Oxygen-deficient metal oxides supported nano-intermetallic InNi3C0.5 toward efficient CO2 hydrogenation to methanol

INTRODUCTION

The rapid growth of carbon-based energy consumption along with the global economic development are responsible for a massive emission of carbon dioxide (CO2), raising its atmospheric concentration from suitable 300 up to 415 ppmv (parts per million by volume) in the last 60 years and causing serious problems of global warming, glacier melting, and ocean acidification (1). CO2 capture and storage technology potentially reduces this emission, but the high storage cost and uncertainty about CO2 leakage greatly limit the application of this technology, while converting CO2 into commodity chemicals is a promising approach to recycle massive quantities of CO2 (25). Among the commodity chemicals, methanol is not only a clean alternative fuel for gasoline and diesel but also an excellent chemical platform to produce olefins and other high value-added chemicals commonly obtained from crude oil, with a worldwide demand of ~50 million tons per year (6, 7). In this context, the catalytic hydrogenation of CO2 to methanol (denoted as CO2-to-methanol) using renewable hydrogen (H2, produced by solar energy, hydropower, and wind power) has been attracting great attention for a CO2 circular economy (8). Moreover, this reversible reaction also shows substantial potential to be used as a H2 storage distribution system for applications in H2-O2 fuel cells (9). However, this is a grand challenge because of the chemical inertness of the CO2 molecule (5, 10). Accordingly, substantial catalytic advances are urgently required for the large-scale hydrogenation of CO2 to methanol.

Over the past decades, photo- and electro-catalytic CO2 hydrogenation to methanol has been greatly advanced but still suffers from low productivity originating from the low photo-/electro-energy density (4, 11). A variety of homogeneous complexes enables a high methanol yield under relatively mild conditions (12), but their practical applications are limited by their high prices and complicated operations. Heterogeneous catalysts are composed of active components loaded on solid supports, endowed with the superior ability to address the operation issues. Nano-copper catalysts supported on metal oxides were extensively studied (5, 10, 13, 14), but severe problems exist, including low methanol selectivity (usually below 60%) caused by the competing reverse water-gas shift (RWGS) reaction and rapid deactivation by Cu sintering (15). A range of precious metal catalysts, such as Pd/In2O3 (16) and Au/ZnO(CeOx/TiO2) (17, 18), were successfully used as a replacement of copper but are compromised by their limited natural abundance. A series of bimetallic oxides were found to be promising for this reaction, such as 5% CO2 conversion, 99.8% methanol selectivity, and 0.295 gMeOH gcat−1 hour−1 methanol productivity on In2O3/ZrO2 (19), and 10% CO2 conversion, 86% methanol selectivity, and 0.73 gMeOH gcat−1 hour−1 methanol productivity on ZnO-ZrO2 (20). Recently, Nørskov and coworkers (21) discovered a Ni5Ga3 intermetallic catalyst with the aid of theoretical calculations, achieving CO2 conversion of 4.9%, methanol selectivity of 44.8%, and methanol productivity of 0.1 gMeOH gcat−1 hour−1 at atmospheric pressure. García-Trenco et al. (22) unveiled a PdIn catalyst, exhibiting CO2 conversion of 0.6%, methanol selectivity of above 80%, and methanol productivity of 0.13 gMeOH gcat−1 hour−1. Despite the fact that no intermetallics have proved to be superior over the catalysts reported ever, it is worthy to explore these emerging catalyst candidates for CO2 hydrogenation, because the intermetallic compounds have facilely tunable components, variable constructions, and reconfigurable electronic structures, and great progress has been made on nano-intermetallic catalysts for CO2 hydrogenation to methanol.

Very recently, the nano-intermetallic compound InNi3C0.5 structured on an Al2O3/Al-fiber, with superior RWGS performance at and above 400°C, was also found to be highly selective for the CO2-to-methanol reaction below 300°C but not active enough (23). Notably, the electronic metal-support interaction (EMSI) is paramount to improve the catalyst performance via tuning the electronic properties of metal nanoparticles by supports (2426). Campbell (24) showed that small platinum clusters experience large electronic perturbation when in contact with ceria, strongly enhancing the catalytic performance for the water-gas shift reaction. Recently, Rodriguez et al. (27) reported the advantages of metal-oxide and metal-carbide interfaces for CO2 conversion to methanol and confirmed that the metal-support interactions modify the electronic properties of metals. Moreover, it is interesting to recognize that zirconia (ZrO2) with oxygen vacancies traps electrons at vacancy centers and modulates the electronic states of as-anchored metal nanoparticles (28, 29). For instance, Ni et al. (29) successfully tuned the electron density of Ni particles by ZrO2 with oxygen vacancies to enhance the hydrogenation of fatty acids to alcohols. These findings may give an interesting hint to develop a high-performance CO2-to-methanol catalyst by dispersing InNi3C0.5 nanoparticles onto reducible oxides that can generate abundant oxygen vacancies. To check this idea, we chose three ZrO2 supports that can generate different amounts of oxygen vacancies, including monoclinic-ZrO2 (m-ZrO2), tetragonal-ZrO2 (t-ZrO2), and amorphous-ZrO2 (a-ZrO2), to tailor the catalysts. Among them, InNi3C0.5/m-ZrO2 shows an excellent performance with 11.2% CO2 conversion and 85.4% methanol selectivity under the typical reaction conditions. A combined study of spectroscopic and electron microscopic methods, and theoretical calculations confirms that the electron structure of InNi3C0.5 is tuned by ZrO2 (especially by m-ZrO2 with abundant oxygen vacancies), accompanied by the gradually enhanced CO2 activation on the InNi3C0.5 surface, thereby leading to remarkable improvement of the CO2 conversion to methanol. According to this inspiring clue, a more efficient and affordable InNi3C0.5/Fe3O4 catalyst, with further enhanced EMSI between InNi3C0.5 and Fe3O4, was tailored by carburizing a In2O3-NiO/Fe2O3 precursor. This catalyst achieves 25.7% CO2 single-pass conversion and 90.2% methanol selectivity at 325°C, gas hourly space velocity (GHSV) of 36,000 ml gcat−1 hour−1, H2/CO2 molar ratio of 10:1, and 6.0 MPa [or a high space time yield (STY) of 2.62 gMeOH gcat−1 hour−1 with 18.8% conversion and 92.8% selectivity using a high GHSV of 115,500 ml gcat−1 hour−1]. Moreover, this catalyst shows high resistance to sulfur poisoning. We take a big step forward in tailoring of a stable and highly active/selective catalyst for efficient synthesis of methanol from CO2 hydrogenation.

