Mechanism of photocatalytic CO₂ methanation on ultrafine Rh nanoparticles

Abstract

Selective hydrogenation of CO₂ to yield CH₄ relies on the appropriate catalysts that can facilitate the cleavage of CO bonds and dissociative adsorption of H₂. Ultrafine Rh nanoparticles loaded on silica nanospheres were used as a class of photocatalysts to significantly improve the selectivity and reaction rate of producing CH₄ from the mixture of CO₂ and H₂ under the illumination of a broadband visible light source. The intense light scattering resonances in the silica nanospheres generate strong electric fields near the silica surface to enhance the light absorption power in the supported ultrafine Rh nanoparticles, promoting the efficiency of hot electron generation in the Rh nanoparticles. The interaction of the hot electrons with the adsorbate species on the Rh nanoparticle surface weakens the C–O bond to facilitate the deoxygenation of CO₂, favoring the production of CH₄ with a unity selectivity at a faster rate in the presence of surface adsorbed hydrogen (H*). The systematic studies on reaction kinetics and diffuse reflectance infrared Fourier transform (DRIFT) spectroscopy under different conditions, including various temperatures, illumination powers, and feeding gas compositions, reveal the reaction mechanism responsible for CO₂ methanation and the role of photoillumination.

---

New concepts

Light absorption in metal nanoparticle photocatalysts leads to the generation of heat (photothermal effect) and the generation of hot electrons (non-thermal effect), both of which can influence the catalytic reactions regarding kinetics and selectivity. The convolution of the photothermal and non-thermal effects makes it challenging to distinguish the contribution of hot-electron-driven chemistry (i.e., the non-thermal effect) and identify the corresponding mechanism. In this work, a composite catalyst of ultrafine Rh nanoparticles (with a diameter < 5 nm) dispersed on sub-micron-sized silica spheres (with a diameter of ∼400 nm) was used to catalyze CO₂ hydrogenation in the presence of H₂. The efficiency of generating hot electrons in metal nanoparticles under photoillumination usually increases as the nanoparticle size decreases and the strength of the excitation electric field increases. The silica spheres in the composite catalyst can exhibit intense light scattering resonances to generate enhanced electric fields near their surfaces to benefit energy absorption in the supported Rh nanoparticles, leading to a high efficiency of hot electron generation in the ultra-small Rh nanoparticles. The sub-micron size of the silica spheres allows packing a thin and loose catalyst bed on a platinum mesh to eliminate the temperature gradient and mass transport gradient in the catalyst, enabling the direct measurement of actual catalyst temperature. The ultrasmall size of Rh nanoparticles and continuous-wave xenon light source eliminates the possible local heating effect. Therefore, the non-thermal effect will make a dominant contribution to the photocatalytic performance. In addition, systematic studies on the reaction kinetics and steady-state DRIFT spectroscopy of adsorbate species allow for determination of the CO₂ methanation reaction mechanism on the catalyst and the role of photoillumination on the rate-determining steps. The results highlight the potential of photocatalyst design to achieve efficient performance in both reaction kinetics and product selectivity.

Introduction

Methanation of CO₂ (Sabatier reaction) is kinetically challenging because of the difficulty in dissociating the strong C–O bond with a bond energy of 8.33 eV and the existence of a competitive reverse water gas shift (rWGS) reaction that produces CO.¹ Supported Rh nanoparticles (NPs) have been widely used to catalyze the hydrogenation of CO₂ due to the approximate binding of carbon-containing intermediates and hydrogen on the nanoparticles’ surface.² Oxides represent a class of popular supports due to the low cost and ease of loading Rh NPs onto the supports using the impregnation method.³ The oxide supports are crucial to maintaining the high dispersity of Rh NPs, particularly at elevated temperatures, and promoting catalytic activity/selectivity relying on the synergistic interactions between the supports and Rh NPs. For instance, oxides of transition metals with variable oxidation numbers (e.g., semiconductor TiO₂, CeO₂, etc.) can be reduced by hydrogen to produce oxygen vacancies that migrate to the Rh/oxide interface to enable the dissociation of the C–O bond of intermediate species adsorbed on the Rh surface, promoting the production of CH₄ against CO.⁴ The high activation energy of oxygen vacancy migration (e.g., 231.6 kJ mol⁻¹ or 2.4 eV in reduced rutile TiO₂)⁵ usually slows the overall reaction kinetics of CO₂ hydrogenation.⁶ The surface Rh atoms interfaced with an oxide support represent only a small fraction of the overall surface Rh atoms, preventing the full capacity of the catalyst Rh NPs from achieving high reaction kinetics. The susceptibility of being reduced by hydrogen at elevated temperatures could sacrifice the stability of oxide supports and the corresponding Rh/oxide composite catalysts. On the other hand, it is promising yet challenging to enable rapid and selective production of CH₄ on the Rh NPs supported on dielectric oxides (i.e., SiO₂ and Al₂O₃) that do not react with hydrogen. However, slow CO₂ hydrogenation catalyzed by the dielectric–oxide-supported Rh NPs favors the production of CO. Increasing temperature represents one effective strategy to accelerate CO₂ hydrogenation, but the highly dispersed Rh NPs will suffer aggregation and fusion into larger NPs at high temperatures, lowering the availability of active surface Rh atoms. Although the selectivity of forming CH₄ can be somehow improved at higher reaction temperatures, robust full methanation is still far from being achieved through thermal catalysis on dielectric-oxide-supported Rh NPs.

