Introduction

Efficient utilization of solar energy1 to produce highly value-added fuels and chemicals is being pursued to reduce dependence on fossil fuel2,3,4,5,6,7. Particularly, photothermal catalysis, converting clean photons into heat to drive catalytic processes via various mechanisms8,9,10,11,12, has attracted wide attention. To boost the solar-to-chemical energy conversion efficiency via photothermal effect13,14, it is critical to generate the required temperature with lower light power and to facilitate the catalytic process thermodynamically and kinetically over a long period of time15,16,17.

For maximizing the photothermal conversion, it is important to consider the net heating power of the catalyst from a heat balance analysis:

$${P}_{heat} ={P}_{light}-{P}_{rad}-{P}_{conduction-convection} \\ ={P}_{sun}-\varepsilon \sigma {T}^{4}-{P}_{conduction-convection} \\ =\int {I}_{light}(\lambda )\alpha (\lambda )d\lambda -\int {I}_{BB}(T,\,\lambda )\varepsilon (\lambda )d\lambda -h({T}_{surface}-{T}_{ambient})$$
(1)

Where ε is average emissivity, σ is Stefan–Boltzmann constant, Ilight(λ) is the energy distribution under the solar illumination of the AM1.5 spectrum, IBB(T, λ) is the energy distribution under the radiation spectrum of a blackbody at a given temperature T, α(λ) is absorptance at a given wavelength, ε(λ) is emissivity at a given wavelength and Pconduction–convection is the power caused by conduction and convection.

It is clear that to maximize the photothermal conversion, the ideal design is to enable maximized sunlight absorption (αlight) (0.28–2.5 μm) and minimized thermal emission (mid-infrared emissivity, εMIR) simultaneously, especially under high-intensity solar irradiation (Fig. 1a, b, Supplementary Fig. 1, Supplementary Tables 13 and Supplementary Note 1)18,19. For example, for a temperature of ~620 K, a common temperature for chemical reactions, the thermal emission from catalysts (under a light input power of 2 W cm−2) results in an energy loss of ~0.83 W cm−2 (accounts for 41.5% of the input light power), thereby significantly decreasing the energy conversion efficiency (Supplementary Note 2). For the part of enhancing the catalytic activity and stability, the ideal design is to facilitate the mass transfer of reactants and products2, while preventing catalysts from sintering and coking, especially at high temperature20,21.

Fig. 1: The engineered optically selective Janus design.
figure 1

a The simulated temperature at given εMIR and absorptance of sunlight. At high light power, εMIR plays a decisive role in the obtained temperature. b Schematic of depressing εMIR to obtain an ideal photothermal catalyst with both broadband and selective spectrum, as well as the AM 1.5 G solar spectrum and the radiation spectrum of a blackbody at 673 K. c Schematic of nanoparticles loaded within macroporous Ti3C2Tx as catalysts for photothermal catalysis. The flat Ti3C2Tx flakes depress εMIR, meanwhile, the porous structure facilitates the mass transfer of reactants, thus the designed catalyst effectively drives the photothermal catalysis process. d The selective spectrum of engineered optically selective catalyst, as well as the AM 1.5 G solar spectrum and the radiation spectrum of a blackbody at 673 K.

Recent years have witnessed significant efforts and advancements in the field of photothermal catalysis. Notably, Ouyang et al. demonstrated the significant enhancement of αlight by employing novel defect engineering on In2O3, leading to improved performance in photothermal catalysis22. He et al. proposed an elegant greenhouse design to confine the photothermal energy23. Ozin and He proposed Nb2C and Mo2TiC2-based catalysts for optimizing full spectrum absorption24,25. Ye et al. innovatively integrated catalysts and spectral selective external reactors, to effectively facilitate photothermal catalysis26,27. While several MXene structures demonstrate promising thermal emission properties28,29,30,31,32,33, and the MXene-based catalysts have been efficiently utilized in catalytic reactions24,25, the ideal photothermal catalysts, which possess three properties simultaneously: high solar absorption, low thermal emission, and activated catalytic process, remain unrealized thus far.

Herein, we propose and demonstrate an optically selective catalyst with Ti3C2Tx-based Janus design composed of overlying flat Ti3C2Tx flake layers and underlying hollow Ti3C2Tx spheres, which minimizes thermal emission, maximizes solar absorption and good catalytic activity simultaneously, thus achieving excellent photothermal catalytic performance (Fig. 1c). Specifically, the high in-plane permittivity (mid-infrared) of overlying flat Ti3C2Tx flake layers enables high αlight (~88%) and low εMIR (~21%), ideal for the photothermal effect (Fig. 1d)28,29,30,31,32,33; while the underlying hollow Ti3C2Tx spheres can facilitate the mass transfer of reactants and products, beneficial for the catalytic activity. Furthermore, this universal architecture can be integrated with various metal nanoparticle catalysts (e.g., ruthenium, palladium, nickel) for a wide range of reactions. When employed in the Sabatier reaction and reverse water-gas shift (RWGS), we demonstrated that reducing the εMIR by ~70% can lead to an approximately 300% increase in catalytic yield. Among them, the CO2 methanation yield reaches 3317.2 mmol gRu−1 h−1 at a light power of 2 W cm−2, setting a performance record among catalysts without active supports.