RESULTS

Structural, morphological, and textural features of InNi3C0.5/ZrO2 catalysts

We initially loaded InNi3C0.5 on m-, t-, and a-ZrO2 supports via incipient wetness impregnation and subsequent carburization (details in Materials and Methods). These catalysts were probed by x-ray diffraction (XRD), clearly identifying the characteristic patterns of InNi3C0.5 (PDF#28-0468), m-ZrO2 (PDF#86-1449), and t-ZrO2 (PDF#50-1089) (Fig. 1A). a-ZrO2 in InNi3C0.5/a-ZrO2 was partially transformed into t-ZrO2 during the carburization process at 600°C, as the tetragonal phase is thermodynamically more stable than the amorphous phase at high temperature (30). The main diffraction peaks of InNi3C0.5 at 41.3° and 48.1° in these catalysts are similarly sharp and strong, indicating comparable grain size and crystallinity of InNi3C0.5. The transmission electron microscopy (TEM) images illustrate that these three catalysts exhibit a uniform dispersion of InNi3C0.5 grains with an average size of 16.0 ± 0.5 nm (Fig. 1, B to D). Moreover, the high-resolution TEM images (Fig. 1, E to G) show the lattice fringes of InNi3C0.5 (1-10) with spacing of 0.267 nm, indicating that the dominant exposed facet of InNi3C0.5 is the (111) plane (see detailed results and analysis in figs. S1 and S2). These three catalysts show rough and porous surface morphology aggregated from irregular-shaped lumps of 300 to 500 nm (fig. S3) and mesoporous feature with dominant mesopore size centered at 10 to 20 nm (fig. S4). The InNi3C0.5/t-ZrO2 catalyst presents the largest specific surface area (SSA) of 52.0 m2 g−1 (table S1), followed by InNi3C0.5/m-ZrO2 (10.0 m2 g−1) and InNi3C0.5/a-ZrO2 (5.0 m2 g−1).

Fig. 1 Structures of ZrO2 supports and corresponding catalysts.

(A) XRD patterns of m-ZrO2, t-ZrO2, a-ZrO2, InNi3C0.5/m-ZrO2, InNi3C0.5/t-ZrO2, and InNi3C0.5/a-ZrO2. TEM images of (B) InNi3C0.5/m-ZrO2, (C) InNi3C0.5/t-ZrO2, and (D) InNi3C0.5/a-ZrO2 (insets: corresponding size distribution of the InNi3C0.5 nanoparticles). a.u., arbitrary units. High-resolution TEM images of a typical InNi3C0.5 nanoparticle supported on (E) m-ZrO2, (F) t-ZrO2, and (G) a-ZrO2 [insets: lattice fringes with distance of 0.267 nm corresponding to the InNi3C0.5 (1-10) crystal plane, 0.363 nm to the m-ZrO2 (110) crystal plane, and 0.360 nm to the t-ZrO2 (100) crystal plane].

Dependence of catalytic performance on ZrO2 type

The InNi3C0.5/ZrO2 catalysts were comparatively investigated for methanol synthesis from CO2 hydrogenation in a continuous-flow fixed-bed tubular reactor, under the reaction conditions of 300°C, 4.0 MPa, H2/CO2 molar ratio of 3:1, and GHSV of 12,000 ml gcat−1 hour−1 (optimized as shown in fig. S5). The InNi3C0.5/SiO2 catalyst offered a high methanol selectivity of 90% but a very low CO2 conversion of only 2.5% (Fig. 2A; much lower than the thermodynamic equilibrium conversion of 11.6%; fig. S6). Our InNi3C0.5/ZrO2 catalysts all raised CO2 conversion markedly compared to InNi3C0.5/SiO2, but interestingly, their catalytic performance showed a remarkable ZrO2-type dependence. Only InNi3C0.5/m-ZrO2 can achieve a CO2 conversion (11.2%) close to thermodynamic equilibrium with an acceptable methanol selectivity of 85.4%. InNi3C0.5/t-ZrO2 and InNi3C0.5/a-ZrO2 delivered moderate conversions of 3.5 to 5.0% with similar methanol selectivity of 85 to 90%. As references, the pure ZrO2 and SiO2 supports were also tested in this reaction but yielded no more than 0.7% CO2 conversion (Fig. 2A and table S2). Moreover, the STY of methanol was calculated to further assess the catalytic performance of the ZrO2-supported InNi3C0.5 catalysts. InNi3C0.5/m-ZrO2 exhibited a high STY of 0.62 gMeOH gcat−1 hour−1 at 300°C, much higher than that of InNi3C0.5/t-ZrO2 (0.30 gMeOH gcat−1 hour−1), InNi3C0.5/a-ZrO2 (0.20 gMeOH gcat−1 hour−1), and InNi3C0.5/SiO2 (0.18 gMeOH gcat−1 hour−1) (Fig. 2B and table S3), as well as most reported catalysts (table S4).

Fig. 2 Catalytic performance of various catalysts.

(A) CO2 conversion and product selectivity (300°C, 4.0 MPa, H2/CO2 = 3:1, and a GHSV of 12,000 ml gcat−1 hour−1). (B) STY of methanol and TOF (300°C, 4.0 MPa, H2/CO2 = 3:1, and a GHSV of 24,000 ml gcat−1 hour−1).

These catalysts have similar surface morphology (aggregation of irregular shaped lumps; fig. S3) and InNi3C0.5 grain size (~16 nm; Fig. 1, B to D), excluding the responsibility for their discrepancy of activity for CO2 hydrogenation. To assess the intrinsic activity, their turnover frequencies (TOFs; defined as the number of reactant consumed on an active site per unit time) were measured. Not surprisingly, InNi3C0.5/m-ZrO2 offered the highest TOF of 72.2 hour−1 (Fig. 2B and table S3), being three and two times as high as that of InNi3C0.5/a-ZrO2 (23.8 hour−1) and InNi3C0.5/t-ZrO2 (34.2 hour−1). Apparently, a special EMSI between ZrO2 and InNi3C0.5 is generated and accounts for the improvement of the catalyst activity, while the EMSI shows strong ZrO2-type dependence. However, the nature of this ZrO2-type dependence of the catalytic performance-relevant EMSI is still not clear.

