Solid catalysts play vital roles as heterogeneous catalysts in numerous applications in the chemical, petrochemical, pharmaceutical, energy, food and automobile industries, facilitating large-scale production and facile product separation. However, conventional heterogeneous catalysts possess three major shortcomings that hinder their wide-spread application, including i) inferior catalytic activity and selectivity caused by a diversity of active sites, ii) tendency towards deactivation caused by agglomeration of catalyst particles, and iii) undesired carbon deposition in transformation of hydrocarbons. 1-4 In this research, we present a new kind of heterogeneous catalyst that consists of a liquid metal with dissolved catalytically active metal atoms, denoted as a heterogeneous droplet catalyst (HDC). Taking galinstan as the liquid metal with Cu as catalytically active metal solute (denoted as Cu-HDC) as an example, we demonstrate that the dissolved Cu atoms promote selective oxidation of methane into primary oxygenates at room temperature with a superb activity of 1940 mmol·gCu-1·h-1 and a selectivity of over 90%, outperforming state-of-the-art industrial catalysts. The liquid metal solvent ensures a high, entropically driven dispersion of the Cu atoms and prevents carbon deposition. The highly promising Cu-HDC catalyst shows a stable operation over 240 hours. By employing in-situ X-ray absorption spectroscopy (XAS), near ambient-pressure X-ray photoelectron spectroscopy (NAP-XPS) and density functional theory (DFT), we show that the narrow d-orbital of the dissolved Cu atoms in HDC facilitates formation of adsorbed methyl radicals *CH3 and impedes their transformation into *CH2 (where the * notation is used to denote an adsorption site) 5,6. These features enhance the selectivity towards primary oxygenates and prevent over-oxidization and formation of C2+ products. As the phase of the HDC is different from those of gas, aqueous and organic solution phases its separation is as facile as that of solid catalysts. Unlike conventional supported catalysts that are designed with well-defined crystal structures, crystal facets and defects, HDCs are characterized by weak interactions between the liquid metal support and the active metal solute, by having mobile surfaces and by dynamic catalytically active centers (Figure 1A). These unprecedented features endow HDCs unique adsorption behavior and adaptive activation of molecules. For this reason, HDCs have the potential to become a major category of catalysts in the field of heterogeneous catalysis.
Galinstan, a gallium-based eutectic alloy is a liquid at room temperature7-10. Different metals can be dissolved into galinstan, e.g., Cu, Ni, Fe, Co, Zn, Pt, Pd, or Au. The major challenge in order to facilitate practical application of HDCs as catalysts is to prevent their aggregation into larger droplets. To reduce the size of the droplets and to increase their specific surface area, we suspended the Cu-HDC droplets in polydimethylsiloxane (PDMS) and solidified the obtained suspension by adding a cross-linking agent and subjecting it to an appropriate thermal treatment (Figure 1B).11,12 The cured PDMS rubber with embedded micro-sized Cu-HDC droplets in it (denoted as [email protected]) can be used (see Supplementary Information: Preparation of [email protected]).
The uniform distribution of Cu in the galinstan alloy was confirmed by energy dispersive X-ray spectroscopy (EDX). The obtained elemental mappings are shown in Figure 2A. The aberration-corrected scanning transmission electron microscope (STEM) images and associated 3D intensity profile demonstrate a disordered arrangement of the Cu atoms (Figure 2B, Figure S1 and Figure 2C). For a better observation, we extracted intensity values from the inverse fast Fourier transform (IFFT) image (Figure 2D) in six different directions. Every profile (Figure 2E) is found to exhibit aperiodicity in both amplitude and frequency, confirming the disordered nature of the Cu atoms. For a verification of the liquid characteristics of Cu-HDC, we employed cryogenic transmission electron microscopy (cryo-TEM). The ring-like selected-area electron diffraction (SAED) pattern recorded at room temperature confirms its amorphous nature (Figure 2F), while the SAED pattern recorded at -196 oC, i.e., below the melting point of galinstan, shows bright spots indicative of a polycrystalline structure. Using X-Ray diffraction (XRD) it was found that galinstan is able to dissolve about 2 wt% of Cu at room temperature (Figure S2). Similar conclusions can be drawn by analyzing oscillations in extended X-ray absorption fine structure (EXAFS) spectra. Figure 2G compares EXAFS spectra of Cu foil, Cu-HDC with 1 wt% and that with 2 wt% Cu, while Figure 2H shows corresponding wavelet transforms for the k3-weighted EXAFS signals (corresponding fitting results are shown in Figure S3 and Table S1, respectively). Fitting of the oscillations indicate that Ga atoms form the first coordination shell of Cu atoms in Cu-HDC, rather than In or Sn (having a higher R-factor). The absence of second-nearest neighbors peaks above 4 Å in the spectrum of Cu-HDC with 1 wt % Cu indicates that agglomeration and crystallization did not occur. In contrast, these peaks emerge in the EXAFS spectrum of Cu-HDC with 2.2 wt% of Cu .
