Introduction
To secure our future energy supply,
(1) it will be necessary to build up storage capacities for extremely large amounts of energy. Such storage buffers are required for temporal matching of demand and supply as the energy generated from renewable sources,
(2) e.g., solar and wind, inherently fluctuates. To limit land use, energy should be stored in materials with a high gravimetric storage density, i.e., chemical bonds.
(3) Hydrogen
(4) (H
2) seems attractive due to its high gravimetric storage density and the mature technology for its production through electrolysis. However, large-scale application of H
2 is problematic due to safety issues associated with large quantity storage, incompatibility with current gas distribution infrastructure, and high costs to construct or upgrade infrastructure. By transforming the H–H bond from green H
2 into C–H bonds of methane (CH
4),
(5) it is possible to produce a green energy carrier with high gravimetric storage density that is actually currently used in our infrastructure. Carbon dioxide (CO
2) serves as a carbon source for this methanation and, therefore, this process can help to reduce CO
2 emitted from point sources.
For this methanation to be sustainable as a whole, it is necessary that all energy required to drive the process directly originates from sunlight, i.e., no transformation of light into electricity (which inherently leads to energy losses) is needed. This process would require improvements over the current state-of-the-art Sabatier process,
(6) which requires high temperatures (300–500 °C), and the use of semiconductor photocatalysts,
(7) which can harvest only the short wavelength part of the solar spectrum and would thus require highly concentrated UV light (
Figure 1).
Plasmonic nanomaterials, e.g., metallic nanoparticles of Au, Ag, and Al, can transform light into heat and hot charge carriers,
(8) which opens up multiple pathways for promoting catalytic processes. Photoinduced increase of the temperature of plasmonic nanoparticle catalysts may provide heat to nearby reactants, resulting, e.g., in increased reaction rates. Alternatively, hot electrons may be generated and transferred to a reactant in close proximity to the particle surface, which can influence, e.g., rate or selectivity of catalytic processes. Third, the photon rate seen by adjacent reactants may be increased because of enhancement of the optical nearfield adjacent to the nanoparticle. A recent review by Baffou and Quidant provides an overview of the main mechanisms involved in plasmon catalysis.
(8) Depending on their type of metal, size, and shape, they can harvest a particular part of the sunlight spectrum.
(9) Combining plasmonic nanoparticles with several sizes and shapes within a single catalyst could be exploited to harvest the energy from the entire solar spectrum reaching the earth’s surface. Recently, Corma, García, and co-workers
(10) used a combination of a metallic and a semiconductor catalyst, i.e., Ni and NiO nanoparticles, for photomethanation with concentrated light at 150 °C. This study does not differentiate between the contribution of the metallic and semiconductor catalyst and is not carried out with terrestrial sunlight. Ye and co-workers
(11) screened a variety of group VIIIb nanoparticles, including Ru and Rh, for photomethanation using highly concentrated light to heat the catalyst to approximately 300 °C (bulk photothermal heating) for carrying out the state-of-art Sabatier process. Liu and co-workers proposed a technique to distinguish between thermal and nonthermal contributions in plasmon catalysis.
(12) For photomethanation of CO
2 using titania-supported Rh nanospheres as a catalyst, they determined an apparent quantum efficiency of 46% for high-intensity UV light (λ = 365 nm, photon flux Φ = 2.7 W cm
–2) and a catalyst temperature of 350 °C. To the best of our knowledge, this is the highest reported apparent quantum efficiency to date for plasmon-catalyzed photomethanation of CO
2. An extensive overview of the literature on photomethanation processes with correspondingly reported catalytic activities has been added to the Supporting Information (SI,
Table S1). Furthermore, Liu and co-workers demonstrated that illumination of Rh nanoparticle catalysts with UV light can be applied to tune the selectivity in the catalytic hydrogenation of CO
2, toward either CH
4 or CO.
(13)
Here, we report a rod-shaped Ru metal catalyst on Al2O3 that facilitates photomethanation. We managed to promote the methanation using sunlight as a sole energy source. Furthermore, we demonstrate a large nonthermal contribution of sunlight to the reaction, resulting in a high photon-to-methane (PTM) efficiency of up to 55%. We attribute the excellent sunlight-harvesting ability of our catalyst and the high photon-to-methane efficiency to its plasmon resonance in the UV–vis–NIR, which relates to the geometry of the catalyst and, thus, provides valuable guidelines for catalyst design. Our highly efficient conversion of H2 to methane is a promising technology to simultaneously accelerate the energy transition and reduce CO2 emissions.