Oxygen vacancy relevant EMSI

The EMSI between InNi3C0.5 and ZrO2 supports was first explored by the quasi–in situ x-ray photoelectron spectroscopy (XPS) technique, with the spectra displayed in Fig. 3. For the InNi3C0.5/SiO2 catalyst, the binding energies of Ni 2p3/2 and Ni 2p1/2 located at 852.7 and 870.1 eV (Fig. 3A) are equal to those of the pure Ni0 metal. By comparison, the binding energy of Ni 2p3/2 for InNi3C0.5/a-ZrO2, InNi3C0.5/t-ZrO2, and InNi3C0.5/m-ZrO2 is respectively lowered from 852.7 to 852.5, 852.4, and 852.1 eV, showing the gradually enriched electron density especially of m-ZrO2–supported InNi3C0.5. Similarly, the electron enrichment in In and C elements is also observed on InNi3C0.5/t-ZrO2 and especially InNi3C0.5/m-ZrO2 (Fig. 3B and fig. S7): For example, the binding energy of In 3d5/2 (31) shifts from 443.6 eV (for InNi3C0.5/SiO2 and InNi3C0.5/a-ZrO2) to 443.3 (for InNi3C0.5/t-ZrO2) and 443.2 eV (for InNi3C0.5/m-ZrO2). The above results confirm the existence of the ZrO2 type–dependent EMSI and indicate that the strongest EMSI takes place between InNi3C0.5 and m-ZrO2. It should be noticed that some Ni2+ (at 855.4 eV) and In3+ (at 444.5 eV) species are observed on all ZrO2-supported InNi3C0.5 catalysts (Fig. 3, A and B), which are likely from the dissolving of Ni2+ and In3+ ions into the ZrO2 lattice (29, 32). In addition, the peak areas of the XPS spectra of surface Ni2+ and In3+ species in all ZrO2-supported InNi3C0.5 catalysts are almost identical, and therefore, we believe that the Ni2+ and In3+ species are not responsible for the discrepancy of their activity for CO2 hydrogenation.

Fig. 3 Electronic states and ability to activate CO2 of ZrO2 supports and corresponding catalysts.

XPS spectra in (A) Ni 2p and (B) In 3d regions of InNi3C0.5/SiO2, InNi3C0.5/m-ZrO2, InNi3C0.5/t-ZrO2, and InNi3C0.5/a-ZrO2. XPS spectra in (C) Zr 3d region for a-ZrO2 (as the reference), InNi3C0.5/m-ZrO2, InNi3C0.5/t-ZrO2, and InNi3C0.5/a-ZrO2, and analysis fittings in table S5. (D) CO2-TPD-MS profiles for unsupported InNi3C0.5 nano-intermetallic, pure SiO2, a-ZrO2, t-ZrO2, and m-ZrO2 supports. (E) CO2-TPD-MS profiles for InNi3C0.5/SiO2, InNi3C0.5/a-ZrO2, InNi3C0.5/t-ZrO2, and InNi3C0.5/m-ZrO2 catalysts. MS signal of the carbonaceous species for CO2 desorption: CO2 signal [mass/charge ratio (m/z) = 44] and CO signal (m/z = 28). (F) Plot of the TOF as a function of the amount of CO desorption (reaction conditions for TOF measurements: 300°C, 4.0 MPa, H2/CO2 = 3:1, and GHSV = 24,000 ml gcat−1 hour−1).

The XPS spectrum of Zr 3d5/2 in pure a-ZrO2 can be deconvoluted into two peaks with binding energies of 181.5 and 182.0 eV (Fig. 3C), respectively, assigned to partially reduced Zrδ+ (denoted as ZrI, δ < 4; related to oxygen vacancy) and stoichiometric ZrO2 (ZrII, Zr4+) (33). In comparison with pure a-ZrO2, the three catalysts have more ZrI species while showing lowered binding energy of Zr 3d5/2 from 181.5 eV to 181.2 to 181.4 eV (Fig. 3C). InNi3C0.5/m-ZrO2 has the highest amount of ZrI species (i.e., oxygen vacancies) and the lowest binding energy of Zr 3d5/2 (i.e., enriched electron density), followed by InNi3C0.5/t-ZrO2 and then InNi3C0.5/a-ZrO2 (table S5 and figs. S8 and S9). This sequence is in accord with the electron density tendency of InNi3C0.5/m-ZrO2 > InNi3C0.5/t-ZrO2 > InNi3C0.5/a-ZrO2 (Fig. 3, A and B, and fig. S7). We are thus confident that the EMSI is gradually enhanced with the increase in ZrI species (i.e., oxygen vacancies) of the ZrO2 supports, which accounts for the ZrO2 type–dependent EMSI and the improvement of CO2 hydrogenation activity especially of the InNi3C0.5/m-ZrO2 catalyst.

To further confirm this oxygen vacancy relevant EMSI and the dependence of EMSI strength on ZrO2 type, these three catalysts were further investigated using hydrogen temperature-programmed desorption (TPD), electron paramagnetic resonance (EPR), and in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) of CO adsorption, with the results shown in fig. S10. All catalysts offer one H2 desorption peak at 110°C (fig. S10A), which is ascribed to the H-species adsorbed on the InNi3C0.5 surface. The area of this peak is almost identical for these catalysts, in line with their same particle size and amount of InNi3C0.5, but InNi3C0.5/m-ZrO2 shows a narrower peak than the other two catalysts, suggesting a quite uniform electron structure of InNi3C0.5 surface in comparison with the other two (34). Most notably, a strong H2 desorption peak appears at 620°C for InNi3C0.5/m-ZrO2, attributed to the H anions (Hδ−) held at coordinatively unsaturated ZrI sites (29, 35), but a very weak peak at 576°C for InNi3C0.5/t-ZrO2 even nothing (at 550° to 650°C) for InNi3C0.5/a-ZrO2. As generally acknowledged, the surface of some reducible oxides (e.g., CeO2, TiO2, and ZrO2) can be partially reduced by H2 in association with the oxygen vacancy formed (36), and the as-formed oxygen vacancies are negatively charged with electron density maximally localized at the vacancy center (28). Clearly, the amount of Hδ− on ZrI sites (i.e., oxygen vacancies) follows the order InNi3C0.5/m-ZrO2 > InNi3C0.5/t-ZrO2 > InNi3C0.5/a-ZrO2 (fig. S10A). Moreover, the EPR spectra of these three InNi3C0.5/ZrO2 catalysts show axial signals with g value of 2.003 (fig. S10B), further indicating the presence of oxygen vacancies (37). As expected, the signal intensity of InNi3C0.5/m-ZrO2 is much higher than that of InNi3C0.5/t-ZrO2 and followed by InNi3C0.5/a-ZrO2. Undoubtedly, InNi3C0.5/m-ZrO2 has a much higher density of oxygen vacancies than the other two, in good agreement with our XPS results (table S5 and figs. S8 and S9) and literature results (29). In addition, as shown in fig. S10C, linearly adsorbed CO with infrared band at 2077 cm−1 (23) is detected on InNi3C0.5/SiO2 at room temperature. This linear adsorption of CO is also observed on the InNi3C0.5/ZrO2 catalysts but exhibits a visible red shift from 2077 to 2075 (InNi3C0.5/a-ZrO2) and further to 2047 and 2046 cm−1 (InNi3C0.5/t-ZrO2 and InNi3C0.5/m-ZrO2, respectively) because of an enhanced electron back-donation from InNi3C0.5 to the anti-bonding orbitals of CO (38). These results confirm again the EMSI-enhanced electron density of InNi3C0.5, which is improved according to the oxygen vacancy density that is dependent on ZrO2 type.