Photocatalysis of CO₂ hydrogenation provides an alternative way to lower the reaction temperature while achieving desirable reaction activity and selectivity and maintaining the stability of the catalysts. Recent examples include the use of Z-scheme semiconductor heterostructures^7–9 and metal/support composites involving the photothermal effect,^10–16 non-thermal effect,^17–19 and metal–support interfacial interactions.²⁰ Photocatalysis of CO₂ hydrogenation on supported Rh NPs has also been recently investigated to improve the reaction kinetics and selectivity of CH₄.^21–24 For example, Liu and co-workers observed that both the photothermal effect and non-thermal effect were capable of accelerating full CO₂ methanation on TiO₂-supported Rh NPs under photoillumination.²² The origin of the non-thermal effect is unclear because the electronic coupling between the metallic Rh NPs and the semiconductor TiO₂ support changes the optical responses of individual components (i.e., surface plasmon resonance absorption in Rh NPs and semiconductor interband absorption in the TiO₂ support). Photoexcitation of the surface plasmon resonances in Al₂O₃-supported Rh nanocubes (37 nm in edge length) could simultaneously accelerate CO₂ hydrogenation and increase CH₄ selectivity from 50–60% to almost 100% when light intensity is high enough.²¹ The increased methanation selectivity is ascribed to the preferred injection of hot electrons into the antibonding orbitals of the C–O bond in the adsorbed HCO intermediates, facilitating cleavage of the C–O bond to favor the formation of CH₄ with a lowered reaction activation energy under light. The interpretation is in contrast to the experimental results showing an independence of activation energy on light intensity. Kim and co-workers observed slight acceleration of the methanation reaction of low pressure CO₂ and H₂ (<0.04 atm) on SiO₂-supported Rh NPs (<5 nm in size) under photoillumination, which was attributed to the direct photoexcitation of adsorbed CO₂ molecules exhibiting a narrowed LUMO–HOMO gap.²⁴ This explanation implies that light absorption in the Rh NPs does not influence the methanation reaction and is in apparent conflict with the superlinear dependence of reaction rate on light intensity observed under strong illumination. The reported controversial conclusions indicate the inaccuracy of the current understanding of the photocatalytic mechanism of CO₂ methanation.²⁵ For instance, cleavage of the C–O bond in intermediates may not represent the only rate-determining step (RDS) that is widely adopted in the electrochemical CO₂ reduction. The inaccurate understanding of the catalytic mechanism limits the potential to maximize the activity and selectivity of photocatalytic CO₂ methanation on highly dispersed Rh NPs, which exhibit a high surface area but weak optical absorption power.

Herein, we report the use of silica nanospheres (SiO_x NSs) as both dielectric light antenna and support for highly dispersed Rh NPs with sizes smaller than 5 nm, in which range metal NPs usually exhibit a high efficiency of generating high-energy hot electrons for photocatalytic reactions upon photoillumination due to the size-dependent quantum behaviors.^26–28 The intense light scattering resonances enhance electric fields locally near the surface of SiO_x NSs to significantly boost absorption of visible light in the small Rh NPs attached to the SiO_x NSs, further favoring hot-electron-driven photocatalysis.²⁹ The synergy of optical responses in SiO_x NSs and small Rh NPs makes the Rh-NP/SiO_x-NS composite particles a unique class of system to maximize the contribution of hot electrons to CO₂ hydrogenation even using the CW visible light source with a power density as low as 1.5 W cm⁻². This merit enables us to decipher the role of photoexcited hot carriers generated in the small Rh NPs in determining the selectivity and kinetics of photocatalytic CO₂ hydrogenation.