Results

Synthesis and characterization of the designed photothermal host and Ru-loaded MXene catalyst (Ru@m-Ti3C2Tx)

To simultaneously achieve high solar absorption, low thermal emission, and activated catalytic process in the catalyst, we employed a two-step approach including a hard-templated method and subsequent impregnation, constructing the photothermal host with selective spectrum and porous structure (Fig. 2a–e, details seen in Methods)34,35,36. To create the ideal photothermal layer, we utilized negative pressure during suction filtration to create the flat Ti3C2Tx flake layers (Fig. 2e). Noted that the intrinsic selective spectrum of Ti3C2Tx is dependent on its structure (details seen in Fig. 3 below). As the presence of channels for advancing mass transfer of reactants and products is crucial for photothermal catalytic process, the PMMA spheres were first wrapped by Ti3C2Tx flakes via polar groups. Subsequently, PMMA spheres were removed through thermal evaporation to obtain a macroporous Ti3C2Tx (m-Ti3C2Tx) membrane, which contains numerous channels or porosities (Fig. 2b–e and Supplementary Figs. 27). After loading the Ru nanoparticle catalyst, we can finally achieve the Ru@m-Ti3C2Tx with a porous side composited of macroporous Ti3C2Tx spheres and a flat side of Ti3C2Tx flakes (thickness ~70 μm, Supplementary Fig. 8).

Fig. 2: Characterizations of designed photothermal host and Ru-loaded MXene catalyst (Ru@m-Ti3C2Tx).
figure 2

a Schematic of the preparation process of Ru@m-Ti3C2Tx. be SEM image of PMMA (b), PMMA@Ti3C2Tx (c) and m-Ti3C2Tx (d, e). f TEM image of Ru@m-Ti3C2Tx. g HRTEM image of Ru@m-Ti3C2Tx and secondary electron image (inset). h HAADF-STEM image and corresponding elemental mapping images of Ru@m-Ti3C2Tx. i High-resolution HAADF-STEM image of a single Ru nanoparticle and the corresponding crystalline structure (inset).

Fig. 3: Comparison of different photothermal effects caused by two different side structures.
figure 3

a Schematic of the high photothermal effect of the flat side due to the low εMIR (~21%). b The Ti3C2Tx (002) pole figure for the flat side of m-Ti3C2Tx. c The spectrum of the flat side of m-Ti3C2Tx within the 0.28–18.3 μm wavelength range, as well as the AM 1.5 G solar spectrum and the radiation spectrum of a blackbody at 673 K. d Schematic of the low photothermal effect of the porous side due to the high εMIR (~91%). e The Ti3C2Tx (002) pole figure for the porous side of m-Ti3C2Tx. f The spectrum for the porous side of m-Ti3C2Tx within the 0.28–18.3 μm wavelength range, as well as the AM 1.5 G solar spectrum and the radiation spectrum of a blackbody at 673 K. g The spectrum simulation for the flat side of m-Ti3C2Tx with varying thicknesses within the 0.28–20 μm wavelength range. h The spectrum simulation for the porous side of m-Ti3C2Tx with varying pore diameters within the 0.28–20 μm wavelength range. i The thermal simulations of different sides. j The measured temperatures from infrared images under different light powers.

To characterize the designed photothermal host and the catalyst Ru@m-Ti3C2Tx, several measurements were performed. Inductively coupled plasma optical emission spectroscopy (ICP-OES) result reveals that the Ru content is ~0.75 wt%, which causes the diffraction peaks associated with metallic Ru could not be detected by X-ray diffraction (XRD) (Supplementary Fig. 6)37. The Ru 3d and Ru 3p peaks were clearly revealed by X-ray photoelectron spectroscopy (XPS), which confirms the presence of Ru in Ru@m-Ti3C2Tx (Supplementary Fig. 9)38. Figure 2f depicts the transmission electron microscopy (TEM) image of a typical Ru-Ti3C2Tx flake. The black dots are uniformly distributed on Ti3C2Tx flakes, which are Ru-particles prepared by the impregnation method. The high-resolution transmission electron microscopy (HRTEM) image at the edge of the layer indicates the lattice fringe spacing of ~1.3 nm, in agreement with the (002) plane of Ti3C2Tx (Fig. 2g). Meanwhile, the high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) image and corresponding elemental mappings confirm that C and Ti elements are distributed throughout the support while Ru mainly exists in the form of nanoparticles (Fig. 2h). To reveal the crystal structure, we further investigated a single Ru nanoparticle. The HAADF-STEM image shows that the distances between adjacent planes are 2.3 and 2.1 Å, corresponding to the lattice spacing of the (1-10) and (10-1) planes of metallic Ru, respectively. The result is consistent with the hexagonal close-packing crystal phase of Ru, thus the Ru nanostructures via this preparation process are considered to be crystalline Ru nanoparticles (Fig. 2i). Meanwhile, we probed the flat side of the Ru@m-Ti3C2Tx and the Ru@Ti3C2Tx catalyst, and we found that the Ru nanoparticle sizes on both sides typically fall within the range of 3–4 nm (Supplementary Figs. 10, 11). Of note, different metal nanoparticles can also be loaded onto our designed Ti3C2Tx host using a similar preparation method (see “Method” section for details) for different reactions.