Methane catalytic oxidation experiments using [email protected] catalysts with different Cu contents were carried out in an autoclave with H2O2 as the oxidant at temperatures in the range 25 - 75 oC13-16. The liquid and gas products were analyzed by 1H nuclear magnetic resonance (NMR) and gas chromatography (GC), respectively. Corresponding data are shown in Figure 3A, revealing that generation of primary oxygenates, the most promising oxidation products of methane13,17, is dominant with a selectivity over 93% under all tested conditions. Control catalytic experiments were carried out using [email protected] (without Cu), Cu-HDC (without PDMS), pure PDMS and pure galinstan as catalysts, but none of these showed an activity and selectivity comparable to [email protected] A linear relationship between production rate and Cu content is found as shown in Figure 3C, implying that Cu-based moieties serve as catalytically active sites. As alluded to above, once the Cu content exceeds about 2 wt%, a solid solution or alloy with crystalline CuGa2 appears as was confirmed by XRD and EXAFS. The formed CuGa2 species is less active in methane oxidation, which explains the departure from the linear relationship in Figure 3C at high Cu loadings. The production rates of undesired carbon dioxide and acetic acid are not significantly affected by the Cu content. The latter shows that the formation of these products is not associated with the presence of Cu. Besides dispersing the liquid metal droplets, the PDMS rubber modulates the wettability and accessibility of the catalyst. We experienced that the liquid metal is quite wettable to water and methanol with contact angles of 8.6o and close to 0o, respectively (Figure S4). This unfavorable wettability to water and methanol hinders catalytic sites from contact with gaseous methane and desorption of generated methanol. The use of PDMS modifies the wettability of the catalyst surface to become more aerophilic, which promotes the adsorption of methane and the desorption of methanol, thus enhancing the catalytic activity. It was found that [email protected] shows a durability of at least 240 hours without any decline in catalytic performance. This outstanding performance originates mainly from two aspects: i) the solution characteristics of Cu-HDC prevents aggregation of Cu atoms, and ii) its fluidity avoids carbon deposition, which was confirmed by EDX analysis of used catalysts (Figure S5 and Table S2). We further investigated the catalytic performance of [email protected] at different temperatures and methane partial pressures. Corresponding results are shown in Figure 3D. As we identified Cu atoms as the catalytically active centers, the catalytic activity of [email protected] was normalized to its Cu content. Even at room temperature and 1.5 MPa CH4, a catalytic performance as high as 1940 mmol·gCu-1·h-1 (with a turnover frequency (TOF) of 124.1 mol·molCu-1·h-1) can be achieved. According to the best of our knowledge, this is the highest production rate of primary oxygenates from methane reported to date (cf. Table S3). Figure S6 shows that the reaction is first order in methane. A value of 31.3 ±1.7kJ mol-1 was obtained for the apparent activation energy (Figure S7). This value is lower than reported for other catalysts and underscores the superior catalytic activity exhibited by [email protected]13,18-20. Figure S8 confirms a linear relationship between the amount of oxygenates formed and the time of reaction. Within the duration of the experiment the catalyst does not show any sign of deactivation. Analysis by 13C and 1H nuclear magnetic resonance (NMR) showed that products (CH3OOH and HOCH2OOH) are only observed when 13CH4 is used as reactant, confirming that methane is the only source of carbon in the products (Figure S9) 17.