CO2 adsorption and activation

Effective adsorption and activation of CO2 on the catalyst is the prerequisite for converting CO2 to methanol. Therefore, CO2-TPD experiments in combination with mass spectrometry (MS) were conducted for the catalysts (InNi3C0.5/m-ZrO2, InNi3C0.5/t-ZrO2, InNi3C0.5/a-ZrO2, and InNi3C0.5/SiO2) as well as for the pure supports (SiO2, m-ZrO2, t-ZrO2, and a-ZrO2) and unsupported InNi3C0.5 as references. The supports desorb little CO2, showing the poor ability to adsorb CO2, while the unsupported InNi3C0.5 desorbs a huge amount of CO2 at 510°C (with trace CO) and 637°C (concomitantly with abundant CO; Fig. 3D), respectively assigned to the nondissociated CO2 adsorption on the 3Ni-In sites and dissociated CO2 adsorption on 3Ni-C sites on the InNi3C0.5(111) plane (23). The InNi3C0.5/SiO2 and InNi3C0.5/a-ZrO2 catalysts also offer two CO2 desorption peaks at 500°C (with trace CO) and 635°C (with comparable CO amount) (Fig. 3E), which are similar to the unsupported InNi3C0.5, indicating the weak EMSI between InNi3C0.5 and SiO2 or a-ZrO2. In contrast, the CO2 desorption peak at 500°C becomes very weak for InNi3C0.5/t-ZrO2 and vanishes for the InNi3C0.5/m-ZrO2 catalysts, while the high-temperature CO2/CO desorption at 650° to 660°C becomes stronger especially for InNi3C0.5/m-ZrO2 with most CO formation from CO2 dissociation (Fig. 3E). Notably, little CO2 is desorbed below 500°C for these four catalysts, indicating the poor ability to adsorb CO2 on supports even after loading InNi3C0.5. Therefore, CO2 should be mainly adsorbed on the InNi3C0.5 surface for these four catalysts. In spite of almost identical total amount of CO2 adsorption on InNi3C0.5, the CO desorption amount is very distinct: InNi3C0.5/m-ZrO2 (111.5 μmol g−1) > InNi3C0.5/t-ZrO2 (50.7 μmol g−1) > InNi3C0.5/a-ZrO2 (39.8 μmol g−1) > InNi3C0.5/SiO2 (30.2 μmol g−1) (table S6), consistent with the ZrO2-type dependence of their EMSI strength. These results explicitly order the CO2 dissociation activity of these four catalysts as follows: InNi3C0.5/m-ZrO2 > InNi3C0.5/t-ZrO2 > InNi3C0.5/a-ZrO2 ~ InNi3C0.5/SiO2. To further investigate the quantitative connection between the CO desorption amount and catalytic performance, the TOFs of these four catalysts are plotted against their CO desorption amount (Fig. 3F), showing a good linear correlation. On the basis of these results, we are confident that the discrepancy of the catalyst activity for the CO2-to-methanol reaction is tightly linked with their different CO2 dissociation activity that is governed by the ZrO2 type–dependent EMSI between InNi3C0.5 and ZrO2 supports.

In-depth understanding of ZrO2 type–dependent EMSI: Density functional theory calculations

To further gain insight into the EMSI between InNi3C0.5 and ZrO2 supports, first-principle calculations were performed. We first established two interfaces between InNi3C0.5 and perfect m-ZrO2 or partially reduced m-ZrO2 (denoted as m-ZrO2-x) (Fig. 4A). Their electron density maps reveal that the electrons are accumulated to some extent at the InNi3C0.5m-ZrO2 interface but more accumulated at InNi3C0.5m-ZrO2-x along with the oxygen vacancy formation on m-ZrO2-x, confirming the EMSI at the interface; electrons also redistribute similarly at InNi3C0.5t-ZrO2 and InNi3C0.5t-ZrO2-x interfaces (Fig. 4B), but with lower electron density than at the InNi3C0.5m-ZrO2 and InNi3C0.5m-ZrO2-x interfaces. These calculations indicate the strongest EMSI between InNi3C0.5 and m-ZrO2-x, coinciding with the conclusions based on the XPS and DRIFTS results. Moreover, the electron structure of surface Ni is also affected, especially d electrons (see the evidences of projected density of states in Fig. 4C), making the CO2 adsorption and concomitant dissociation ability of 3Ni-In close to that of 3Ni-C sites. The difference in the CO2 adsorption energies on 3Ni-In and 3Ni-C is reduced from 0.47 eV on unsupported InNi3C0.5(111) to 0.41, 0.39, 0.37, and 0.24 eV on InNi3C0.5(111) surfaces supported on m-ZrO2(−111), t-ZrO2(011), t-ZrO2-x(011), and m-ZrO2-x(−111), respectively, again indicating the strongest EMSI between InNi3C0.5(111) and m-ZrO2-x(−111). Furthermore, the interfacial adhesion work (Wad) of the four systems was also calculated, and InNi3C0.5(111)/m-ZrO2-x(−111) exhibits the largest Wad of 2.89 J/m2, followed by InNi3C0.5(111)/t-ZrO2-x(011), InNi3C0.5(111)/m-ZrO2(−111), and InNi3C0.5(111)/t-ZrO2(011), further consolidating the strongest EMSI between InNi3C0.5(111) and m-ZrO2-x(−111). This markedly enhanced EMSI between InNi3C0.5 and m-ZrO2 imparts high CO2 dissociation activity of the 3Ni-In sites quite comparable to the 3Ni-C sites, rationally explaining why InNi3C0.5/m-ZrO2 offers a single strong CO2/CO desorption peak in Fig. 3E. Also, the CO2 chemical adsorption configurations are quite similar on the unsupported and ZrO2-supported InNi3C0.5 surfaces (Fig. 4, D and E), and therefore, the difference in CO2 adsorption energy on these systems is attributed to the electronic rather than geometrical effect.

