rather than further hydrogenation of CO* to CH₃OH or CH₄.

On both DFMs, the peak at 1460 cm⁻¹ may correlate best with a surface acetate intermediate (CH₃COO*), reported to form by reaction between hydroxyl groups (OH*) and carbonyls (CO*), hydrogenation of CO* (to form H–C–O* intermediates), or between CO* and methyl groups (CH₃*).^55–57 Interestingly, acetate intermediates are more often associated with CO methanation or Fischer–Tropsch rather than CO₂ methanation, which we also confirmed by comparing DRIFTS spectra during co-fed CO₂ hydrogenation and CO hydrogenation over non-DFM Ru/Al₂O₃ (Fig. S11); importantly, a 1460 cm⁻¹ peak was only observed during CO hydrogenation. To further confirm the assignment of an acetate species, DFT calculations were performed to compute infrared active bands of acetate and formate adsorbed on Na-free (Fig. S12, Table S5) and Na-modified hydrated alumina surfaces (Fig. S13, Table S6). Stable adsorbed acetate and formate species on Na/Al₂O₃(110)_2H2O are shown in Fig. 4A–F. Acetates coordinated at Na, both Na and Al, and Al (Fig. 4A–C, respectively), all displayed vibrational signatures between 1400 and 1500 cm⁻¹, corresponding well to the observed 1460 cm⁻¹ peak. Among the computed species, Al-coordinated CH₃COO* was the most stable (−1.99 eV, Table S6) and formed chelating bidentate motifs, exhibiting major peaks at 1530 and 1462 cm⁻¹ associated with υ_as(OCO) vibration and υ_s(OCO) stretch mixed with CH₃ bending, respectively. The bridged acetates coordinated to two surface Na sites and that coordinated to a Na and Al site exhibited peaks at 1448 and 1430 cm⁻¹, respectively, indicating that there is likely a mixture of CH₃COO* geometries and site coordination. In contrast, HCOO*—regardless of coordination and geometry—did not display bands in the 1400–1500 cm⁻¹ range (Fig. 4D–F), further corroborating that the observed 1460 cm⁻¹ peak in the DRIFTS spectrum can be assigned to a surface acetate. Formates did, however, exhibit peaks at ca. 1600 and ca. 1350 cm⁻¹, corresponding well to the 1575 cm⁻¹ and 1350 cm⁻¹ envelopes observed in DRIFTS. HCOO* geometries also had adsorption energies similar to CH₃COO* (Table S6), indicating that they would be stable on the DFM surface and suggesting that they are likely present in the observed DRIFTS spectra. It is worth mentioning that CO₃*(P) species were also assigned to similar wavenumbers (1559/1349 cm⁻¹); however, we conclude that their contributions are likely minor compared to that of the hydrogenated intermediates, given that their calculated adsorption energies are notably higher (−0.99 eV versus −1.99 eV, Table S4). In summary, experimentally observed envelopes at 1575, 1460, and 1350 cm⁻¹ during reactive desorption correlated well with predicted IR active modes of adsorbed CH₃COO* and HCOO*, verifying the formation and presence of these key intermediates during RCC.

Fig. 4 Computed vibrational spectra and images of the corresponding structures for Na-coordinated, Na, Al-coordinated, and Al-coordinated acetate (CH₃COO*) (A–C, respectively) and formate (HCOO*) (D–F, respectively) on Na/Al₂O₃(110)_2H2O. A black dashed line indicates the 1460 cm⁻¹ wavenumber observed in the simulated RCC experiment. Atom colors: O-red, Al-grey, H-green, C-dark blue, Na-yellow.