The remarkable catalytic performance of Cu-HDC prompted us to investigate the underlying mechanism of the catalytic reaction. In-situ XAFS was used to study the local chemical environment of the Cu atoms during reaction. Measurements were performed using a home-made gas-tight reaction chamber equipped with a vitreous carbon window which allowed us to obtain the EXAFS signal (Figure S10). Spectra were collected of Cu-HDC in pristine form, immersed in 30% hydrogen peroxide and under reaction conditions (Figures 3F-3H). Curve fitting of the spectrum obtained for pristine Cu-HDC only reveals the Cu-Ga bond. A peak at ca. 1.4 Å appears in the spectrum of Cu-HDC immersed in 30% hydrogen peroxide. Curve fitting shows that this peak corresponds to Cu atoms bonded to a single O atom at a distance of about 1.9 Å. The EXAFS spectrum recorded under 2.0 MPa CH4 reveals that the Cu-O peak is shifting to 1.5 Å, corresponding to a slight increase of the Cu-O interatomic distance to about 2.0 Å. Exposure to CH4 leads to formation of Cu-O-CH3 and, hence, weakening of the Cu-O bond21. To verify these findings, we performed ambient-pressure X-ray photoelectron spectroscopy (APXPS) (Figure 3I, Figure S11). The C1s spectra recorded under 0.5 mbar CH4 reveal the existence of CHx (285.6 eV) and molecular CH4 (287.0 eV). New peaks appear when adding 0.5 mbar of O2 due to the formation of OCH3 (286.5 eV) and COx (289.8 eV). The C1s peak corresponding to OCH3 is found to dominate in a mixed CH4/O2 atmosphere, suggesting that OCH3 is a stable intermediate in methane oxidation (Figure 3J)22,23. The data further confirms the high selectivity of the Cu-HDC catalyst as the relative intensity of the COx peak is found very small (10%).
First-principle DFT calculations were performed to investigate the high catalytic activity of Cu-HDC droplets towards oxygenates. As expected, the computations disclose a low energy barrier for migration of Cu atoms from the interior of the droplets to their surface, suggesting a high possibility for formation of Cu-centered active sites for the catalytic surface reaction (Figure S12). These calculations were performed assuming the liquid metal to consist of pure gallium (rather than a mixture of gallium, indium and tin). We further calculated the reaction pathways for partial oxidation of CH4 on the surfaces of Cu-HDC, pure gallium and on Cu(111) as shown in Figures S13, S14 and S15, respectively. In these figures, Roman numbering is used to designate sequential stages in the mechanism. The calculations reveal that surface Cu and two adjacent Ga atoms abstract one O atom from H2O2 (II), forming a Cu-O-Ga2 motif on the surface of Cu-HDC (III), acting as the site for CH4 adsorption (Figure S13). A bond distance of 1.88 Å is calculated for the Cu-O bond, which is consistent with the value determined by in-situ XAFS. Upon adsorption of CH4 (III), one of its H atoms is abstracted by the *O atom, whereby the *CH3 methyl radical is formed (IV). This is the decisive step in the overall reaction with an energy barrier (TS1) of 1.07 eV (by comparison with 1.38 eV (Figure S11) and 1.56 eV (Figure S12) calculated for similar reaction steps occurring on pure gallium and Cu(111), respectively). Two reaction steps may follow: (Va) *CH3 reacts with H2O2 to form CH3OH, or (Vb) *CH3 is further dehydrogenated to *CH2. The *CH2 radical is the precursor intermediate of oxidation of carbon to higher oxidation states. The calculated energy barrier for reaction step (Vb) is 1.36 eV (TS2), which is high and implies that oxidation of CH4 on Cu-HDC would not proceed any further. It may be noted that the energy barrier for this reaction occurring on Cu(111) is found to be only 1.03 eV (Figure S12). The obtained results clearly demonstrate that reaction via reaction step (Va) is the more favorable pathway to produce CH3OH (VI) on Cu-HDC. The formed CH3OH may be adsorbed on the Cu atom (VII). Subsequent H-abstraction (VIII) and reaction of the formed intermediate *OCH3 with H2O2 (IX) will generate CH3OOH (X).