Fig. 4 Density functional theory studies.

Three-dimensional (3D) interfacial configuration and electron density difference map for (A) InNi3C0.5/m-ZrO2 (without oxygen deficiency) and InNi3C0.5/m-ZrO2-x, and for (B) InNi3C0.5/t-ZrO2 (without oxygen deficiency) and InNi3C0.5/t-ZrO2-x. Top row: side view (left) and top view (right) of 3D interfacial unit cell. Middle row: depletion regions, blue; accumulation region, yellow. Bottom row: 2D configuration; cutting plane: the best plane of the chosen three atoms marked in circles in the middle 3D structures. (C) Total and partial density of states (TDOS and PDOS) for InNi3C0.5 and InNi3C0.5/m-ZrO2-x. Chemically adsorbed CO2 on 3Ni-In and 3Ni-C sites of (D) unsupported InNi3C0.5(111) surface and (E) defective m-ZrO2-x(−111)–supported InNi3C0.5(111) surface. The C─O bond (the one parallel to the surface) length is provided.

Tailoring more advanced InNi3C0.5 catalyst

Inspired by above interesting findings, we believed that there is a possibility to build more advanced InNi3C0.5 catalysts via EMSI tailoring by using other oxides to replace m-ZrO2. Given that the EMSI is tightly related to the reducible oxides enriched with oxygen vacancies, some universal transition metal oxides such as ZnO, TiO2, CeO2, and Fe2O3 were used to support InNi3C0.5 nanoparticles (Fig. 5A and fig. S11). In particular, InNi3C0.5/Fe3O4, obtained by carburization treatment of an In2O3-NiO/Fe2O3 precursor (details in Materials and Methods), delivers a superior performance over the other candidates, for example, enabling CO2-to-methanol with 20.0% CO2 conversion and 91.2% CH3OH selectivity at 325°C, GHSV of 30,000 ml gcat−1 hour−1, H2/CO2 molar ratio of 8:1, and 4.0 MPa, and even with 25.7% CO2 conversion and 90.2% CH3OH selectivity at 325°C, GHSV of 36,000 ml gcat−1 hour−1, H2/CO2 molar ratio of 10:1, and 6.0 MPa (tables S4 and S7). Notably, no matter how harsh the reaction conditions became in the present work, methanol selectivity always stayed at 90 to 93% with CH4 selectivity no more than 0.2%. The InNi3C0.5/Fe3O4 catalyst was further examined at 325°C and a fixed GHSV (for CO2) of 10,500 ml gcat−1 hour−1 but varied reaction pressure and H2/CO2 molar ratio; excitingly, when increasing the reaction pressure and H2/CO2 molar ratio from 4.0 MPa and 3:1 up to 6.0 MPa and 10:1, the STY of methanol gradually increased from 1.35 (with 10.0% conversion and 90.0% selectivity) to 2.62 gMeOH gcat−1 hour−1 (with 18.8% conversion and 92.8% selectivity; table S7). We also evaluated our InNi3C0.5/Fe3O4 catalyst under the reported reaction conditions, and obviously, the InNi3C0.5/Fe3O4 catalyst exhibited higher STY of methanol than the reported ones under the identical reaction conditions: for example, 1.01 versus 0.73 gMeOH gcat−1 hour−1 for ZnO-ZrO2 at 320°C, 5.0 MPa, H2/CO2 molar ratio of 3:1, and 24,000 ml gcat−1 hour−1 (20); 1.16 versus 0.86 gMeOH gcat−1 hour−1 for In@Co at 300°C, 5.0 MPa, H2/CO2 molar ratio of 4:1, and 27,500 ml gcat−1 hour−1 (39); 1.30 versus 1.01 gMeOH gcat−1 hour−1 for Pd-In2O3 at 280°C, 5.0 MPa, H2/CO2 molar ratio of 4:1, and 48,000 ml gcat−1 hour−1 (40); 0.308 versus 0.288 gMeOH gcat−1 hour−1 for hexagonal In2O3 at 280°C, 5.0 MPa, H2/CO2 molar ratio of 6:1, and 9000 ml gcat−1 hour−1 (41). All things considered, such InNi3C0.5/Fe3O4 catalyst outperforms all ever-reported promising catalysts (5, 19, 20, 3944) in terms of methanol STY, methanol selectivity (>90%), and intrinsic activity represented by TOF [for example, 133.7 hour−1 based on total number of active sites (3Ni-In and 3Ni-C) or 89.1 hour−1 based on total number of surface Ni atoms for our InNi3C0.5/Fe3O4, higher than 74.2 hour−1 based on total number of surface Cu atoms for Cu-Zn-ZrO2 (44); see detailed comparison in table S4]. As expected, the InNi3C0.5/Fe3O4 catalyst exhibits further improved ability for CO2 dissociative adsorption evidenced by much higher MS signal of CO than CO2 in the CO2-TPD profiles in comparison with the InNi3C0.5/ZrO2 catalysts (Fig. 5, B and C, and table S6), thereby leading to a remarkable increase of the TOF to 133.7 hour−1. This breakthrough is due to the strong EMSI between InNi3C0.5 and Fe3O4, evidenced by lowered binding energy values of In, Ni, and C in InNi3C0.5 (remarkably against unsupported InNi3C0.5 and slightly against InNi3C0.5/m-ZrO2; Fig. 5D and table S5).

Fig. 5 Characterization and catalytic performance of InNi3C0.5/Fe3O4.

(A) XRD patterns of the InNi3C0.5/Fe3O4 catalyst. (B) CO2-TPD-MS profiles for In2O3-NiO/Fe2O3 catalyst precursor and InNi3C0.5/Fe3O4 catalyst. MS signal of the carbonaceous species for CO2 desorption: CO2 signal (m/z = 44) and CO signal (m/z = 28). (C) Plot of TOF as a function of the amount of CO desorption (reaction conditions for TOF calculations: 300°C, 4.0 MPa, H2/CO2 = 3:1, and GHSV = 24,000 ml gcat−1 hour−1). (D) XPS spectra in Ni 2p, In 3d, and C 1s regions of InNi3C0.5/Fe3O4. (E) CO2 conversion and CH3OH/CO/CH4 selectivity along with the time on stream over the InNi3C0.5/Fe3O4 catalyst (temperature of 250° to 350°C, H2/CO2 of 3:1 to 8:1, and GHSV of 12,000 to 42,000 ml gcat−1 hour−1).