Results and discussion

Typical Rh-NP/SiO_x-NS samples with Rh NPs (loading content of 2.14 wt%) of different sizes were synthesized using an in situ reduction method (ESI,† Part I including Fig. S1). X-ray photoelectron spectroscopy of the as-synthesized composite particles exhibits strong signals of Rh, revealing the high-density loading of Rh NPs on the surface of SiO_x NSs (Fig. S2, ESI†). The diffuse reflectance spectra (DRS) of the composite particles in the ultraviolet-visible region exhibit distinct absorption peaks at ∼450 nm and ∼700 nm regardless of the size of Rh NPs, while the bare SiO_x NSs exhibit zero absorption in the visible spectral region (Fig. S3c, ESI†), indicating that the Rh NPs are responsible for the optical absorption. On the other hand, an aqueous dispersion of Rh NPs with similar sizes shows a peak-less feature in the absorption spectrum. The spectral difference between the supported and freestanding Rh NPs highlights that the SiO_x NSs enhance optical absorption in Rh NPs due to the strong light scattering resonances in the SiO_x NSs.^30,31 In contrast, the Rh NPs supported on alumina (Al₂O₃) NPs with irregular shapes (Fig. S3a, ESI†) do not show absorption peaks because of the lack of light scattering resonances in the Al₂O₃ NPs. The DRS spectrum of the Rh NPs loaded on mesoporous silica nanoparticles with a size of <50 nm (Fig. S3b, ESI†) also exhibits less pronounced peaks but overall higher intensity than the Rh NPs on SiO_x NSs, indicating the lack of light scattering resonances in the small mesoporous silica nanoparticles and deeper penetration of light in the mesoporous nanoparticle powder. The diffuse light penetrating the powder sample exhibits random propagation directions and reduced intensity to favor the photothermal process upon absorption in the buried nanoparticles. Therefore, using SiO_x NSs as a support for Rh NPs minimizes the penetration of diffuse light to weaken the photothermal process, while the strong light scattering resonance on the surface of the SiO_x NSs enhances light absorption in small Rh NPs to benefit the non-photothermal process for photocatalytic CO₂ hydrogenation. The catalytic performance was evaluated in a fix-bed reactor using a thin layer of Rh-NP/SiO_x-NS composite particles (∼0.7 mm in thickness, ∼7 mg in mass) on a platinum mesh, which can minimize the photothermal effect.³² The loose packing of the particles favored the isotropic and uniform mass transport of gas flow and heat transfer, ensuring fast dissipation of heat generated from light absorption (i.e., photothermal effect) to eliminate significant temperature deviation from the set temperature and temperature gradient in the thin catalyst layer (ESI,† Part II and Table S1). With respect to a set temperature of the reactor, the photothermally induced temperature change in the catalyst bed increases with the light intensity but decreases with the set temperature (Fig. S4, ESI†). The temperature changes of the catalyst bed are not significant (<15 °C) using the CW visible light with a power intensity up to 1.5 W cm⁻² in this work. These minor changes in temperature originated from the photothermal effect and induce only small perturbations of reaction kinetics, making it feasible to determine the presence of a non-thermal effect in the photocatalytic reactions with significantly accelerated kinetics. For example, catalytic CO₂ hydrogenation on the 2.3-nm Rh-NP/SiO_x-NS catalyst at 330 °C in a H₂-rich atmosphere (i.e., 440 mL h⁻¹ H₂, 35 mL h⁻¹ CO₂, 85 mL h⁻¹ Ar at 1 atm) produces CO and CH₄ with a CH₄ selectivity of only 71% in the dark. Illuminating the catalyst with visible light increases the steady-state production rate of both CO and CH₄ with a preference for CH₄, improving the CH₄ selectivity (Fig. 1a). The production rate of CH₄ increases by 60 times under illumination of 1.5 W cm⁻² power density while the increase of CO production rate is less than one time, corresponding to a CH₄ selectivity of 99%. The production rate of CH₄ monotonically increases with light power density and reaches a plateau under light stronger than 1.4 W cm⁻² due to the limitation of mass diffusion. In contrast, the production rate of CO increases with light power density and then decreases when the light is stronger than 1.1 W cm⁻² (Fig. 1b). The volcano-type dependence of the CO production rate indicates that surface-adsorbed CO (CO*) represents the intermediate to be consumed for forming CH₄. The increase in production rate under photoillumination corresponds to a decrease in the apparent activation energy of the methanation reaction (Fig. S5a, ESI†), indicating that both thermal energy and optical energy (no-thermal effect) can simultaneously determine the methanation kinetics on the Rh-NP/SiO_x-NS composite catalysts. Fig. 1c and d present the production rate and selectivity of CH₄ under photoillumination of different power densities and at different temperatures. Increasing both temperature and light power density favors the methanation kinetics and selectivity. The reaction kinetics is more sensitive to photoillumination at a lower reaction temperature. For example, the production rate of CH₄ increases by 350 times at 290 °C under the light of 1.5 W cm⁻², and the corresponding CH₄ selectivity increases to 97% from 35% (i.e., the value in the dark). The temperature-dependent photosensitivity can be described by the change of reaction activation energy as a function of light power density. The activation energy of both the CO₂ consumption reaction and CH₄ production reaction monotonically decreases with the light power density, i.e., from 82.1 kJ mol⁻¹ (0.85 eV) in the dark to 35.5 kJ mol⁻¹ (0.368 eV) under light of 1.5 W cm⁻² and from 115.8 kJ mol⁻¹ (1.20 eV) in the dark to 38.6 kJ mol⁻¹ (0.400 eV) under light, respectively (Fig. S5b, ESI†). The leveling of activation energy at a lower value leads to a larger and equivalent reaction rate for both consuming CO₂ and producing CH₄ under light with enough power density, resulting in a CH₄ selectivity close to unity (Fig. 1d). The results in Fig. 1c and 1d highlight the potential of the non-thermal effect in the photoexcited Rh NPs to promote both the reaction kinetics and selectivity of CO₂ methanation simultaneously.

Fig. 1 Dependence of CO₂ hydrogenation reaction rate and selectivity on the light power intensity. (a) Gas chromatography (GC) graphs of the products of CO₂ hydrogenation reactions at 330 °C, showing the variation of signals of CH₄ and CO under different photoillumination. (b) Production rates of CH₄ and CO and the selectivity for CH₄ calculated from (a). (c) Production rate of CH₄ and (d) selectivity for CH₄ as a function of reactor temperature and light power intensity. The light power intensity refers to the corrected light absorbed in the Rh NPs with a size of 2.26 nm. The gas flow into the reactor contained 35 mL h⁻¹ CO₂, 440 mL h⁻¹ H₂, and 85 mL h⁻¹ Ar.