Detailed analysis of its unique photothermal effect

To verify the enhanced photothermal effect from low εMIR in the designed Ru@m-Ti3C2Tx membrane, a detailed analysis of the relationship among surface structure–spectrum–thermal was performed. Specifically, the spectral features of Ru@m-Ti3C2Tx strongly depend on the surface structure irradiated under sunlight, which could be identified by the crystal orientations of nanoflakes33,39. On the flat side, the overall εMIR of m-Ti3C2Tx (2.8–18.3 μm) is only ~21%, which remarkably prohibits it from thermal radiation loss, thus promoting localized high temperature on this side of the designed catalyst (Fig. 3a). Typically, as a two-dimensional material, the multilayer overlapping structure of Ti3C2Tx forms the crystallographic orientation of a certain characteristic crystal plane (002), which can be determined by X-ray pole figure40,41. For this measurement, the 2θ value was fixed and the m-Ti3C2Tx membrane was rotated from 0° to 360° at series of tilt angles from 0° to 75°. Figure 3b and Supplementary Fig. 12a show a (002) pole figure for the flat side of m-Ti3C2Tx, with one spot at a tilt angle of 0°. The out-of-plane orientation of flat side was detected, corresponding to the (002) plane. In this orientation, similar to the natural crystalline structure of Ti3C2Tx, the flat side of the m-Ti3C2Tx membrane shows an excellent spectral selectivity, which exhibits an 88% absorptance in the sunlight wavelength range and a εMIR of 21% (Fig. 3c).

To further demonstrate the low εMIR on the flat side, we also performed the full-wave simulation using the finite-element method-based software package COMSOL Multiphysics (see Method for details). In the calculation model, periodic close-packing Ti3C2Tx spherical pores are used for simplicity. We used the permittivity data of Ti3C2Tx from previous work33. Meanwhile, we characterized the thickness of the top layer to be ~150 nm (Supplementary Fig. 13). Calculated emissivity spectra at the normal incidence for flat side with different pore diameters are shown in Fig. 3g and Supplementary Fig. 14. The relatively low εMIR is nearly independent with pore diameters. The phenomenon can be explained by the large real and imaginary permittivity of Ti3C2Tx. The infrared light is highly reflected from the top flat side due to the drastic impedance mismatching between reactant gas and Ti3C2Tx slabs, and the large imaginary permittivity of Ti3C2Tx also leads to the strong absorption of residual light even with hundreds of nanometers-thick slabs. As a result, light can hardly transmit to the porous Ti3C2Tx layer, and the low εMIR is nearly unchanged with different diameters of Ti3C2Tx pores.

Furthermore, to understand the attenuation of the photothermal effect of conventional photothermal catalysts from high εMIR, the irradiated surface was experimentally reconstructed. As we all know, multiple scattering is a common phenomenon in optics, which increases emissivity at the corresponding waveband. The porous side which consists of massive ~5 μm macroporous Ti3C2Tx spheres increase the εMIR via multiple scattering, providing an in-situ control group (Fig. 3d). According to the (002) pole figure for the porous side of m-Ti3C2Tx, there is non-orientation on the porous side (Fig. 3e and Supplementary Fig. 12b). It could be caused by the large angular bending of the Ti3C2Tx flakes. Figure 3f shows that the overall εMIR for the porous side of m-Ti3C2Tx (2.8–18.3 μm) rises to ~91%, which significantly increases the thermal emission. According to the simulation, strongly enhanced εMIR is found for porous side. The near-unity εMIR is due to the multiple scattering effect generated in Ti3C2Tx pore layers (Fig. 3h). Although the shell of a single pore is thick, the multiple scattering effect effectively increases the time and strength of light-matter interaction, leading to near-unity and broadband emission in infrared wavelengths. As a result, the engineered porous side of the m-Ti3C2Tx membrane possesses no crystal orientation and exhibits high emission characteristics in the 0.28–18.3 μm wavelength range, with an absorptance of 92% in the sunlight wavelength range and an εMIR increasing to 91% (Fig. 3f).