Results and Discussion
In search of a catalyst that consists of plasmonic metallic particles with a variety of shapes and sizes so that it can harvest a large part of the solar energy, we prepared RuO
2 nanocrystals supported on γ-Al
2O
3 (6.0 wt % Ru) via a wet impregnation technique (
Experimental Section). This material was used as a precatalyst for the methanation at 1 sun light intensity (1 kW m
–2) and 150 °C reactor temperature (
Experimental Section), which leads only to formation of CH
4 and H
2O as products. After 2 h, 174 ± 9 mmol CH
4 g
Ru–1 was obtained under these conditions, whereas only a fraction of this amount (51 mmol CH
4 g
Ru–1) was produced without illumination at the same reactor temperature. Thus, the catalyst is substantially more active under illumination with terrestrial sunlight when compared to dark.
To investigate the stability of the catalyst under reaction conditions, we used the catalyst in seven sequential 30 min reactions at 1 sun and 150 °C (
Figure 2b). Reproducibly, the first two runs gave a lower yield of methane (first run 36 ± 2 mmol CH
4 g
Ru–1) than the runs 3–7 (57 ± 2 mmol CH
4 g
Ru–1), indicating that the RuO
2 precatalyst is reduced in situ to the metallic Ru catalyst during the first reaction runs. No catalyst deactivation was observed from the third to the seventh run. Preactivation of the RuO
2 precatalyst with H
2 for 2 h at 250 °C allowed us to obtain a more active Ru catalyst. Use of this catalyst for the methanation reaction at 150 °C and 1 sun illumination resulted in a 1.6-fold increase in yield after a 1 h reaction time (135 mmol CH
4 g
Ru–1,
Figure 2c) versus the use of the RuO
2 precatalyst (85 mmol CH
4 g
Ru–1,
Figure 2a). Again, only a fraction of methane was produced without illumination at the same reactor temperature (53 mmol CH
4 g
Ru–1 in 1 h,
Figure 2c).
To differentiate between thermal and nonthermal effects contributing to the sunlight-fueled methanation, we performed kinetic studies using the activated Ru catalyst and different combinations of reactor temperature (between 25 and 190 °C) and sunlight intensity (dark, 1, 5.7, 8.5 sun). In all experiments, the catalyst temperature was monitored by placing a thermocouple directly under the catalyst bed (SI,
Figure S3). In a comparative experiment using an infrared camera, we confirmed the correctness of the catalyst temperature determined with the thermocouple (SI,
Figure S6). As depicted in
Figure 3a, the catalyst temperature increases substantially upon illumination.
For the experiment performed at
Treactor = 150 °C and 1 sun solar intensity, the catalyst temperature
Tcat increased to 164 °C. Higher solar intensities resulted in a further increase in the catalyst temperature, with a maximum of 231 °C obtained for
Treactor = 150 °C and 8.5 sun. To differentiate between thermal and nonthermal contributions of sunlight illumination, we studied the reaction kinetics for different combinations of reactor temperature (
Treactor between 25 and 190 °C) and sunlight intensity (dark, 1, 5.7, 8.5 sun) and compared the reaction rates to those obtained in dark at the same catalyst temperature. The results of this study are depicted in the Arrhenius plot (
Figure 3b). The results clearly demonstrate that at the standard solar intensity of 1 sun and catalyst temperatures up to 189 °C the contribution of illumination with sunlight is largely thermal since the obtained reaction rates are nearly identical to those obtained in dark for the same catalyst temperature. At solar intensities of 5.7 and 8.5 sun, the contribution is largely thermal up to a catalyst temperature of approximately 200 °C. At higher catalyst temperatures of 212 °C (
Treactor = 150 °C + 5.7 sun), 221 °C (
Treactor = 125 °C + 8.5 sun), and 231 °C (
Treactor = 150 °C + 8.5 sun), respectively, the reaction rates are 1.93, 2.12, and 2.73 times higher than the calculated rate obtained in dark at the same catalyst temperature. From the kinetics study, we calculated a photon-to-methane (PTM) efficiency to quantify the nonthermal contribution of solar illumination as follows

with
kl = reaction rate upon illumination,
kd = reaction rate in dark at the same catalyst temperature as in the illumination reaction, Φ = photon flux, and
A = illuminated area. This PTM carries the same definition as the apparent quantum efficiency introduced by Liu and co-workers.