The photocatalytic methanation reaction rate (ν_CH4) exhibits the power law dependence on the power density (I) of incident light: ν_CH4 ∝ Iⁿ, where n is the power law exponent (Fig. 2a). When the temperature is higher than 310 °C, the methanation reaction rate is linearly dependent on the power density of weak light (I < 0.4 W cm⁻²) and the value of n becomes larger than one under intense light with I > 0.4 W cm⁻². The transition of power law from linear dependence to superlinear dependence, i.e., from n = 1 to n > 1, is consistent with the mechanism of photocatalytic hot-carrier-driven reactions that are ascribed to the non-thermal effect.³³ This transition of power law corresponds to the change from single-electron to multiple-electron transitions that are required to excite adsorbates to accumulate sufficient vibrational energy higher than the activation barrier of the methanation reaction.^34,35 As the temperature lowers, the value of n in the superlinear region increases, and the linear region even disappears at the temperature below 310 °C. The n values exhibit a strong positive correlation with the reciprocal of thermodynamic temperature (T) (Fig. 2b) regardless of the size of Rh NPs, indicating that the methanation kinetics is less sensitive to light power density at a higher temperature. The linear dependence of n on 1/T agrees with the theoretical model proposed for desorption induced by multiple electronic transitions (DIMET), in which with ħw and ΔE_R representing the energy quantum of vibrational states and the reorganization energy of adsorbates, respectively.³⁴ For a given adsorbate species, the energy quantum of vibrational states is defined, but the reorganization energy is smaller at a higher temperature, resulting in a smaller n for the power law.^36–39 The results shown in Fig. 1c can be fitted with ν_CH4 = a × ν_CH4,dark²(T) × I^n(T) with n(T) = 5253/T − 5.77 (Fig. S6, ESI†), confirming the synergy of thermal energy and optical energy in promoting the methanation reaction kinetics. The minor contribution of the photothermal effect to methanation kinetics can be calculated from the Arrhenius equation with the activation energy of the dark reaction and the photothermally induced temperature increase as shown in Fig. S4b (ESI†). Subtracting the photothermal contribution from the difference in methanation reaction rates between dark and light conditions gives the quantitative contribution of the non-photothermal effect. The apparent quantum efficiency (AQE) of the non-photothermal effect is calculated by comparing the number of additional CH₄ molecules formed from the non-thermal effect and the number of photons absorbed by the catalysts (ESI,† Part III and Fig. S7). The positive values of AQE confirm the existence and contribution of the non-photothermal effect to promote the methanation kinetics. The value of AQE increases with both the temperature of the catalyst bed and the power density of incident light (Fig. 2c and Fig. S8, ESI†). The AQE can reach 26% at 360 °C under light with a power density above 1 W cm⁻², highlighting the high efficiency of the non-photothermal effect in driving the methanation reaction on ultrafine Rh NPs supported on SiO_x NSs.

Fig. 2 (a) Photocatalytic CH₄ production rate (i.e., the difference of the rate measured under photoillumination and in the dark) as a function of light power intensity. The log–log plots allow easy determination of the exponent n in the power law through linear fitting the data points. “n = 1” means a linear dependence of CH₄ production rate on light power intensity, and “n > 1” means a superlinear dependence of CH₄ production rate on light power intensity. The change from linear to superlinear dependence with an increase in the light power intensity indicates the switch of single to multiple electronic transitions in surface adsorbate induced by hot electron injection. The catalyst contained Rh NPs with a size of 2.26 nm. (b) The power n in the superlinear region as a function of the reciprocal of the reactor temperature when using catalyst Rh NPs with different sizes (i.e., 1.71 nm, 2.26 nm, and 4.20 nm). (c) Apparent quantum efficiencies (AQE) as a function of reactor temperature and light power intensity. The gas flow into the reactor contained 35 mL h⁻¹ CO₂, 440 mL h⁻¹ H₂, and 85 mL h⁻¹ Ar.

Under an atmosphere with high H₂ partial pressure (p_H2 = 0.786 atm), the reaction order of consuming CO₂ with respect to the partial pressure of CO₂ (p_CO2) is 0.42 in the dark at 330 °C while the reaction order of producing CH₄ is close to zero (i.e., 0.006) (Fig. 3aversusFig. 3b, black diamond symbols). The independence of CH₄ production rate on p_CO2 indicates that the higher chemical potential associated with a higher p_CO2 is mainly responsible for producing CO. The coverage of CO* intermediate on Rh NPs remains constant to react with hydrogen of a given pressure to produce CH₄. Photoillumination significantly increases the reaction rate of both CO₂ consumption and CH₄ production, which exhibit less difference than the dark reaction. The reaction order of producing CH₄ remains close to zero, but the reaction order of consuming CO₂ drops to 0.25 under visible light of 1.5 W cm⁻². The comparisons indicate that photoillumination accelerates both the dissociation of CO₂ and the methanation of intermediate CO*. Under an atmosphere with a low H₂ partial pressure (p_H2 = 0.41 atm), the reaction order of consuming CO₂ with respect to p_CO2 is always positive while the reaction order of producing CH₄ is negative under both dark and light conditions (Fig. 3aversusFig. 3b, square symbols). The contrast value sign of the reaction orders indicates that the reactivity of hydrogen plays the most crucial role in the methanation of CO* and the dissociation of CO₂ to CO* is less sensitive to hydrogen. The dependence of reaction rate on p_H2 shows the corresponding orders of consuming CO₂ and producing CH₄ are 0.53 and 0.73, respectively, at p_CO2 = 0.0625 atm and in the dark (Fig. 3c and d, black squares). The difference in reaction orders is significantly amplified under visible light of 1.5 W cm⁻²; for example, that of consuming CO₂ remains essentially unchanged but that of producing CH₄ increases to 2.55. The distinct change in reaction orders indicates that photoillumination significantly improves the reactivity of hydrogen and hydrogen-assisted deoxygenation of CO* most likely due to the increased coverage of adsorbed hydrogen atoms (H*). The higher reactivity of H* than H₂ favors the surface hydrogenation reaction, and the faster consumption of H* leads to a stronger dependence on p_H2 with a higher reaction order. Moreover, the consumption rate of CO₂ is much higher than the production rate of CH₄ at low p_H2 = p_CO2 = 0.0625 atm under light, i.e., 600 μmol g⁻¹ s⁻¹versus 11 μmol g⁻¹ s⁻¹, indicating the capability of photoillumination in promoting the cleavage of the C–O bond of CO₂* (and possible derivatives) to form CO* on the surface of Rh NPs.