To further verify the enhanced photothermal effect from low εMIR, light–thermal conversion efficiency calculation, infrared images, and thermal simulation were employed to demonstrate the reduction of thermal loss. Benefit from the low εMIR, the light–thermal conversion efficiency (ηlight-th, at 698 K) of the flat side can reach a high value of 74.2% far higher than that of the porous side (32.2%) (details seen in Methods). This difference will be more obvious in the high-temperature region, as mentioned above. Under 2 W cm−2 light power, the highest temperatures at the center of the flat side can reach 742 K, while the porous side only reaches 656 K (Fig. 3i). This similar phenomenon also can be observed under other light powers. The temperature difference between the two sides increases with the light power (Supplementary Fig. 15). Furthermore, the infrared images also prove the temperature difference between the two different sides, and this difference also increases with the light power (Fig. 3j and Supplementary Fig. 16). The flat side of Ru@m-Ti3C2Tx possess higher heating rate than the porous side of Ru@m-Ti3C2Tx (Supplementary Fig. 17). Meanwhile, to systematically assess the photothermal process of the catalyst, we performed experiments and simulations to investigate the its cooling process. As a result, the cooling rate of the flat side of Ru@m-Ti3C2Tx is lower than that of the porous side of Ru@m-Ti3C2Tx in the high-temperature region, which attributes to the reduction of thermal radiation loss (Supplementary Figs. 18, 19 and Supplementary Note 3). The heat profile is also investigated to prove the uniform distribution of heat in the architecture (Supplementary Fig. 20 and Supplementary Note 4). Through the detailed analysis of surface structure–spectrum–thermal, it reveals that the enhanced photothermal effect originates from low εMIR on the flat side of the Ru@m-Ti3C2Tx membrane, which could serve as an ideal photothermal catalyst.

Boosting photothermal catalytic performance

This photothermal host design, composited of macroporous Ti3C2Tx spheres and Ti3C2Tx flakes, aims to enhance the photothermal effect via low εMIR and to provide channel for the mass transfer of reactants and products. To reveal the advantages of our catalyst design, CO2 hydrogenation performance was tested under full-Arc Xe lamp irradiation for the Ru@m-Ti3C2Tx membrane. Ti3C2Tx membrane and Ru@Ti3C2Tx membrane were also prepared and used as references (see Method for details). First, to clarify whether the photothermal host affects the Sabatier reaction, a series of control experiments were performed under light power 0.1–2 W cm−2 for pure Ti3C2Tx membrane. At different light powers, nearly no methane and intermediates can be detected, clearly indicating that Ti3C2Tx itself is not active to CO2 methanation (Supplementary Figs. 2123). As shown in Supplementary Fig. 24, the catalytic activity of the Ru@Ti3C2Tx membrane increases with the light power. However, its CH4 production rate normalized by the mass of Ru only reached 128.5 mmol gRu−1 h−1 at 2 W cm−1, which could result from insufficient contact of the reactant with the Ru catalyst. In sharp contrast, the porous side of Ru@m-Ti3C2Tx (Ru@m-Ti3C2Tx-p) membrane, despite having higher εMIR, possesses much higher CH4 production rate (the average value of 733.2 mmol gRu−1 h−1 at 2 W cm−1) at each light power compared with Ru@Ti3C2Tx membrane, which could be attributed to rapid mass transfer of reactants as well as products (Fig. 4a, the blue line). The contribution of mass transfer has been further verified by simulation and in-situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) (Supplementary Figs. 2529, Supplementary Table 4 and Supplementary Note 5).

Fig. 4: Photothermal catalytic performance of the designed optically spectral catalyst and mechanism analysis.
figure 4

a CH4 production rate on the two different sides of the Ru@m-Ti3C2Tx catalyst. b CO2 conversion rate and selectivity on the flat side of the catalyst. c CO2 conversion rate and selectivity on the porous side of the catalyst. d Stability test for hydrogenation of CO2 at 2 W cm−2 without external heating. e Performance comparison of CO2 reduction among this work and previous works. f In-situ DRIFTS over flat side of Ru@m-Ti3C2Tx. g CO2 conversion rate on other catalysts with our design strategy. h Energy landscape for CO2 hydrogenation to CO and CH4 on the typical (0001) surface of Ru particles. To improve legibility, H2 is omitted from the labels after the initial state. All error bars represent the standard deviations of three independent measurements and the bars indicate mean values.