As evidenced in the DRIFTS spectra, the Na sorbent sites in these DFMs dominate the interactions with CO₂. To understand differences between surface species and reactive intermediates with and without Na sorbent sites, RCC on a Na-free Ru/Al₂O₃ catalyst was probed with DRIFTS (Fig. 5). After CO₂ exposure (green spectrum, Fig. 5A), the DRIFTS spectrum exhibited a sharp peak at 2042 cm⁻¹, which is assigned to linear atop CO* species on Ru.⁵⁸ Notably, the peak associated with Na-stabilized Ru carbonyl species (1859 cm⁻¹, Fig. 2B) was not observed, further supporting our initial assignment of this peak to the presence of Na species. Further, the complex envelope between 1650 and 1200 cm⁻¹ was distinct in shape compared to that on the Ru+Na DFM, and indicated various Al₂O₃-carbonate structures: (i) a bicarbonate species (HCO₃*) designated by peaks at 1648, 1439, and 1228 cm⁻¹, (ii) a monodentate species indicated by a peak at 1520 cm⁻¹, and (iii) a formate species, likely formed by interaction between bicarbonate species and residual H* on the catalyst surface persisting after pre-reduction, designated by peaks at 1590, 1392, and 1375 cm⁻¹, all in accordance with reported peak positions for these surface species. After a He purge (dark gray spectrum, Fig. 5A), the intensity of all peaks decreased; in particular, the peaks associated with the bicarbonate species were no longer detectable, indicating complete desorption of this surface species that is critical to RCC performance. The remaining peaks correspond to formate and monodentate species.

Fig. 5 DRIFTS spectra of Ru/Al₂O₃ (A) after 30 min of CO₂ exposure followed by a 1 h purge with He and (B) reactive desorption in H₂ at 300 °C.

Upon exposure to H₂ (Fig. 5B), the carbonyl 2042 cm⁻¹ peak rapidly decreased, and in its place, peaks at 2040, 2006, and 1970 cm⁻¹ appeared. While there was a significant decrease in the intensities of these new peaks, 2040 and 1970 cm⁻¹ features persisted even after 1 h of H₂ exposure. According to a previous report by Wang et al., these peaks can be associated with geminal di-carbonyl species on low coordination Ru sites and are designated as a spectator species during CO₂ methanation.^15,59 Thus, the decrease in the peak intensities detected here can be described as a simple desorption of the carbonyl species, rather than consumption for conversion to CH₄. The remaining features, associated with formate and monodentate carbonate species, were consumed almost to completion during the H₂ exposure.

The geometries and activity of adsorbed CO_x species are considerably altered by the presence of Na sites. On the Ru+Na DFM, Na dominates as the primary adsorption site, offering strongly bound (but chemically reactive) bidentate and polydentate carbonate structures, compared to weakly bound bicarbonate and monodentate geometries on Al₂O₃ that desorb before reactive desorption begins. The presence of Na also impacts the distribution of carbonyl species, largely evidenced by a peak at ca. 1850 cm⁻¹ that only appears on the Ru+Na DFM; notably, all carbonyls on the Ru+Na DFM are reactive and produce CH₄, compared to the persistence of an unreactive geminal di-carbonyl on the Na-free Ru/Al₂O₃. Lastly, acetate emerges as a key intermediate in the presence of Na, compared to formates on a Na-free Ru/Al₂O₃, likely as a result of the different coverage and distribution of carbonyls on the two surfaces.

Ultimately, we are able to provide insight on the mobility of surface species in RCC. It is debated whether RCC mechanisms are driven by CO₂ desorption/re-adsorption, CO₃* migration, or H* migration. We hypothesize that migration of a variety of surface species may occur under these conditions. We observed that the formation of reactive hydrogenated intermediates (e.g., acetate, formate) is rapid and associated with the Na sorption site, likely driven by the migration of H* from Ru active sites to surface carbonates on the Na site. As proposed in co-fed CO₂ hydrogenation, formates likely undergo hydrogenation to CO*, which migrate to Ru sites for further hydrogenation to CH₄. Additionally, CO* and CH_x* form acetates that can occupy diverse Na and Al site types and form a relatively more stable intermediate. These reported surface phenomena are coupled with no CO₂ slip (i.e., no detection of gas phase CO₂), further reducing the probability that CO₂ is desorbed into the gas phase to be readsorbed to the surface for reaction (i.e., CO₂ desorption/re-adsorption mechanism).