Fig. 3 Dependence of CO₂ methanation kinetics on the partial pressure of CO₂ (p_CO2) and H₂ (p_H2). (a) and (b) Influence of p_CO2 on (a) consumption rate of CO₂ and (b) production rate of CH₄ in the dark or under photoillumination of a light power intensity of 1.5 W cm⁻². p_H2 was constant at either 0.410 atm or 0786 atm. (c) and (d) Influence of p_H2 on (c) consumption rate of CO₂ and (d) production rate of CH₄ in the dark or under photoillumination. (e) Dependence of the production rate of CH₄ on p_H2 under photoillumination of different light power intensities. p_CO2 was constant at 0.0625 atm. The total gas flow rate was 560 mL h⁻¹ and Ar was used as the balance gas. The catalyst Rh NPs had a size of 2.26 nm. (f) Fitted plots of the data shown in (d) and (e) using the kinetic equation of .

The results in Fig. 3a–d indicate that the non-photothermal effect (representing the major contribution of photoillumination in the current work) in Rh-NP/SiO_x-NS composite catalysts can increase the adsorption coverage of H* and facilitate C–O bond cleavage to favor CO₂ methanation kinetics. Light absorption in Rh NPs excites plasmons that decay to generate high-energy hot electrons with the population and energy depending on the intensity of incident light and the size of NPs.⁴⁰ The non-photothermal effect originates from the photoexcited electronic structures in the Rh NPs, influencing the interactions with the surface adsorbates in CO₂ hydrogenation. At a low p_H2 of 0.0625 atm, the production rate of CH₄ increases with the light power density up to 0.7 W cm⁻² and then decreases with even stronger light. Only intense light can significantly facilitate the cleavage of the first C–O bond of CO₂ to dissociate CO₂ into CO, which consumes a large portion of H* atoms to lower the availability of H* for CO* methanation. When p_H2 is high enough to exclude the limitation of H₂ diffusion, the production rate of CH₄ increases with the power density of incident light, indicating light absorption in Rh NPs favors activation of H₂ and C–O cleavage in the CO-derived intermediates. Regardless of the light intensity, the production rate of CH₄ increases with p_H2 to reach a maximum followed by a decrease (Fig. 3e). In the bell-like dependence of CH₄ production on light power density the negative correlation is ascribed to the fact that the cleavage of C–O bond in CO* methanation is not fast enough to consume H*. Shifting the maximum-point p_H2 from ∼0.62 atm in the dark (black squares, Fig. 3d) to ∼0.22 atm under light of 0.5 W cm⁻² (brown cycles, Fig. 3e) indicates that weak light can favor H₂ activation but not for the C–O cleavage at the same level. The maximum-point p_H2 is pushed back to high values under intense light (e.g., 0.42 and ∼0.7 atm for light of 1.0 and 1.5 W cm⁻², respectively), implying that intense light favors C–O cleavage to accelerate methanation with fast consumption of H*.

The analysis of results in Fig. 3 reveals that the non-thermal effect induced by light absorption in Rh NPs can promote H₂ activation (i.e., increase of H* coverage under a given p_H2) and C–O bond cleavage of intermediate adsorbates to accelerate CO₂ methanation kinetics. The C–O bond cleavage becomes predominant only under intense light although significant H₂ activation can always be observed under light. Moreover, surface H* atoms react with stable surface adsorbates (i.e., CO₂ and CO) to form appropriate hydrogenated intermediates to facilitate the C–O bond cleavage, promoting the formation of CH₄. In situ diffuse reflectance infrared Fourier transform (DRIFT) spectroscopy has been used to identify the most possible intermediates (peaks assignments summarized in Table S2, ESI†) and the influence of light on reaction kinetics. The adsorption of CO₂ on Rh NPs decreases with temperature, and the first C–O bond cleavage (aka direct dissociation of CO₂) occurs to produce CO* with low yield at a high enough temperature (>300 °C) (Fig. S9a, ESI†). In the presence of H₂ (p_H2 : p_CO2 = 1 : 1), the signal of CO₂* significantly reduces but the signal of CO* becomes dominant even at temperatures as low as 30 °C (Fig. S9b, ESI†), confirming the critical role of hydrogen in accelerating cleavage of the first C–O bond of CO₂*. The appearance of an infrared (IR) absorption peak around ∼2800 cm⁻¹ at low temperatures (e.g., 50 °C) is ascribed to the formation of bidentate formate (bi-HCOO*).⁴¹ This peak diminishes with an increase in temperature and the peak intensity of CO* increases, indicating that bi-HCOO* is most likely the hydrogenated intermediate species to facilitate cleavage of the first C–O bond of CO₂. When a gas mixture with p_H2 : p_CO2 = 1 : 1 flows through the catalyst bed at 50 °C, both CO* and bi-HCOO* accumulate on the surface of Rh NPs to reach plateaus of surface coverage after ∼30 min in the dark (Fig. 4a and c). The appearance of plateaus corresponds to the steady-state reaction conditions. It becomes much faster for CO* to reach a plateau of a higher coverage under light, and the accumulation time is only ∼12 min (Fig. 4b and c). The absence of IR absorption signal of bi-HCOO* under light indicates that the dissociation of bi-HCOO* is too fast to accumulate on the surface of Rh NPs. Therefore, photoillumination promotes cleavage of one C–O bond in bi-HCOO* to produce CO* at a higher rate, resulting in a higher coverage of CO* on the surface of Rh NPs (Fig. 4c, top panel). Similar results are observed in the H₂-rich atmosphere (e.g., p_H2 : p_CO2 = 10 : 1) except for the non-zero signal of bi-HCOO* under light (Fig. S10, ESI†). The accumulation of bi-HCOO* under light indicates that high H₂ partial pressure (corresponding to high coverage of H*) can accelerate the formation rate of bi-HCOO* intermediate.