Moreover, InNi3C0.5/Fe3O4 offers a much smaller InNi3C0.5 particle size of 7 nm than that (~16 nm) of the InNi3C0.5/ZrO2 catalysts. Coupling this feature with the markedly improved activity for InNi3C0.5/Fe3O4 leads to a big reduction of the loading of InNi3C0.5 to 11.4 weight % (wt %), almost one-fourth of that (42.8 wt %; table S8) of InNi3C0.5/m-ZrO2. Low loading of In together with using cheap Fe3O4 as support makes InNi3C0.5/Fe3O4 more affordable, which is also an important consideration in practical application. Another advantage of our InNi3C0.5 nano-intermetallic catalysts is the promising stability (Fig. 5E and figs. S12 and S13), and especially the InNi3C0.5/Fe3O4 catalyst shows satisfying activity/selectivity maintenance throughout the entire 500-hour testing in a wide range of reaction conditions (Fig. 5E) without any sintering (fig. S12). Notably, no any FeCx species was detected in the InNi3C0.5/Fe3O4 catalyst even after 500-hour testing, according to the 57Fe Mössbauer spectroscopy and Fe 2p XPS spectra (fig. S14 and Supplementary Text). By comparison, most literature catalysts are suffering from rapid deactivation because of the easy carbon deposition and/or catalyst sintering (15, 45, 46), such as the conventional CuZnAl catalyst, which loses more than 50% of its initial activity within 100-hour reaction under the identical reaction conditions (19). Our InNi3C0.5/Fe3O4 catalyst also shows pleasing tolerance to sulfur poisoning even in the presence of 50 ppmv H2S in the feed gas (fig. S15).

We demonstrate an outstanding oxide-supported InNi3C0.5 nano-intermetallic catalyst for efficient methanol synthesis from CO2. First, interesting ZrO2 type–dependent activity of the InNi3C0.5/ZrO2 catalysts is observed, which is tightly linked with the EMSI strength governed by the type of ZrO2 phase. Evidenced by experimental CO2-/H2-TPD, XPS, EPR, CO-DRIFTS spectral studies, and density functional theory (DFT) calculations, InNi3C0.5/m-ZrO2 achieves markedly enhanced EMSI that therefore endues InNi3C0.5 with high electron density, due to the higher oxygen deficiency of m-ZrO2 compared to t-ZrO2 and a-ZrO2. As a result, the InNi3C0.5/m-ZrO2 catalyst shows superior activity for dissociative adsorption of CO2 and subsequent hydrogenation to form methanol over the two others. Inspired by this finding, a more advanced InNi3C0.5/Fe3O4 catalyst, with further enhanced EMSI effect, is developed via carburization of an In2O3-NiO/Fe2O3 precursor. As expected, over this catalyst, the CO2 dissociative adsorption is markedly improved and therefore leads to a remarkable increase of the catalyst activity. This catalyst is also stable, highly resistant to sulfur poisoning, and cost-efficient because of the low In loading and cheap Fe3O4 support used. Our results will stimulate attempts to discover highly active/selective intermetallic catalysts by optimizing EMSI effect through combining theoretical and experimental studies, which might lead to commercial exploitation of an efficient CO2 hydrogenation to methanol process.

MATERIALS AND METHODS

Catalyst preparation

Synthesis of zirconia supports. Three zirconia supports with different phases (monoclinic, tetragonal, and amorphous zirconia, denoted as m-ZrO2, t-ZrO2, and a-ZrO2, respectively) were synthesized according to the following methods. m-ZrO2 was synthesized by a precipitation method: Zr(NO3)4·5H2O (6.968 g) was dissolved in deionized water (100 ml), followed by dropwise addition of a 100-ml aqueous solution of (NH4)2CO3 (3.119 g) in 30 min under vigorous stirring at 70°C to form a precipitate. The suspension was continuously stirred at 70°C for 2 hours, followed by aging at ambient temperature overnight, filtering, and washing several times with deionized water. Subsequently, the as-obtained sample was dried at 110°C for 4 hours and calcined at 500°C in static air for 3 hours to yield the m-ZrO2 support. The t-ZrO2 was synthesized by a combined precipitation and reflux digestion method (47): ZrOCl2·8H2O (16.106 g) was dissolved in deionized water (100 ml), followed by dropwise adding 200-ml NH4OH solution (1 M) under vigorous stirring. The resulting material was heated in the mother liquor at 105°C under reflux for 240 hours, while the pH was maintained at 10, followed by aging, filtering, washing (until to no detectable chlorine anions in the filtrate by AgNO3), drying at 110°C for 4 hours, and calcining at 800°C in static air for 3 hours to yield the t-ZrO2 support. a-ZrO2 was synthesized by a precipitation method assisted with surfactant (30, 48): Pluronic P123 (EO20PO70EO20, 6.960 g) and ZrOCl2·8H2O (13.075 g) were dissolved in deionized water (200 ml) with vigorous stirring at 80°C; subsequently, a NH4OH solution (1 M) was dropwise added to the obtained solution until a pH of 11. The obtained suspension was digested at 100°C for 240 hours, followed by aging at ambient temperature overnight, filtering, washing several times with deionized water, and drying at 110°C for 12 hours. Last, the product was calcined for 4 hours at 450°C in static air to obtain the a-ZrO2 support.

Synthesis of supported InNi3C0.5 catalysts. The InNi3C0.5/m-ZrO2 catalyst was taken as an example to describe the synthesis procedures (fig. S16): First, In(NO3)3·4H2O (0.487 g) and Ni(NO3)2·6H2O (1.140 g) were dissolved in deionized water (1.500 g) under stirring at ambient temperature for 15 min. Then, the as-prepared m-ZrO2 support (0.500 g) was impregnated with the as-obtained aqueous solution, followed by ultrasonication for 2 hours, aging at ambient temperature overnight, drying in air at 100°C for 12 hours, and calcining in static air at 350°C for 2 hours. Then, the resulting In2O3-NiO-ZrO2 catalyst precursor was packed into a continuous-flow fixed-bed tubular reactor made of stainless steel (inner diameter of 8 mm with length of 768 mm) and carburized in a stream of a mixture of H2 and CO2 (30 ml min−1, H2/CO2 molar ratio of 3:1) at 600°C for 3 hours under atmospheric pressure. InNi3C0.5/t-ZrO2, InNi3C0.5/a-ZrO2, InNi3C0.5/SiO2 (for comparison; commercial SiO2, Sinopharm Chemical Reagent Co. Ltd.), and InNi3C0.5/MOx (MOx = ZnO, TiO2, CeO2, and Fe3O4) were synthesized following the same procedures. The catalysts with different InNi3C0.5 loadings were obtained by varying the adding amounts of indium nitrate and nickel nitrate precursors (table S8).