Furthermore, the reduction of the εMIR can increase the photothermal effect and further improve the methane yield of photothermal CO2 reduction. As shown in Fig. 4a, the red line, catalytic activity based on the flat side of the Ru@m-Ti3C2Tx (Ru@m-Ti3C2Tx-f) increases with the light power, which outperforms Ru@Ti3C2Tx and Ru@m-Ti3C2Tx-p under all light powers, especially at high intensities. Under the highest power of 2 W cm−2, the average CH4 production rate of Ru@m-Ti3C2Tx-f reaches 3317.2 mmol gRu−1 h−1, which is over 4 times higher than that of Ru@m-Ti3C2Tx-p (Supplementary Fig. 30). On the other hand, to achieve the same production rate (733.2 mmol gRu−1 h−1), Ru@m-Ti3C2Tx-f could reduce about 0.4 W cm−2 power compared with that of Ru@m-Ti3C2Tx-p at 2 W cm−2. Furthermore, the high product selectivity for CH4 can be detected on both Ru@m-Ti3C2Tx-p and Ru@m-Ti3C2Tx-f, probably due to the same size of Ru nanoparticles which does not vary with the irradiated surface (Fig. 4b, c)42. In order to eliminate potential differences in catalytic activity due to the placement state of the catalysts, the Arrhenius plots of Ru@m-Ti3C2Tx-p and Ru@m-Ti3C2Tx-f were constructed to assess kinetic performance. The results demonstrate that no noticeable changes were observed in the apparent activation energies (Ea) of these two sides of catalysts, indicating that the reaction energy remains unchanged with the placement state (79.8 ± 4.2 kJ mol−1 for Ru@m-Ti3C2Tx-p and 76.5 ± 8.6 kJ mol−1 for Ru@m-Ti3C2Tx-f) (Supplementary Fig. 31).

To demonstrate the high stability of our catalyst with the proposed design, we first performed 20-h consecutive testing and another 10-h consecutive testing for Ru@m-Ti3C2Tx-f (Fig. 4d). Between the two tests, the Ru@m-Ti3C2Tx-f was kept in argon gas at room temperature for two months. During the first 20-h test, both the CO2 conversion rate and product selectivity are almost no change. Surprisingly, the next 10-h test still exhibits stable selectivity in spite of the slightly fluctuating yield of CH4. The stable activity and selectivity can be attributed to the fact that the strong physical barrier between Ti3C2Tx layers resists the Ostwald ripening of Ru nanoparticles (see following post-catalysis characterizations). As shown in Supplementary Fig. 32, the hollow structure of Ru@m-Ti3C2Tx is nearly not changed and the overall structure was kept intact after 30 cycles of testing. Meanwhile, the absorptance of visible light still reaches 87.4%, and the emissivity in the mid-infrared region remains at 22.0% (Supplementary Fig. 33). Besides, the Ru nanoparticle still possesses good crystallinity and high dispersion, and its combination with the Ti3C2Tx has no obvious change (Supplementary Figs. 34, 35). The high stability may attribute to the demonstrated structural stability of the MXene analogue and the potential interaction between nanoparticle and Ti3C2Tx43,44. We compared the CH4 production rate to typical reports of CO2 methanation with or without an external heater (Fig. 4e and Supplementary Table 5). The results indicate that, at high optical power levels, whether based on the mass of Ru or the entire catalyst, our catalyst exhibits significantly higher production rates compared to the reported works, and it achieves a yield record among catalysts without active supports.

Our generic design strategy is expected to serve the vast majority of photothermal catalytic reactions. In order to prove the efficient catalysis performance resulting from the reduction of the εMIR, Pd, and Ni catalysts were also loaded on to m-Ti3C2Tx as active metal catalysts. The low yields are obtained compared with Ru-based catalysts, which ascribed to the types of active metals testified by previous works15. However, both catalysts have several times the performance improvement reaching an excellent yield by the reduction of the εMIR without active supports (Fig. 4g). Furthermore, to demonstrate the generality of the Janus design towards various MXene materials, we also performed structural design on two other MXene, i.e. Ti3CNTx and Ti2CTx. As a result, both materials exhibit excellent selective spectra and complete porous structures (Supplementary Figs. 36, 37).