Fig. 4 Time-dependent DRIFT spectra of adsorbate species on the Rh-NP/SiO_x-NS composite catalyst. (a) and (b) DRIFT spectra recorded from CO₂ hydrogenation at 50 °C (a) in the dark and (b) under photoillumination of a light power density of 1.5 W cm⁻². The gas flow contained 35 mL h⁻¹ H₂, 35 mL h⁻¹ CO₂, and 960 mL h⁻¹ Ar, providing a H₂-lean condition. The three sharp peaks between 1650 cm⁻¹ and 2100 cm⁻¹ originate from different adsorption states of CO*. The total area of these three peaks is proportional to the overall CO* on the catalyst surface. The peak around 2800 cm⁻¹ is ascribed to bi-HCOO* surface species. (c) Comparison of the integrated area of CO* peaks (top panel) and bi-HCOO* peak (bottom panel) in the dark and under photoillumination. The integration was derived from the spectra shown in (a) and (b). (d) and (e) DRIFT spectra recorded after CO₂ hydrogenation reaction at 330 °C while exposing the catalyst to a H₂-rich condition with a gas flow containing 350 mL h⁻¹ H₂ and 650 mL h⁻¹ Ar. The pre-hydrogenation reaction was performed using conditions the same as in Fig. 1, leaving the steady-state coverage of CO* to allow studying the deoxygenation of CO*. (f) Time-dependent area of CO* derived from the spectra shown in (d) and (e), showing the faster deoxygenation kinetics under photoillumination.

When temperature increases to 150 °C and above, the total signal of CO* under the steady-state condition is independent of gas composition, temperature, and photoillumination (Fig. S11 and S12, ESI†). The constant steady-state coverage of CO* on a given Rh NP catalyst suggests that the methanation of CO* has to be fast enough to match the rate of converting CO₂ to CO* to achieve high CH₄ selectivity. Otherwise, CO* detaches from the surface of Rh NPs to form gaseous CO. Methanation of CO* is limited by the cleavage of the C–O bond. Co-adsorption of H* redshifts the IR absorption peaks of CO* due to the slight weakening of the C–O bond. The presence of H* also favors the linear configuration of CO* neighboring with H* (i.e., rhodium carbonyl hydride) (Fig. S13, ESI†). The activation energy for directly cleaving the C–O bond of CO* on Rh catalysts decreases from 213 kJ mol⁻¹ (2.2 eV) to 143 kJ mol⁻¹ (1.48 eV) in the presence of one H* or 130 kJ mol⁻¹ (1.35 eV) in the presence of two H*.⁴² Although the pre-adsorbed H* on Rh NPs can help reduce CO₂ to CO*, a continuous supply of H₂ is necessary to drive the hydrogenation of CO* into CH₄ (Fig. S14, ESI†). The surface coverage of CO* is almost constant during the CO₂ hydrogenation reaction regardless of the light condition at an elevated temperature (Fig. S11 and S12, ESI†). We can use the residual CO* after stopping the CO₂ supply as the benchmark to evaluate the influence of light on the CO* methanation kinetics at 330 °C. A continuous flow of H₂ (p_H2 = 0.35 atm) can hydrogenate CO* to produce CH₄ at the measurable rate in the dark (Fig. 4d), decreasing the coverage of CO* on the surface of Rh NPs until the complete consumption of CO*. Photoillumination can accelerate the methanation of CO* with a higher production rate of CH₄ and fast consumption of CO* (Fig. 4e). For example, converting 80% CO* under light of 1.5 W cm⁻² takes 81 min while 124 min is required in the dark (Fig. 4f). A similar difference is also observed when a continuous flow of low-pressure H₂ (p_H2 = 0.035 atm) is applied (Fig. S15, ESI†). It is worth noting that no CH₄ formation can be detected with low-pressure H₂ in the dark (Fig. S15a, ESI†). The absence of the CH₄ signal might be ascribed to either that the formation rate of CH₄ is too low to be measured by the GC or that the low chemical potential of H₂ cannot overcome the energy barrier of cleaving the C–O bond in CO* but combine with CO* to form hydrogenated intermediate to desorb from the Rh surface. The most possible intermediate could be formaldehyde (H₂CO*), which can equilibrate with rhodium carbonyl hydride (L2-CO*) (Fig. S13 and S16, ESI†). Light illumination of Rh NPs promotes dissociation of H₂ to increase the adsorption coverage of H* to accelerate the formation of L2-CO* intermediate and facilitates cleavage of the C–O bond in CO* (Fig. S17, ESI†), resulting in a faster production of CH₄. When Rh NPs are replaced with Ag NPs that have very weak adsorption of hydrogen (H*),⁴³ the reduction of the CO₂ reaction primarily produces CO at much lower rates (Fig. S18, ESI†), confirming that the presence of sufficient H* on the metal surface is crucial for CO₂ methanation in the cleavage of both C–O bonds.