(12) Both terms represent the ratio of the nonthermal share of the reaction rate to the rate of incident photons, and reported values are therefore directly comparable. We decided to introduce the term “photon-to-methane efficiency” because it accurately represents the efficiency of solar methanation reactions as the ratio of CH
4 molecules produced to incident photons. We consider the term “quantum efficiency” confusing in this respect since it is defined as the ratio of photocurrent versus the rate of incident photons to characterize optoelectronic devices. Hence, it describes the share of energy from (sun)light that is converted to electrical energy. In the Sabatier process, solar energy is not converted to chemical energy since the reaction is exothermic.
The PTMs obtained for the reactions at
Treactor = 150 °C + 5.7 sun (
Tcat = 212 °C),
Treactor = 125 °C + 8.5 sun (
Tcat = 221 °C), and
Treactor = 150 °C + 8.5 sun (
Tcat = 231 °C), respectively, are 22.3, 25.0, and 54.8%. Compared to the best PTM reported for photomethanation of CO
2 by Liu and co-workers,
(12) we realized a substantially higher PTM (54.8 vs 46%) at lower temperature (231 vs 350 °C), lower light intensity (0.85 vs 2.7 W cm
–2), and for broad-spectrum sunlight rather than for monochromatic UV light of a wavelength of 365 nm.
In addition to the high PTM achieved in this process, we have demonstrated that conventional heating of the reactor is not required to promote the methanation (
Figure 4). At a solar intensity of 8.5 sun and
Treactor = 25 °C, the reaction proceeded with a rate of 15.5 mmol CH
4 g
Ru–1 h
–1 (
Figure 4a). At 25 °C and 10.1 sun, the photomethanation rate (52 mmol CH
4 g
Ru–1 h
–1) was approximately 18 times higher than the previously reported
(7h) 2.8 mmol CH
4 g
Ru/SiO2–1 h
–1 for
Treactor = 25 °C and 24.7 sun (
Figure 4b). Furthermore, for all sunlight intensities (5.7, 8.5, and 10.1 sun), the obtained reaction rates were higher than for the dark reaction (calculated values based on the Arrhenius plot in
Figure 3b), demonstrating a nonthermal contribution to the reaction rate. The nonthermal contribution to the reaction rate increases with increasing solar intensity.
To gain insight into this highly efficient photomethanation, we studied the RuO
2 precatalyst (SI,
S12 and S14) and the activated Ru catalyst with transmission electron microscopy (TEM,
Figure 5a–j), X-ray diffraction analysis (XRD,
Figure 5k), and X-ray photoelectron spectroscopy (SI,
S16). These studies confirm that the RuO
2 precatalyst, which consists of single crystalline nanorods and has the orthorhombic RuO
2 structure (SI,
Figure S14), is transformed into the activated Ru catalyst by solid-state recrystallization toward metallic Ru. This is performed either under typical reaction conditions or upon reduction with H
2 (preactivated catalyst). TEM confirms that the anisotropic shape of the RuO
2 nanoparticles is largely retained in the activated Ru catalyst. However, the recrystallization process is initiated by many separate nucleation events of the metallic phase, leading to polycrystalline Ru rods, as confirmed by the selected area electron diffraction (SAED) pattern in
Figure 5j.
Analysis of Ru on the Al
2O
3 catalyst (obtained through reduction under typical reaction conditions) by bright-field (BF)TEM and high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) (
Figure 5a,b, respectively) and energy-dispersive X-ray spectroscopy (EDX) mapping of Ru in the HAADF sample (
Figure 5c) shows that Ru nanostructures are present in a variety of shapes, among others nanospheres and (anisotropic) nanorods, with a variety of sizes. Studying several samples, we determined that the average diameter of the nanospheres was 11 ± 4 nm and that the average length and width of the nanorods were 153 ± 45 and 11 ± 10 nm, respectively (TEM analysis, SI,
S17). A representative, detailed TEM study of two Ru nanocrystals supported on a few Al
2O
3 crystals found in our catalyst is displayed in
Figure 5d–j. HAADF-STEM and EDX-mapping of the elements Ru, O, and Al (
Figure 5f–i) proved the metallic nature of the nanorods. BF-TEM (
Figure 5e) and HAADF-STEM (
Figure 5d) imaging allowed us to determine that the Ru nanorods are polycrystalline. This is further evidenced by the presence of rings in the SAED pattern in
Figure 5j. Finally, the crystallinity of the catalyst is confirmed by XRD analysis (
Figure 5k), giving reflections for the metallic Ru located at 2θ = 38.39 and 44.01°. From Scherrer analysis of the peak width in the XRD patterns, we obtained an average grain size of 11 nm. This is in agreement with the dimensions of the polycrystalline domains in the nanorods as well as with the average diameter of the nanospheres.