To confirm the origin of the as-produced CH4, the 13C isotope labeling experiment was performed for the photothermal hydrogenation. The products were examined by gas chromatography-mass spectrometry (GC–MS). A major signal at a mass/charge ratio of 17 on the mass spectrum corresponding to 13CH4 appears, which verifies that the as-detected CH4 originates from the CO2 hydrogenation process (Supplementary Fig. 38). We also performed in-situ DRIFTS under reaction conditions to reveal the possible reaction mechanism of CO2 methanation on our Ru@m-Ti3C2Tx. The measurements were performed at 300 °C in the reaction gas mixture of 80 vol% H2/20 vol% CO2. As shown in Fig. 4f and Supplementary Fig. 39, the peak at 1649 cm−1 is characteristic of surface CO2* species, namely bicarbonate45,46. As the reaction proceeded, the formate species gradually emerged according to the peaks at 1585, 1393, 1374, and 1355 cm−1, which means the transformation of bicarbonate to formate47,48. Meanwhile, the peak at 1040 cm−1 for CH3O* is detected, demonstrating the further hydrogenation of HCOO*. Consequently, the signal intensities attributed to CH4 species (located at 3016 and 1304 cm−1) increased gradually as the reaction progressed. The peak assignments of the surface species are listed in Supplementary Table 6. The reaction diagram of CO2 hydrogenation on the (0001) surface of Ru was explored by using density functional theory (DFT) calculations to further reveal the reaction mechanisms (Fig. 4h)49,50. The adsorption and conversion of CO2 on the Ru site are more favorable than other sites. Although both reactions CO2* + H* → HOCO* and CO2* + H* → HCOO* are exothermic. The reaction CO2* + H* → HCOO* (ΔrG = −0.956 eV) is energetically more preferred than CO2* + H* → HOCO* (ΔrG = −0.555 eV). So, we conclude that HCOO* is the dominating species of CO2 reduction on Ru surface, which is in consistent with the in-situ DRIFTS observations. Whereafter, the C-O bond cleavage of HOCO* and the subsequent desorption at the catalytic site render the entire process highly energetically unfavorable (ΔrG = 2.355 eV). In sharp contrast, the C-O bond cleavage of HCOO*, followed by its successive hydrogenation until the final formation of CH4, results in the entire process highly energetically favorable (ΔrG = −0.989 eV), which is also consistent with the in-situ DRIFTS observations. Therefore, it can be inferred that CO2 hydrogenation on the surface of Ru exhibits good selectivity towards CH4.

Discussion

In summary, we demonstrated an optically selective catalyst based on 3D Ti3C2Tx structures, which possess low εMIR, high αlight, and fast mass transfer kinetics. Ti3C2Tx flakes with high in-plane permittivity in the mid-infrared enable the ideal εMIR to reduce the thermal emission of catalysts, while the designed 3D porous structure composed of Ti3C2Tx spheres provides numerous channels for mass transfer of reactants/products. In addition, this universal architecture can be integrated with various nanoparticle catalysts for a wide range of reactions. Taking Sabatier reaction and RWGS as demonstrations, we achieved around 3 times and 4 times improvement in yield for CH4 and CO compared to that without this design, respectively. The CO2 methanation yield reaches 3317.2 mmol gRu−1 h−1 at a light power of 2 W cm−2, setting a performance record among catalysts independent of active supports. We believe that the design of the optically selective catalyst not only provides a new route for the reduction of CO2 in photothermal catalysis but also facilitates various photothermal-driven chemical reactions toward green chemistry.

Methods

Preparation of Ti3C2Tx membrane

According to the previous work, the preparation of Ti3C2Tx includes five steps30. First, 2 g lithium fluoride and 40 mL hydrochloric acid were stirred in a Teflon beaker (100 mL) for 30 min. Second, 2 g MAX-Ti3AlC2 powder was added to the above solution, and the reaction temperature was adjusted to 35 °C, and kept this temperature for 24 h. Third, the obtained reaction liquid was divided into four parts for centrifugation (959 × g), and then the supernatant was discarded. 40 mL deionized water was added into four portions of precipitate, and they were thrown into high-power ultrasonic machine (750 W) for 10 min. The process was repeated by several times until the pH of the supernatant reached 5. Fourth, 40 mL ethanol was added to each precipitate, followed by ultrasound for 1 h (this will facilitate the separation of the layered Ti3C2Tx), followed by centrifuged (7826 × g), and the precipitate was collected. Finally, 20 mL deionized water was added into four portions of precipitate followed by ultrasonication (750 W, 20 min) and centrifugation (959 × g), and the black supernatant was collected as a few-layer Ti3C2Tx dispersion (the specific concentration can be adjusted by adding deionized water). The Ti3C2Tx membrane could be obtained via filtering as-prepared 20 mL few-layer Ti3C2Tx dispersion (1 mg mL−1). In addition, the Ti3CNTx and Ti2CTx were prepared in a similar method using MAX-Ti3AlCN and MAX-Ti2AlC as precursors, respectively.

Preparation of PMMA sphere

According to the previous work, the preparation of ~5 μm PMMA spheres is through a typical dispersion polymerization process36. Primarily, 50 mL of methanol, 118.4 mg of azobisisobutyronitrile, and 1.578 g of polyvinyl pyrrolidone (PVP, molecular weight of 58,000) were added sequentially to a 100 mL flask and stirred to dissolve. Subsequently, the argon was pumped into the solution for 3–4 min to fill the flask with a protective atmosphere. Finally, 4.383 mL methyl methacrylate and a stirrer were added to the flask, and the polymerization was carried out with magnetic stirring at 55 °C for 24 h. PMMA spheres can be stored after cleaning with methanol.