The migration of intermediates could also conceptually coincide with the migration of active phases (i.e., Ru and Na species). Based on the fact that water is formed during the CO₂ hydrogenation reaction (co-fed or RCC), one would presume that Na species hydrate and dehydrate through the course of the reaction. The Na–O–Na dimer (as depicted in Fig. S8) may disperse to two adjacent NaOH species when hydrated and migrate to similar alumina surface sites. This migration would theoretically depend on the binding energy of NaOH to alumina and proximity of other alumina sites and may be aided by surface hydration, where Na⁺ moves from one bound water site to another with protons moving in the other direction. As this concept pertains to acetate/formate surface migration, similarly, one must consider binding energy, proximity of Na binding sites for migration, and species moving in the opposite direction for charge balance. Mechanistically, Na migration and acetate/formate migration are completely independent of each other. Na sites may be strongly bound to the alumina surface or may move with acetate/formate, and one would presume that protons are the charge compensating species moving in the other direction. Yet, regardless of whether the intermediates traverse in the bound state or in the ionic state, the ultimate result is that they must traverse close to a Ru site to be further reacted.

Proposed mechanistic differences between reactive desorption and co-fed hydrogenation over Ru+Na/Al₂O₃

To investigate potential differences in the reaction mechanisms of CO₂ methanation during RCC cycles compared to co-fed hydrogenation, a mixed stream of CO₂ and H₂ was fed over Ru+Na/Al₂O₃ in the DRIFTS reactor cell. The DRIFTS spectra after 2 h of reaction at 300 °C and a CO₂ : H₂ ratio of 1 : 4 are displayed with the black solid line in Fig. 6A, and the corresponding effluent gas analysis is presented in the blue region of Fig. 6B. Several differences in surface species populations were notable in the spectrum after 2 h of co-fed reaction compared to RCC reactive desorption. In the co-fed condition, the carbonate envelopes were centered at 1604 and 1340 cm⁻¹ (compared to 1650 and 1350 cm⁻¹), indicating a different distribution of carbonate species during co-fed hydrogenation. Also, a set of triplet peaks between 3000 and 2600 cm⁻¹ corresponded well with the C–H bond of a Na-formate species (NaOOCH).⁵⁴ These triplet peaks were not observed in the previous reactive desorption experiment (comparison in Fig. S14), and coupled with the shift in the carbonate envelopes, formates appear to be abundant. The carbonyl region was also distinct with higher intensity peaks in the lower wavenumber region (1850–1750 cm⁻¹), indicating a higher surface population of bridged carbonyls on Ru.

Fig. 6 (A) In situ DRIFTS spectra on Ru+Na/Al₂O₃ during co-fed hydrogenation (CO₂ : H₂ = 1 : 4, 10% CO₂, 40% H₂, balance He) at 300 °C and subsequent hold in H₂ without CO₂. (B) Gas-phase product detection during experiment.

A particularly clear difference was the absence of a peak at 1460 cm⁻¹, indicating that the corresponding acetate intermediate observed during reactive desorption was not formed/stabilized during co-fed hydrogenation. Importantly, the different surface speciation led to a different product distribution (Fig. 6B). During co-fed hydrogenation (blue region), a mix of CO and CH₄ was detected, in stark contrast to the high selectivity to CH₄ observed during RCC reactive desorption. After 2 h of co-fed CO₂ and H₂, the flow of CO₂ was stopped, and the DRIFTS spectra and product evolution were continuously monitored under flowing H₂ (Fig. 6A, gray spectra, Fig. 6B, yellow region). Upon the change in reactant gas flow, there was significant consumption of the surface species and a slight increase in intensity at 1460 cm⁻¹ (zoom on region in Fig. S15), accompanied by a rapid decrease in the signals for CO₂ and CO, and rapid, selective formation of CH₄, mirroring the evolution of products during RCC reactive desorption.