To gain insight into the sunlight absorption properties of the Ru catalyst, we determined the UV–vis–NIR diffuse reflectance of a powder sample. The (100 –
R) spectrum of the KBr-diluted powder, which is directly related to its absorption, is displayed in
Figure 6a. The spectrum clearly shows that the Ru catalyst features a broad and asymmetric absorption, which spans the UV–vis–NIR range with a maximum at ∼310 nm. Given the large size and shape polydispersity of the Ru nanoparticles present in our Ru catalyst sample, it is important to disentangle their relative contribution to the catalyst’s UV–vis–NIR light absorption properties. For this purpose, we performed optical simulations based on the boundary element method
(14) (BEM, SI,
S21 for details) to determine the extinction spectra of the two most representative nanoparticle geometries present in our catalyst sample: spherical Ru nanoparticles [diameter (ϕ) = 11 nm, SI,
S17 for details] and Ru nanorods (length × width = 153 nm × 14 nm, SI,
S17 for details). Based on the previous work performed on polycrystalline Au nanorods and nanoshells, we anticipate that the polycrystallinity of the Ru nanorods will not significantly influence its plasmonic properties.
(15) Figure 6b shows the calculated averaged cross sections (comparable to the extinction cross sections) in the 250–850 nm wavelength range (full range spectra in SI,
Figure S28) of those individual nanoparticles dispersed in a γ-Al
2O
3 medium [which is the actual catalyst support,
n(γ-Al
2O
3) ∼1.766 at 2.059 eV].
(16) The calculated spectra resemble closely the experimental spectral response of the Ru catalyst shown in
Figure 6a. The calculated spectrum of the spherical Ru nanoparticle is characterized by a steep extinction for λ < 400 nm, which coincides with the onset of its characteristic UV plasmon band
(17) (centered at ∼205 nm, SI,
Figure S28a), by a weak shoulder at ∼445 nm, and by an essentially featureless, and nonzero, extinction for λ > 600 nm. The Ru nanorod also features a steep extinction for λ < 340 nm (its characteristic UV plasmon band is centered at ∼184 nm, SI,
Figure S28b). However, differently from the spherical Ru nanoparticle, it features a significant extinction in the vis–NIR range resulting from a broad and symmetric band centered at ∼620 nm and from a (weaker) band peaking at ∼430 nm. Optical simulations performed at different angles of light incidence (SI,
Figure S29) indicate that the nanorod’s vis–NIR plasmon bands at ∼430 and ∼620 nm occur only when the angle of light incidence is ≠ 0° with respect to its longitudinal axis, i.e., when its longitudinal plasmon modes are being excited. On the other hand, the excitation of the nanorod’s UV plasmon band at ∼184 nm (∼2-fold more intense than the vis–NIR ones) occurs mainly upon illumination at 0° light incidence, i.e., upon transverse plasmon mode excitation. Given the large size of the Ru nanorod considered herein, it is likely that the nanorods may accommodate not only dipolar plasmon resonances but also higher order plasmon modes, as previously reported for other metal nanoparticles of comparable dimensions.
(18) An accurate plasmon mode assignment for both the UV and vis–NIR plasmon bands of the Ru nanorod would therefore require a thorough optical modeling investigation that goes beyond the scope of this work.
In any case, our optical modeling results indicate that nonspherical, anisotropic Ru nanoparticles, such as the large and high aspect ratio Ru nanorods present in our catalyst sample, feature not only a characteristic plasmon band in the UV but also plasmon bands that span the visible and NIR range. Given that light absorption governs the overall contribution to light extinction for both nanoparticle types (SI,
Figure S30), it can be concluded that the broad-band absorption of our γ-Al
2O
3-supported Ru catalyst (
Figure 6a) is the result of all plasmon band contributions from the randomly distributed spherical and nonspherical Ru nanoparticles of various sizes present in our catalyst sample and from the plasmon coupling effects that may arise as a result of their sometimes small interparticle separation distances.