To understand the superior activity and selectivity of Cu-HDC over Cu(111) in methane oxidation, a comparative computational analysis of the electronic structures of the TS1 transition states was carried out. In both cases, the *CH3 radical resides on top of the Cu atoms, denoted as Cu-HDC_TS1 (Figure S13) and Cu(111)_TS1 (Figure S15) , respectively. As shown in Figure 4, the orbital interaction between *CH3 and Cu-HDC is substantially stronger than that between *CH3 and Cu(111), which leads to a substantial lower energy barrier for CH4 activation on Cu-HDC by comparison with that on Cu(111). The projected density of states (pDOS) of the Cu 3d orbitals in Cu-HDC (Figure 4A) exhibits as a narrow peak, suggesting a very weak coupling between the Cu 3d orbitals and the surrounding Ga orbitals. In contrast, the 3d orbitals in Cu(111) are split and broadly distributed (Figure 4B). The difference in the Cu 3d orbital distributions accounts for the different orbital interactions between Cu and *CH3 in the respective TS1 transition states. The Cu 3d and *CH3 non-bonding (nb) orbitals are strongly coupled in Cu-HDC_TS1, but hardly interact with each other in Cu(111)_TS1. As a result, a Cu-C bond is formed when *CH3 is adsorbed onto CuGa125, while *CH3 almost retains its radical character when adsorbed onto the Cu(111) surface. These results are strengthened by the projected crystal orbital Hamilton population (pCOHP) calculations (Table S4), which reveal a Cu-C bond strength of +0.24 |e| for adsorption of *CH3 onto Cu-HDC and +0.03 |e| in the case of adsorption onto Cu(111). Moreover, the differential charge density profiles indicate that there is an appreciably higher charge density between the Cu and C atoms in Cu-HDC_TS1 (Figure 4C) than in Cu(111)_TS1 (Figure 4D). The *CH3 radical resides directly atop of the Cu atom on the Cu-HDC surface, with a binding energy of -1.21 eV (Figure 4E), while it moves to a hollow site on the Cu(111) surface, with a much larger binding energy of -1.81 eV (Figure 4F). Such a big difference in the binding strength can be explained in two alternative ways. One, *CH3 forms three Cu-C bonds of length 2.23 Å when it adsorbs at the hollow site on Cu(111), while there is only one Cu-C bond of length 2.02 Å when it adsorbs atop a Cu atom on the Cu-HDC surface. Albeit the individual Cu-C bond is weaker in CH3-Cu(111) than in CH3-Cu-HDC, two more Cu-C bonds are formed in the former complex resulting in an overall stronger binding strength. Another explanation is that the d-π orbital hybridization between Cu and C atoms contributing to the Cu-C bonding is appreciably stabilized in the Cu(111)_*CH3 complex, while it is barely changed in Cu-HDC_*CH3, as shown in Figure 4A and 4B, respectively. The much stronger binding of *CH3 on the Cu(111) surface is further confirmed by the pCOHP analysis (Table S4), showing a larger d-p population (integral of -pCOHP) on the Cu-C bond (+0.73 |e| for Cu(111)_*CH3 vs. +0.62 |e| for Cu-HDC_*CH3). On the basis of above analysis it is concluded that, because of the narrow 3d orbital distribution, activation of CH4 is more favorable on Cu-HDC, while the occurrence of over-oxidation and formation of C2+ products is less likely than on the Cu(111) surface. Hence, Cu-HDC outperforms Cu(111) in terms of both activity and selectivity of methane oxidation.
In conclusion, in this research we have demonstrated that a liquid metal with a catalytically active metal solute can act as a highly efficient and durable heterogeneous catalyst. By employing a liquid metal with dissolved Cu, we achieved a state-of-the-art catalytic performance in the selective oxidation of methane, with an activity of 1940 mmol·gCu-1·h-1 and a selectivity over 90% at room temperature. In-situ XAFS and NAP-
XPS experiments and DFT calculations reveal that the active Cu atoms present at the surface of the liquid metal droplet facilitates adsorption of methane, generates methyl radicals and prevents over-oxidation and formation of C2+ products. We believe that our findings offer new opportunities and exciting challenges for discovery of novel heterogeneous catalysts.