The evolution of surface species during co-fed hydrogenation and reactive desorption over a Ru+Na/Al₂O₃ is summarized in Fig. 7. In co-fed CO₂ methanation over Ru+Na/Al₂O₃, the Na component dominates interactions with CO₂ (i.e., not Al₂O₃ as observed on Na-free Ru/Al₂O₃), wherein the CO₂ was found to adsorb as surface carbonates (CO₃*). The formation of HCOO* was also observed during the co-fed reaction, presumably as CO₃* are hydrogenated by Ru-activated H*, following hypotheses that Al–HCO₃* species are hydrogenated to HCOO*.¹⁵ A high surface coverage of Ru–CO* groups was observed, mostly in the linear and bridged configurations, and there was a notable absence of unreactive gem-dicarbonyls over this DFM.

Fig. 7 Schematic of surface species evolution in co-fed hydrogenation and reactive carbon capture over a Ru+Na/Al₂O₃ DFM.

Under RCC conditions, CO₂ also preferentially adsorbed onto Na sites as CO₃* species. Additionally, it adsorbed as CO* on Ru sites, though in lower abundance compared to co-fed hydrogenation. Once H₂ was introduced, CO* and some CO₃* species were rapidly consumed. It is likely that the CO₃* species were hydrogenated via H* from Ru, much like in co-fed hydrogenation, to form a formate intermediate. However, a distinct peak at 1460 cm⁻¹ indicates the formation of an acetate intermediate, assigned through computational modeling and CO methanation experiments. This is likely formed between CH_x* and CO* species as reported for CO methanation and based on our reaction conditions.⁵⁵ The acetate species is consumed, albeit slowly, and forms methane with high selectivity. Falbo et al. reported that acetates on Ru/Al₂O₃ started decomposing at 250 °C, supplying CO* to the surface.⁵⁵ Furthermore, Na-acetate has been reported to decompose to form methane in biomass hydrotreating applications.⁶⁰ Thus, it is likely that the surface Na-acetate formed under RCC conditions also decomposes to form methane, given that the reactive desorption step is carried out at 300 °C. We propose that the formation of this acetate species contributes to the high methane selectivity in reactive desorption: in the co-fed environment, there is a constant population of CO* occupying the Ru sites which must also activate H₂. To activate H₂, some CO must desorb, resulting in the observed CO in the product stream, and thus, a lower CH₄ selectivity. However, in RCC reactive desorption, the CO* on Ru is quickly reacted or drawn away to form acetate species, liberating Ru sites for H₂ activation. The acetate species then serve as a temporary reservoir of CO* species for hydrogenation when more hydrogenating sites are available, accompanied by liberating an equivalent of CH₄ when they decompose. The net result is no desorption of CO and high methane selectivity.

Conclusions

This work provides a comprehensive understanding of the key interactions between CO₂ and DFM components – sorbent and catalytic metal – and how reactive desorption mechanisms proceed over a Ru+Na DFM. We first investigated the CO₂ adsorption characteristics on several Al₂O₃-supported group 1 and group 2 metal oxides, which showed that Na/Al₂O₃ maintains the highest gravimetric and molar adsorption capacity over a range of flue gas temperatures and adsorbed CO₂ in diverse geometries (bidentate, polydentate, bicarbonate). Catalytic metals (Ru, Pd) introduced to Na/Al₂O₃ generally increased adsorption capacity, with the greatest increase on Ru, and without major perturbation of adsorbed CO₂ binding geometries on Na sites. In subsequent reactive desorption on the DFMs, Ru+Na/Al₂O₃ selectively produced CH₄ while Pd+Na/Al₂O₃ selectivity produced CO, indicating tunability of product selectivity subject to catalytic metal. Through operando DRIFTS and computational modeling, formate and acetate species were identified as likely intermediates on both DFMs. Comparison of RCC over a Na-free Ru/Al₂O₃ showed that Na plays a key role in stabilizing carbonates and altering the distribution of carbonyls. Notably, (i) carbonates species bind stronger on Ru+Na/Al₂O₃, (ii) the acetate intermediate was only detected on Ru+Na/Al₂O₃, and (iii) unreactive geminal di-carbonyls were only formed on Ru/Al₂O₃. Finally, a bound acetate intermediate does not persist under co-fed conditions over Ru+Na/Al₂O₃, accompanied by lower selectivity to CH₄ when compared to RCC. We posit that the acetate species during RCC draw CO* species away from Ru sites during RCC, which are required for hydrogen activation, leading to selective CH₄ formation. In summary, we propose that the characteristics and relative populations of adsorbed CO_x species are affected by (1) the presence of strongly basic sites in DFMs when compared to sorbent-free catalysts, and (2) the step-wise introduction of CO₂ and H₂ compared to co-fed CO₂ and H₂. The unique synergy between the DFM materials properties and RCC process operation thus results in a reaction mechanism and catalytic performance that cannot be directly extrapolated from traditional co-fed hydrogenation literature.