Preparation of Ru@m-Ti3C2Tx, Ru@Ti3C2Tx, Pd@m-Ti3C2Tx, Ni@m-Ti3C2Tx, m-Ti2CTx and m-Ti3CNTx

The preparation of Ru@m-Ti3C2Tx mainly consisted of two processes: a hard template method and further an impregnation method34. First, as-prepared 20 mL few-layer Ti3C2Tx dispersion (1 mg mL−1) and 40 mL PMMA sphere dispersion (2 mg mL−1, dispersed by deionized water) were stirred in a 100 mL flask for 24 h. Note that the induced electrostatic self-assembly effect comes from the surface functional groups of Ti3C2Tx and PMMA. Next, the PMMA@Ti3C2Tx membrane was obtained by suction filtration and further annealed in flowed argon (100 sccm) at 500 °C to form an m-Ti3C2Tx membrane. Here, the m-Ti2CTx and m-Ti3CNTx can be synthesized using a similar method by replacing Ti3C2Tx with Ti2CTx and Ti3CNTx, respectively. Finally, the RuCl3·3H2O solution was added dropwise to Ti3C2Tx membrane support and m-Ti3C2Tx membrane support and annealed at 350 °C under mixture gas (10% H2 and 90% Ar) for 1 h to obtain Ru@Ti3C2Tx and Ru@m-Ti3C2Tx, respectively. Palladium and nickel are also loaded by a method similar to those described above. PdCl2 and NiCl2 are the sources of palladium and nickel, respectively.

Characterization

XRD was performed on a Bruker D8 X-ray diffractometer with Cu Kα radiation. BET was performed on a Micromeritics ASAP 2460. SEM images were obtained by Tescan MIRAS3 SEM. TEM images were obtained with a Talos F200E TEM. The UV wavelength range in the 0.28–2.5 μm range was measured by ultraviolet-visible spectroscopy (UV-3600, Shimadzu) with an attached integrating sphere (ISR-3100). The NIR wavelength within the 2.5–18.3 μm range was measured by an FTIR spectrometer (Nicolet IS50, ThermoFisher) equipped with a gold integrating sphere (IntergatIR MIR, Pike). In-situ FTIR spectra were measured by an FTIR spectrometer (Nicolet IS50, ThermoFisher) equipped with a reaction tank (Hefei In-situ Technology). TGA was performed on a TG/DTA7300 (Seiko In.). The elemental valence states were investigated by X-ray photoelectron spectroscopy (Thermo Scientific K-Alpha). The invisible temperature data was collected via coaxial laser infrared thermometer. The infrared images were taken using a FLUKE Tix 580 infrared camera. The Ru content was measured on an Agilent 5110 ICP-OES spectrometer.

Structural optical simulation

All calculations are performed using the finite-element method (FEM) based software package COMSOL Multiphysics. In the simplified calculation model, three-dimensional square close-packing MXene spherical pores with an equal diameter and a MXene slab layer are used, see the schematic view of a unit cell in Fig. 3g, the number of MXene spherical pores along the z-direction is dependent on the total thickness (60~70 μm) and pore diameters. The shell thickness of MXene pores is 8 nm. The periodic boundary condition is used along x- and y- directions. Two ports at the top (reflection plane) and the bottom surfaces (transmission plane) of the FEM model are used for calculating the reflectivity (R) and the transmissivity (T), the absorptivity (A) or emissivity (E) is defined as A(E) = 1-R-T. The distance between ports and adjacent MXene structures is larger than the calculated wavelength. To simulate the ultrathin shell of pores, the transition boundary condition with an 8-nm thickness is utilized for each MXene pore. Free tetrahedral meshing with a maximum size of 1 μm is adopted for pores and adjacent air volumes. Swept meshes with 5 distributions and 40 distributions are utilized for the slab region and the air volume at both ends, respectively.

Thermal-field simulation

The thermal-field simulation is through COMSOL Multiphysics® software. The size and structure of the reactor in the simulation are the same as that in the experiment, which is shown in Supplementary Fig. 40. The heat source, whose power density (Pin) is set from 0.1 to 2 W cm−2, is added above the material to simulate solar radiation used in the experiment. The solar-radiation absorption of the catalyst is simulated by multiplying the Pin by the measured absorptivity (0.88 and 0.92) in the visible band. The thermal emission of the catalyst is simulated by the surface-to-surface-radiation module, where the emissivity of the catalysts is set to 0.21 and 0.91, respectively. The heat-transfer-in-solid module is used to simulate the thermal convection between the reactor and the external environment with a heat transfer coefficient of 5 W m−2 K−1, where the temperature of the external environment is fixed at 293.15 K. The reactor body is defined as AISI 316 L stainless steel with 18.4 W m−1 K−1 thermal conductivity, 7980 kg m−3 density, and 502 J kg−1 K−1 constant pressure specific heat capacity. The optical window is defined as Sapphire crystal with 42.5 W m−1 K−1 thermal conductivity, 4000 kg m−3 density, and 780 J kg−1 K−1 constant pressure specific heat capacity. The solid support is defined as quartz with 1.4 W m−1 K−1 thermal conductivity, 2210 kg m−3 density, and 730 J kg−1 K−1 constant pressure specific heat capacity.