Catalyst characterization

X-ray powder diffraction (XRD) measurements were conducted on a Rigaku Ultima IV diffractometer (Japan), using a Cu Kα radiation source generated at 30 kV and 25 mA in the 2θ angle range of 10° to 60° at a scanning speed of 10° min−1 with a step size of 0.02°. The catalyst micromorphology and nanostructure were observed by a scanning electron microscope (Hitachi S-4800, Japan; accelerating voltage: 3.0 kV) and TEM (FEI-Tecnai G2 F30, USA; accelerating voltage: 200 kV). Nitrogen adsorption-desorption isotherms were taken on a Quantachrome Autosorb-3B instrument (USA) at −196°C.The samples were evacuated at 300°C for at least 6 hours before the measurements. The SSA was calculated from the adsorption branch using standard Brunauer-Emmett-Teller theory. The pore size distribution and total pore volume were determined using the Barrett-Joyner-Halenda method based on the adsorption isotherm. Quasi–in situ XPS analyses were carried out on an AXIS SUPRA system (Shimadzu/Kratos) equipped with an in situ reactor chamber, using a standard Al Kα x-ray source (300 W) with an analyzer pass energy of 40.0 eV. The circular catalyst chips (1.5 mm diameter) were pretreated in H2/CO2 mixture (H2/CO2 molar ratio of 3:1, 30 ml min−1) at 300°C for 2 hours in the reactor chamber and then cooled down to room temperature and transferred into the spectrometer chamber without exposure into air. All binding energies were referenced to the adventitious C1s line at 284.8 eV. EPR was performed on a Bruker EMXPLUS spectrometer at a microwave frequency of 9.83 GHz (X-band) with catalyst sample of 0.038 g. Spectra were collected accumulating 1 scan for field sweeps of 5000 G at −196°C with a microwave power of 0.2 mW. The 57Fe Mössbauer spectra were recorded on a conventional spectrometer (Wissel MS-500, Germany) in transmission geometry with constant acceleration mode. 57Co(Pd) was used as the radioactive source. The spectra were fitted with the appropriate superposition of Lorentzian lines. The real In and Ni contents of the InNi3C0.5 catalysts were quantitatively analyzed by the inductively coupled plasma–atomic emission spectroscopy on Optima 8300 (PerkinElmer, USA).

H2 and CO2 TPD (H2-/CO2-TPD) measurements were performed on a TP 5080 multifunctional automatic adsorption/desorption instrument (Xianquan Industrial and Trading Co. Ltd., P.R. China) with a TCD and an online mass spectrometer (ProLine Dycor, AMETEK Process Instrument, USA). For each trial, the sample (0.1 g) was treated in a H2 flow (30 ml min−1) at 300°C for 1 hour and flushed by a He flow (30 ml min−1) at 300°C for 30 min to clean its surface. After cooling to room temperature in a He flow, the catalyst sample was exposed to a H2 (or CO2) flow for 30 min for saturation adsorption of H2 (or CO2), and afterward, the carrier gas (ultrahighly purified N2 for H2-TPD or He for CO2-TPD) was switched into the reactor at a flow rate of 30 ml min−1 until stable baseline appeared before implementing. The TPD profiles were then recorded from room temperature to 850°C at a heating rate of 10°C min−1.

In situ DRIFTS experiments for CO adsorption on the catalysts were carried out on a Bruker Tensor 27 spectrometer, equipped with a mercury-cadmium-telluride detector and a Harrick Scientific HV-CDRP-4 reaction cell fitted with ZnSe windows. The catalyst sample of 0.020 g was placed into the cell chamber, treated at 400°C for 2 hours in a H2 flow (30 ml min−1), purged with a He flow (30 ml min−1) at 400°C for 1 hour, and cooled down to room temperature in He for taking a reference spectrum. Then, the catalyst was exposed to pure CO flow (10 ml min−1) for 30 min and subsequently purged with He (30 ml min−1) for 30 min, for taking CO-DRIFT spectrum. All spectra were recorded by collecting 32 scans from 4000 to 400 cm−1 at a resolution of 4 cm−1.

Reactivity tests

The CO2-to-methanol reaction was evaluated in a continuous-flow fixed-bed tubular reactor made of stainless steel (inner diameter of 8 mm with length of 768 mm) that was heated by an electronic furnace. Typically, the as-carburized catalyst with granule size between 100 and 125 μm (0.500 ± 0.002 g) was packed into the reactor, and the catalyst bed at the center of the reactor was supported by quartz wool at both ends. The reaction temperature, pressure, GHSV, and H2/CO2 molar ratio were varied in the range of 250° to 350°C, 1.0 to 6.0 MPa, 12,000 to 115,500 ml gcat−1 hour−1, and 3:1 to 10:1, respectively.

The effluent gas was quantitatively analyzed by an online Agilent 7820 gas chromatograph equipped with a thermal conductivity detector (TCD) and a flame ionization detector. The postreactor line was maintained at 150°C to prevent product from condensing. All the reaction data were collected after running for at least 3 hours under steady-state conditions. The gas sample was withdrawn every 30 min, and more than eight measurements were taken for each reaction parameter. The products from this reaction were CH3OH, CO, and CH4. The CO2 conversion (%) and product selectivity (%) were calculated by the standard normalization method based on carbon atom balance according to the following equationsCO2 conversion (%)=(1fCO2ACO2,outfiAi,out+fCO2ACO2,out)×100%i selectivity (%)=(fiAi,outfiAi,out)×100%where Ai,out and fi are the chromatographic peak area at the outlet and the relative molar calibration factor of the individual product i (i: CH3OH, CO, and CH4), respectively.

STY of methanol, expressed as grams of CH3OH per gram catalyst per hour (gMeOH gcat−1 hour−1), was calculated according to the following equation CH3OH STY=FCO2,in×XCO2×SCH3OH×MWCH3OHWcat×Vmwhere FCO2,in is the volumetric flow rate of CO2 (milliliter hour−1), XCO2 is the CO2 conversion, SCH3OH is the CH3OH selectivity, MWCH3OH is the molecular weight of CH3OH (32 g mol−1), Wcat is the overall mass of catalyst (g), and Vm is the ideal molar volume of CO2 at standard temperature and pressure.