The involvement of adsorbed hydrogen in CO₂ methanation is consistent with the significant kinetic dependence on the partial pressure of H₂. The high reaction orders of CH₄ production with respect to p_H2 is not related to the mass transport limitations because of the thin catalyst bed and the stoichiometrically excess amount of H₂ (ESI,† Part IV including Fig. S19). Five possible reaction paths have been proposed for CO₂ methanation on catalyst surfaces based on the reported experiments and theoretical calculations: (i) CO₂* → CO* → C*, (ii) CO₂* → CO* → H_xCO* → CH_x*, (iii) CO₂* → HCOO* → CO* → C*, (iv) CO₂* → HCOO* → CO* → HCO* → C*/CH_x*, and (v) CO₂* →HCOO* → CO* → H₂CO* → CH_x*. By considering different rate-determining steps (RDS), the kinetic equations of the CH₄ production rate as a function of p_H2 can be determined (ESI,† Part V and Table S3). The results in Fig. 3d and e are then fitted with more than forty possible kinetic equations. The goodness of fitting corresponds to the possibility of the reaction mechanisms (Table S4, ESI†). The direct dissociation of CO₂ and CO in the reaction paths (i, ii, and iii) cannot fit the measured data well, resulting in a value of R² lower than 0.97 at 330 °C in the dark. The large fitting deviations could exclude the mechanisms involving the direct dissociation of CO₂ and CO. In contrast, the hydrogen-assisted dissociation paths (iv and v) offer an appropriate mechanism that is possible to have fitted R² larger than 0.99 (see the top row, Table S4, ESI†). The best-fitting equation is , which agrees with the experimental data regardless of the illumination condition (Fig. 3f). The corresponding methanation mechanism has the RDS of H₂CO* (or L2-CO*) + 2H* → CH₂* + H₂O* + *, highlighting the importance of H* in the full deoxygenation of CO₂ for forming CH₄. Photoillumination generates hot electrons in the Rh NPs to facilitate the deoxygenation of L2-CO* in the presence of high-density H*, increasing the CH₄ production rate.

Conclusion

According to the results of reaction kinetics and steady-state DRIFT spectroscopy, the CO₂ methanation mechanism catalyzed by the Rh NPs supported on SiO_x NSs is depicted in Fig. 5. The Rh surface has an appropriate H adsorption coverage that can bind with or interact with the neighboring CO₂* and CO* to facilitate the cleavage of C–O, enabling the production of CH₄ even though CO* can detach from the Rh surface as a side product. Fast reaction kinetics of the hydrogen-assisted deoxygenation of CO* is crucial to minimize the release of CO side product. Photoillumination of the catalyst using a visible light source excites surface plasmon resonances in the Rh NPs to generate hot electrons, which interact with molecular orbitals of CO* to weaken the C–O bond. The weakened C–O accelerates the deoxygenation of CO* to produce methane at a higher rate in the presence of high-density H*. The deoxygenation of the first O in CO₂* can also benefit from the hot-electron-induced weakening effect in the C–O bond under photoillumination to accelerate the formation rate of the intermediate CO*, which can provide enough CO* to support the accelerated CH₄ production. Since the deoxygenation of the two O atoms from CO₂ requires the presence of H*, the accelerated CH₄ production under photoillumination requires faster adsorption kinetics for the reaction H₂ → 2H*, which can generate H* at a higher rate. This correlation implies that photoillumination also favors the dissociative adsorption of H₂ on the Rh NPs, which is consistent with previous studies using Au nanoparticles with strong surface plasmon resonances.⁴⁴ The strategy combining the reaction kinetics and steady-state DRIFT spectroscopy allows for deconvolution of the CO₂ methanation reaction mechanism on catalysts. Using SiO_x NSs to support the ultrafine Rh NPs as catalysts minimizes the photothermally induced temperature change in the catalyst bed, making it feasible to evaluate the non-thermal photocatalysis. The results and conclusions shed light on photocatalyst design to achieve efficient performance in both reaction kinetics and product selectivity. For example, using Rh-NP/SiO_x-NS composite catalysts under visible light of 1.5 W cm⁻² at 330 °C achieved a near unity selectivity of CH₄ production from CO₂ hydrogenation and two orders of magnitude enhancement of production rate compared to the reaction in the dark.

Fig. 5 Proposed CO₂ deoxygenation and hydrogenation mechanism on the Rh-NP/SiO_x-NS catalyst. The primary role of the photo-generated hot electrons in the Rh NPs is to facilitate cleavage of the C–O bond.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the startup and OVPR seed grant from Temple University. Partial characterizations were performed with the use of TMI (Temple Materials Institute) facilities. We are thankful to Mr. sichuan Huang for the help with data fitting and Dr. Lauren Profitt for ICP characterizations.