Light–thermal conversion efficiency calculation

The spectrally averaged sunlight absorption (αlight) is defined as:

$${\alpha }_{light}=\frac{\int {I}_{light}(\lambda )\alpha (\lambda )d\lambda }{{I}_{light}}$$
(2)

The spectrally averaged thermal emissivity is calculated by

$${\varepsilon }_{MIR}(T)=\frac{\int {I}_{BB}(T,\,\lambda )\varepsilon (\lambda )d\lambda }{\sigma {T}^{4}}$$
(3)

Hence, the light–thermal energy conversion efficiency can be obtained by29

$${\eta }_{light-th}(T)={\alpha }_{light}-{\varepsilon }_{MIR}\frac{\sigma ({T}^{4}-{T}_{0}^{4})}{{I}_{light}}$$
(4)

Here, Ilight(λ), IBB(T, λ), α(λ) and ε(λ) represent the light power, energy distribution under the radiation spectrum of a blackbody at a given temperature T, and absorptance and the emissivity at the wavelength λ, respectively. And σ is the Stefan–Boltzmann constant, T is the operating temperature, and T0 is the ambient temperature.

Computational methods

We explored the origin of the as-produced CH4 by calculating various different reaction pathways. All the structures were optimized by the density functional theory (DFT) calculations as implemented in the Vienna Ab initio Simulation Package (VASP)51,52. The interaction between valence electrons and ion cores were accurately described using the projector-augmented wave (PAW) method52,53. The Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional54,55 was adopted to treat the interaction between valence electrons. The DFT-D2 method56 was used to describe the van der Waals interaction between molecules and Ru (0001) substrate. We used a plane wave basis set with an energy cutoff of 450 eV in the calculations. All the structures were fully relaxed, and the convergence criteria for energy and force were set at 10−5 eV and 10−4 eV/Å, respectively. The relative energy ΔrG of the two systems was obtained from ΔrG = G1G2, where G2 is the Gibbs free energy of the second system, G1 is the Gibbs free energy of the first system.

CO2 hydrogenation measurements

The gas-phase CO2 hydrogenation experiments were conducted in a batch reactor (100 mL) with internal gas circulation (CEL-HPR, CEAU-light). In the CO2 hydrogenation tests, ~4 mg of catalyst membrane was placed on a hollowed-out shelf. Ru loading mass is around 0.75 wt%. The infrared images were taken using a FLUKE Tix 580 infrared camera. A 300 W Xe lamp (PLS-SXE300D, Beijing Perfect light) was used to drive photothermal CO2 hydrogenation. The reactor is cleaned by bubbling and evacuating the mixture of CO2 and H2 (1:4) thrice. Subsequently, when the injection pressure reached 4 bar, the reactor was sealed, the Xe lamp was turned on, and the photothermal catalytic reaction began. After the reaction, 1 mL of gas was taken out of the reactor and injected into the gas chromatography (GC-2014, Shimadzu) to detect the composition of the gas (the amounts of CO and CH4 were analyzed with a flame ionization detector installed in the same GC).

The 13C isotope labeling experiment was also performed in the above reactor with 13CO2 (99.9 atom %, Sigma–Aldrich) and H2. Before the catalytic reaction, the reactor was also degassed and purged with a mixture of 13CO2 and H2 (1:4) thrice to eliminate the interference of impurity gas to the experiment. The 13C isotope product gases were measured using an Agilent 6890 A gas chromatographic mass spectrometer.

The CH4 production rate (RCH4), CO production rate (RCO) and CO2 conversion rate (RCO2) were defined as follow:

$${R}_{CH4}=\frac{{n}_{CH4}}{{m}_{cat}{\omega }_{Ru}t}$$
(5)
$${R}_{CO}=\frac{{n}_{CO}}{{m}_{cat}{\omega }_{Ru}t}$$
(6)
$${R}_{CO2}={R}_{CH4}+{R}_{CO}$$
(7)

The selectivity of CH4 is defined as follow:

$${S}_{CH4}=\frac{{R}_{CH4}}{{R}_{CH4}+{R}_{{{\rm{CO}}}}}$$
(8)

n is the yield of products (mmol), mcat is the mass of the catalyst (g), ωRu is the loading percent of Ru, and t is the irradiation time (h).