Thermodynamic analysis and TOF calculations

The thermodynamic analysis was performed by using the HSC Chemistry 6.0 software, with the results shown in fig. S6. In the analysis, the reaction pressure was increased from 0 to 100 bar (10 MPa) at different temperatures (from 200° to 400°C). The reactants of H2 and CO2 (with the molar ratio of 3:1) and aimed products CH3OH and H2O were considered (CO2(g) + 3H2(g) ↔ CH3OH(g) + H2O(g)). The theoretical equilibrium conversion of CO2 (Xe) was calculated according to the following equation Xe=(AA0A)×100%where A is the initial amount of CO2 (kmol) and A0 is the amount of CO2 (kmol) at thermodynamic equilibrium.

To assess the intrinsic activity of the supported InNi3C0.5 catalysts for the CO2-to-methanol reaction, the TOF was measured (with CO2 conversion below 10.0% at 300°C, 4.0 MPa, and 24,000 ml gcat−1 hour−1; table S3), which was defined as the produced methanol per active site per hour TOF=FCO2Vm×XCO2×SCH3OH×NAWcat×x×Nmumwhere FCO2 is the volumetric flow rate of CO2 (milliliter hour−1), Vm is the ideal molar volume of CO2 at standard temperature and pressure, XCO2 is the CO2 conversion, SCH3OH is the CH3OH selectivity, NA is the Avogadro constant, Wcat is the mass (0.5 g) of the supported InNi3C0.5 catalysts, and x is the mass fraction (42.8%) of InNi3C0.5 in the supported InNi3C0.5 catalysts. Nnum is the number of available surface active sites (i.e., the total number of 3Ni-In and 3Ni-C) per gram InNi3C0.5, which can be calculated according to the following equation (23) Nnum=SA×25.0%A(3Ni-In or 3Ni-C)where SA is the exposed SSA (m2 g−1) of the InNi3C0.5 nanoparticles in the supported InNi3C0.5 catalyst [assuming that all exposed surfaces of the supported InNi3C0.5 nanoparticles were InNi3C0.5 (111) surface], 25.0% is the percentage of the total area of the surface active sites (the 3Ni-In and 3Ni-C sites) in the total surface area of InNi3C0.5 (111) surface, and A(3Ni-In or 3Ni-C) is the area of one 3Ni-In or 3Ni-C active site (one 3Ni-In site has equal area to one 3Ni-C site of 3.013 × 10−20 m2). The SA can be estimated on the basis of their TEM-visualized particle size according to the following equation (3) SA=6ρ×dInNi3C0.5where ρ is the density of bulk InNi3C0.5. The corresponding TEM-visualized particle size distribution of InNi3C0.5 nanoparticles is shown in Fig. 1 (B to D) and table S1.

DFT calculations

We used spin-polarized DFT as implemented in the Vienna Ab initio Package. The self-interaction problem inherent with this functional has been partly removed by applying the DFT + U approach, where the Hubbard’s U parameter for the 4d orbitals of the Zr ions was set to 4 eV (49). Core-valence and electron-electron interactions were treated by the projector augmented wave method (50) and the Perdew-Burke-Ernzerhof generalized gradient approximation (51, 52). The energy cutoff for the planewave basis set was 400 eV. Geometry optimization was considered to be converged when the forces was <0.03 eV/Å. Reciprocal space was sampled only at the Γ-point because of the large supercell. The interface structure of InNi3C0.5(111)/t-ZrO2(011) was constructed using (2 × 4) InNi3C0.5(111) and (3 × 3) t-ZrO2(011) with the lattice mismatch less than 3%, and the interface structure of InNi3C0.5(111)/m-ZrO2(−111) was constructed using (4 × 4) InNi3C0.5(111) and (3 × 3) m-ZrO2(−111) with the lattice mismatch ~4% (Fig. 4, A and B). The bottom Ni at the left corner of InNi3C0.5(111) was used as the reference atom to construct different interfacial structures at top site of O, top site of Zr, and bridge site of O-Zr of ZrO2. Three-layer thickness of InNi3C0.5 was chosen to build the interfacial structures to save the cost of the computation where the interfacial structures of InNi3C0.5(111)/m-ZrO2(−111) contain 360 atoms and the geometry optimization is very time-consuming and presents an experimental weight ratio of about 1:1 with the ZrO2 support. All atoms of InNi3C0.5 and the top ZrO2 unit were allowed to relax. For the defective interfacial structures, the interfacial oxygen atoms were removed yielding an oxygen vacancy of 31.25 and 22%, close to the experimental value of ~30 and ~24% for InNi3C0.5/m-ZrO2-x and InNi3C0.5/t-ZrO2-x, respectively (the oxygen vacancy concentration was calculated according to XPS results; see table S5, fig. S9, and Supplementary Text).

To describe the interfacial binding strength qualitatively, the ideal adhesion work of the interface was defined as follows (53)Wad=EZrO2+EInNi3C0.5EInNi3C0.5/ZrO2A

The first, second, and third terms on the right side of the equation are the total energies of the optimized single ZrO2 surface, single InNi3C0.5, and InNi3C0.5/ZrO2 interface, respectively. A is the interfacial area. The larger the adhesion work, the stronger the interfacial binding of InNi3C0.5 with ZrO2.

Acknowledgments: We appreciate R. Prins (ETH Zurich) for helpful discussions. Funding: This work was supported by a Key Basic Research Project (18JC1412100) from the Shanghai Municipal Science and Technology Commission, the National Natural Science Foundation of China (22072043, 21773069, 21703137, 21703069, and 21473057), and the National Key Basic Research Program (2011CB201403) of the Ministry of Science and Technology of the People’s Republic of China. Author contributions: Y. Lu and G.Z. conceived the idea for the project. C.M. and P.C. conducted material synthesis. C.M. performed structural characterizations and catalytic test. X.-R.S. performed DFT calculations. Y. Lu, G.Z., C.M., X.-R.S., and Y. Liu analyzed the catalytic results. C.M., G.Z., X.-R.S., and Y. Lu drafted the manuscript. Y. Lu directed the research. All authors discussed and commented on the manuscript. Competing interests: Y. Lu, P.C., G.Z., and Y. Liu have a patent application related to this work filed with the Chinese Patent Office on 15 October 2017 (201710956080.1). The authors declare no other competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials.

Be the first to comment

Leave a Reply

Your email address will not be published.


*