Conflicts of interest

There are no conflicts to declare.

Data availability

The following data are included in the supplementary information (SI): metal content, CO₂ chemisorption data, and DRIFTS analysis of adsorbed CO₂ for sorbents; structural models of slabs and adsorbed CO₂ used in computational analyses; computed vibrational spectra for adsorbed CO₂ and related species; additional in situ DRIFTS data on DFMs with relevant time-dependent spectra; surface energies of various slab terminations. See DOI: https://doi.org/10.1039/d5ey00320b. All computational files have been made freely available at: C. Jeong-Potter, N. Mehra, C. A. Farberow and D. A. Ruddy, Identifying mechanistic differences between co-fed CO₂ hydrogenation and reactive CO₂ capture using Ru and Pd dual function materials, Materials Cloud Archive, 2026, 32, https://doi.org/10.24435/materialscloud:nr-g4.

Acknowledgements

This work was authored by the National Laboratory of the Rockies (NLR) for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. This work was supported by the Laboratory Directed Research and Development (LDRD) Program at NLR. A portion of this research was performed using computational resources sponsored by the U.S. Department of Energy's Office of Critical Minerals and Energy Innovation and located at NLR. The views expressed in the article do not necessarily represent the views of the DOE or the U.S. Government. The U.S. Government retains and the publisher, by accepting the article for publication, acknowledges that the U.S. Government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for U.S. Government purposes. The authors thank Dr W. Wilson McNeary (NLR) for his assistance with elemental analysis.