Notes and references

1. S. Wu, S. Wu and Y. Sun, Sol. RRL, 2021, 5, 2000507.
2. A. Erdőhelyi, Catalysts, 2020, 10, 155.
3. N. M. Martin, P. Velin, M. Skoglundh, M. Bauer and P.-A. Carlsson, Catal. Sci. Technol., 2017, 7, 1086–1094.
4. É. Novák, K. Fodor, T. Szailer, A. Oszkó and A. Erdöhelyi, Top. Catal., 2002, 20, 107–117.
5. E. Iguchi and K. Yajima, J. Phys. Soc. Jpn., 1972, 32, 1415–1421.
6. Z. Wu, W.-J. Yin, B. Wen, D. Ma and L.-M. Liu, J. Phys. Chem. Lett., 2023, 14, 2208–2214.
7. J. Hua, S. Feng, C. Ma, H. Huang, K. Wei, X. Dai, K. Wu, H. Wang and Z. Bian, J. Environ. Chem. Eng., 2023, 11, 111290.
8. M. Ni, Y. Zhu, C. Guo, D.-L. Chen, J. Ning, Y. Zhong and Y. Hu, ACS Catal., 2023, 13, 2502–2512.
9. F. Raziq, M. Z. Rahman, S. Ali, R. Ali, S. Ali, A. Zada, X. Wu, J. Gascon, Q. Wang and L. Qiao, Chem. Eng. J., 2024, 479, 147712.
10. C. J. Bueno-Alejo, A. Arca-Ramos, J. L. Hueso and J. Santamaria, Catal. Today, 2020, 355, 678–684.
11. Z. Geng, Y. Yu and J. Liu, J. Phys. Chem. C, 2023, 127, 17723–17731.
12. Z. Li, R. Shi, J. Zhao and T. Zhang, Nano Res., 2021, 14, 4828–4832.
13. K. Lorber and P. Djinović, iScience, 2022, 25, 104107.
14. D. Mateo, P. Maity, G. Shterk, O. F. Mohammed and J. Gascon, ChemSusChem, 2021, 14, 5525–5533.
15. Y. Zhou, P. Zheng, F. Wang, F. Gu, W. Xu, Q. Lu, T. Zhu, Z. Zhong, G. Xu and F. Su, Nano Res., 2023 DOI:10.1007/s12274-023-6226-5.
16. R. Jiang, X. Yue, K. Wang, Z. Yang, W. Dai and X. Fu, Energy Fuels, 2022, 36, 11636–11646.
17. A. Anouar, R. García-Aboal, P. Atienzar, A. Franconetti, N. Katir, A. El Kadib, A. Primo and H. Garcia, Adv. Sustainable Syst., 2022, 6, 2100487.
18. Y. Chen, Y. Zhang, G. Fan, L. Song, G. Jia, H. Huang, S. Ouyang, J. Ye, Z. Li and Z. Zou, Joule, 2021, 5, 3235–3251.
19. X. Yue, K. Wang, Z. Yang, W. Ni, Z. Zhang, W. Dai and X. Fu, Chem. Eng. J., 2023, 458, 141491.
20. K. Wang, S. He, Y. Lin, X. Chen, W. Dai and X. Fu, Chin. J. Catal., 2022, 43, 391–402.
21. X. Zhang, X. Li, D. Zhang, N. Q. Su, W. Yang, H. O. Everitt and J. Liu, Nat. Commun., 2017, 8, 14542.
22. X. Zhang, X. Li, M. E. Reish, D. Zhang, N. Q. Su, Y. Gutiérrez, F. Moreno, W. Yang, H. O. Everitt and J. Liu, Nano Lett., 2018, 18, 1714–1723.
23. X. Li, H. O. Everitt and J. Liu, Nano Res., 2019, 12, 1906–1911.
24. C. Kim, S. Hyeon, J. Lee, W. D. Kim, D. C. Lee, J. Kim and H. Lee, Nat. Commun., 2018, 9, 3027.
25. U. Ulmer, T. Dingle, P. N. Duchesne, R. H. Morris, A. Tavasoli, T. Wood and G. A. Ozin, Nat. Commun., 2019, 10, 3169.
26. G. V. Hartland, L. V. Besteiro, P. Johns and A. O. Govorov, ACS Energy Lett., 2017, 2, 1641–1653.
27. K. D. Rasamani and Y. Sun, J. Chem. Phys., 2020, 152, 084706.
28. S. Stewart, Q. Wei and Y. Sun, Chem. Sci., 2021, 12, 1227–1239.
29. N. Zhang, C. Han, Y.-J. Xu, J. J. Foley Iv, D. Zhang, J. Codrington, S. K. Gray and Y. Sun, Nat. Photonics, 2016, 10, 473–482.
30. L. K. Ausman and G. C. Schatz, J. Chem. Phys., 2008, 129, 054704.
31. X. Dai, K. D. Rasamani, G. Hall, R. Makrypodi and Y. Sun, Front. Chem., 2018, 6, 494.
32. X. Bian, Y. Zhao, C. Zhou and T. Zhang, Angew. Chem., Int. Ed., 2023, 62, e202219340.
33. L. Zhou, M. Lou, J. L. Bao, C. Zhang, J. G. Liu, J. M. P. Martirez, S. Tian, L. Yuan, D. F. Swearer, H. Robatjazi, E. A. Carter, P. Nordlander and N. J. Halas, Angew. Chem., Int. Ed., 2021, 118, e2022109118.
34. T. Olsen and J. Schiøtz, Phys. Rev. Lett., 2009, 103, 238301.
35. C. Frischkorn, Surf. Sci., 2005, 593, 67–78.
36. P. Szymanski, A. L. Harris and N. Camillone, J. Phys. Chem. A, 2007, 111, 12524–12533.
37. H. Ueba and B. N. J. Persson, J. Phys.: Condens. Matter, 2008, 20, 224016.
38. J. Güdde and U. Höfer, J. Phys.: Condens. Matter, 2006, 18, S1409–S1424.
39. T. Avanesian and P. Christopher, J. Phys. Chem. C, 2014, 118, 28017–28031.
40. L. V. Besteiro and A. O. Govorov, J. Phys. Chem. C, 2016, 120, 19329–19339.
41. X. Wang, H. Shi and J. Szanyi, Nat. Commun., 2017, 8, 513.
42. R. Zhang, F. Liu, X. Zhao, B. Wang and L. Ling, J. Phys. Chem. C, 2015, 119, 10355–10364.
43. S. S. Kumar and H. Lim, Sustainable Energy Fuels, 2023, 7, 3560–3583.
44. S. Mukherjee, F. Libisch, N. Large, O. Neumann, L. V. Brown, J. Cheng, J. B. Lassiter, E. A. Carter, P. Nordlander and N. J. Halas, Nano Lett., 2013, 13, 240–247.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3nh00506b

---