(19) Importantly, our results demonstrate that a catalyst comprising solely Ru nanorods such as the ones presented in this study features a plasmonic absorption profile that matches that of the terrestrial AM 1.5 solar spectrum across the UV–vis–NIR (
Figure 6c), and it clearly surpasses by manifold that of a similar catalyst that comprises only small spherical Ru nanoparticles.
Because the aspect ratio of the Ru nanorods is essential to the position of the LSPR (similar to Au nanorods) and our Ru rods are highly polydispersed, length is 153 ± 45 nm and width is 11 ± 10 nm, the broad-band character of the extinction spectrum (
Figure 6a) is even more pronounced than the results obtained through optical simulations (
Figure 6b). This is beneficial for harvesting sunlight and should make the photomethanation largely wavelength independent. To experimentally validate the direct connection of the impact of the broad-band absorption of our Ru catalyst on its photomethanation activity, we performed experiments at 150 °C and 1 sun light intensity in the presence of a 395 nm and of a 495 nm longpass filter. Such filters block all sunlight with wavelengths below 395 and 495 nm (SI,
Figures S4 and S5), thus decreasing the overall light intensity reaching the plasmonic particles (
Figure 7). Comparing the formed methane per g
Ru in the photomethanation with and without the use of filters, the photomethanation with longpass filters gives a lower amount of methane g
Ru–1 (
Figure 7a, full spectrum: 100%; filter <395 nm: 92.3%; filter <495 nm: 82.9% all in 60 min). However, when normalizing the methane production results shown in
Figure 7a with respect to the solar intensity (
Figure 7b), essentially the same amount of methane per g
Ru per sun is formed in those three experiments (
Figure 7b, values for 60 min reaction time: full spectrum, 1 sun light intensity: 100%; filter <395 nm, 0.93 sun light intensity corrected to 1 sun light intensity: 99.3%; filter <495 nm, 0.81 sun light intensity corrected to 1 sun light intensity: 102.3%), indicating that the photomethanation is wavelength independent and thus that a broad-band sunlight absorption of the catalyst is responsible for the methanation.
Furthermore, we studied the photomethanation using Ru spheres instead of rods. The Ru spheres supported on γ-Al
2O
3 were prepared via wet immersion and direct calcination under a reducing atmosphere (
Experimental Section). HAADF-STEM analysis shows that the diameter of the randomly distributed spheres is 0.88 ± 0.13 nm (
Figure 8a), and through XRD analysis, we confirmed that the catalyst consists of γ-Al
2O
3 and metallic Ru (SI,
S15). Since individual Ru nanospheres do not display a plasmon resonance in the solar spectrum, which was demonstrated by the calculations performed in this study for 11 nm Ru spheres, and confirmed for a broad range of sizes by other research groups,
(17) the resulting catalyst powder is black and displays broad-band absorption properties between 300 and 800 nm (SI,
S20). Through kinetic studies, we investigated the reaction rate obtained with this catalyst in dark and upon illumination with sunlight (1, 5.7, and 8.5 sun). Based on these results, we were able to calculate the PTM′ using the equation reported above with
kd′ obtained at the same
Treactor instead of the same
Tcat. For comparison, we calculated the PTM′ obtained with the rods at the same reaction conditions and with
kd obtained at the same
Treactor. The rods show significantly higher values for PTM′, and the difference (Δ)PTM = PTM′
rods – PTM′
spheres is displayed in
Figure 8b.
In all cases, ΔPTM is larger than zero, which demonstrated that the PTM′ is higher for rods than for spheres. Furthermore, the difference in PTM′ increases with increasing light intensity from 3.3% at 1 sun to 29.3% at 5.7 sun with similar catalyst temperatures for both spheres and rods under these conditions. At 8.5 sun, ΔPTM further increases to 41.5% and the catalyst temperature of the spheres surpasses that of the rods by 15 °C. The fact that the Ru rods are far more efficient catalysts than the Ru spheres upon solar illumination supports the hypothesis that the plasmon resonance in the UV–vis–NIR plays a crucial role in this catalytic process. For Ru spheres, this is very weak and positioned in the UV; for Ru rods, the LSPR is stronger and covers the entire solar spectrum (vide supra). Consequently, the rods are much more efficient in harvesting sunlight and using the corresponding energy to drive the photomethanation, which is a plausible explanation for the distinct difference in activity. This difference in the LSPR of Ru spheres and rods is in line with previously reported plasmonic systems.
(20)