References

1. IPCC, in Special Report: Global Warming of 1.5 °C, 2018.
2. S. Stucki, A. Schuler and M. Constantinescu, Int. J. Hydrogen Energy, 1995, 20, 653–663.
3. C. Graves, S. D. Ebbesen, M. Mogensen and K. S. Lackner, Renewable Sustainable Energy Rev., 2011, 15, 1–23.
4. F. V. Vázquez, J. Koponen, V. Ruuskanen, C. Bajamundi, A. Kosonen, P. Simell, J. Ahola, C. Frilund, J. Elfving, M. Reinikainen, N. Heikkinen, J. Kauppinen and P. Piermartini, J. CO₂ Util., 2018, 28, 235–246.
5. M. Fasihi, O. Efimova and C. Breyer, J. Cleaner Prod., 2019, 224, 957–980.
6. M. S. Duyar, M. A. Arellano and R. J. Farrauto, Appl. Catal., B, 2015, 168, 370–376.
7. M. S. Duyar, S. Wang, M. A. Arellano-treviño and R. J. Farrauto, J. CO₂ Util., 2016, 15, 65–71.
8. I. S. Omodolor, H. O. Otor, J. A. Andonegui, B. J. Allen and A. C. Alba-Rubio, Ind. Eng. Chem. Res., 2020, 59, 17612–17631.
9. S. Wang, R. J. Farrauto, S. Karp, J. H. Jeon and E. T. Schrunk, J. CO₂ Util., 2018, 27, 390–397.
10. M. A. Arellano-treviño, N. Kanani, C. W. Jeong-potter and R. J. Farrauto, Chem. Eng. J., 2019, 375, 121953.
11. U.S. Energy Information Administration, Henry Hub Natural Gas Spot Price, https://www.eia.gov/dnav/ng/hist/rngwhhdm.htm, (accessed 16 July 2025).
12. Methanex, Historical Methanol Posted Price, https://www.methanex.com/our-products/about-methanol/pricing/, (accessed 16 July 2025).
13. P. Melo Bravo and D. P. Debecker, Waste Disposal Sustainable Energy, 2019, 1, 53–65.
14. S. B. Gharamaleki, T. R. Reina and M. S. Duyar, Prog. Energy, 2025, 7, 012001.
15. X. Wang, Y. Hong, H. Shi and J. Szanyi, J. Catal., 2016, 343, 185–195.
16. M. A. Nolen, S. A. Tacey, S. Kwon and C. A. Farberow, Appl. Surf. Sci., 2023, 637, 157873.
17. M. A. Abir, R. E. Phillips, J. Z. M. Harrah and M. R. Ball, Catal. Sci. Technol., 2024, 14, 4506–4521.
18. C. Zhang, Z. Xiao, X. Guo, X. Tan, J. Gu, J. Li, G. Li, J.-J. Zou and D. Wang, Chem. Eng. J., 2025, 519, 165414.
19. R. Ye, J. Ding, T. R. Reina, M. S. Duyar, H. Li, W. Luo, R. Zhang, M. Fan, G. Feng, J. Sun and J. Liu, Nat. Synth, 2025, 4, 288–302.
20. A. J. Al Abdulghani, S. Ganguly, R. H. Hagmann, Z. Sun, M. Alvear, L. O. Mark, E. Nikolla, Y. J. Pagán-Torres and I. Hermans, J. Am. Chem. Soc., 2025, 147, 27438–27448.
21. C. Jeong-Potter, M. A. Arellano-Treviño, W. W. McNeary, A. J. Hill, D. A. Ruddy and A. T. To, EES. Catal., 2024, 2, 253–261.
22. L. Moreno Bravo, F. C. Meunier and J. Kopyscinski, Appl. Catal., B, 2025, 361, 124591.
23. L. Proaño, E. Tello, M. A. Arellano-Trevino, S. Wang, R. J. Farrauto and M. Cobo, Appl. Surf. Sci., 2019, 479, 25–30.
24. L. Proaño, M. A. Arellano-Treviño, R. J. Farrauto, M. Figueredo, C. Jeong-Potter and M. Cobo, Appl. Surf. Sci., 2020, 533, 147469.
25. S. Sapru, K. D. Hart, C. Zhou, G. Liccardo, J. Oh, M. J. Hollobaugh, J. Osio-Norgaard, A. Majumdar, B. D. Chandler and M. Cargnello, ACS Nano, 2025, 19, 2484–2496.
26. M. A. Arellano-Treviño, Z. He, M. C. Libby and R. J. Farrauto, J. CO₂ Util., 2019, 31, 143–151.
27. C. Jeong-Potter, A. Porta, R. Matarrese, C. G. Visconti, L. Lietti and R. Farrauto, Appl. Catal., B, 2022, 310, 121294.
28. G. Kresse and J. Hafner, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 49, 14251–14269.
29. G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 11169–11186.
30. G. Kresse and J. Furthmüller, Comput. Mater. Sci., 1996, 6, 15–50.
31. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868.
32. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1997, 78, 1396.
33. P. E. Blöchl, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 50, 17953–17979.
34. K. Momma and F. Izumi, J. Appl. Crystallogr., 2008, 41, 653–658.
35. M. Gajdoš, K. Hummer, G. Kresse, J. Furthmüller and F. Bechstedt, Phys. Rev. B: Condens. Matter Mater. Phys., 2006, 73, 045112.
36. A. Hjorth Larsen, J. Jørgen Mortensen, J. Blomqvist, I. E. Castelli, R. Christensen, M. Dułak, J. Friis, M. N. Groves, B. Hammer, C. Hargus, E. D. Hermes, P. C. Jennings, P. Bjerre Jensen, J. Kermode, J. R. Kitchin, E. Leonhard Kolsbjerg, J. Kubal, K. Kaasbjerg, S. Lysgaard, J. Bergmann Maronsson, T. Maxson, T. Olsen, L. Pastewka, A. Peterson, C. Rostgaard, J. Schiøtz, O. Schütt, M. Strange, K. S. Thygesen, T. Vegge, L. Vilhelmsen, M. Walter, Z. Zeng and K. W. Jacobsen, J. Phys.: Condens. Matter, 2017, 29, 273002.
37. D. Karhánek, Self-Assembled Monolayers Studied by Density-Functional Theory, University of Vienna, 2010.
38. P. Gruene, A. G. Belova, T. M. Yegulalp, R. J. Farrauto and M. J. Castaldi, Ind. Eng. Chem. Res., 2011, 50, 4042–4049.
39. C. J. Keturakis, F. Ni, M. Spicer, M. G. Beaver, H. S. Caram and I. E. Wachs, ChemSusChem, 2014, 7, 3459–3466.
40. A. Porta, R. Matarrese, C. G. Visconti, L. Castoldi and L. Lietti, Ind. Eng. Chem. Res., 2021, 60, 6706–6718.
41. A. Bermejo-López, B. Pereda-Ayo, J. A. González-Marcos and J. R. González-Velasco, J. CO₂ Util., 2019, 34, 576–587.
42. L. Li, S. Miyazaki, S. Yasumura, K. W. Ting, T. Toyao, Z. Maeno and K. Shimizu, ACS Catal., 2022, 12, 2639–2650.
43. T. Sasayama, F. Kosaka, Y. Liu, T. Yamaguchi, S.-Y. Chen, T. Mochizuki, A. Urakawa and K. Kuramoto, J. CO₂ Util., 2022, 60, 102049.
44. L. C. Wirner, F. Kosaka, T. Sasayama, Y. Liu, A. Urakawa and K. Kuramoto, Chem. Eng. J., 2023, 470, 144227.
45. N. C. Nelson, M.-T. Nguyen, V.-A. Glezakou, R. Rousseau and J. Szanyi, Nat. Catal., 2019, 2, 916–924.
46. X. Jiang, X. Nie, X. Guo, C. Song and J. G. Chen, Chem. Rev., 2020, 120, 7984–8034.
47. X.-G. Wang, L.-Y. Qiao, L.-Y. Xu, Y.-Y. Zeng, Z.-S. Huang, S.-S. Zong, J.-X. Wang, Z.-F. Zhou and Y.-G. Yao, Chem. Eng. J., 2024, 500, 156937.
48. C. Jeong-Potter, M. Abdallah, S. Kota and R. Farrauto, Ind. Eng. Chem. Res., 2022, 61, 10474–10482.
49. F. Solymosi and A. Barko, J. Catal., 1986, 101, 458–472.
50. M. Bowker, P. Stone, R. Bennett and N. Perkins, Surf. Sci., 2002, 497, 155–165.
51. S. Y. Chin, C. T. Williams and M. D. Amiridis, J. Phys. Chem. B, 2006, 110, 871–882.
52. L.-Y. Gan and Y.-J. Zhao, J. Chem. Phys., 2010, 133, 094703.
53. X. Wang, H. Shi, J. H. Kwak and J. Szanyi, ACS Catal., 2015, 5, 6337–6349.
54. K. Ito and H. J. Bernstein, Can. J. Chem., 1956, 34, 170–178.
55. L. Falbo, C. G. Visconti, L. Lietti and J. Szanyi, Appl. Catal., B, 2019, 256, 117791.
56. A. de Klerk, Chapter 4: Fischer Tropsch Synthesis in Fischer–Tropsch Refining, 2011.
57. B. H. Davis, Catal. Today, 2009, 141, 25–33.
58. B. T. Loveless, C. Buda, M. Neurock and E. Iglesia, J. Am. Chem. Soc., 2013, 135, 6107–6121.
59. S. Xu, S. Chansai, S. Xu, C. E. Stere, Y. Jiao, S. Yang, C. Hardacre and X. Fan, ACS Catal., 2020, 10, 12828–12840.
60. J. A. Onwudili and P. T. Williams, Green Chem., 2010, 12, 2214.

---

Footnote

† Present address: Department of Biomedical and Chemical Engineering, Syracuse University, Syracuse, New York, 13244, USA. E-mail: cjeongpo@syr.edu

---