Introduction

The wide geological distribution of natural gas resources leads to an undesirable loss of methane (CH4) especially at remote locations via flaring or direct emission into the atmosphere1,2. One possible strategy to mitigate such an issue is to convert CH4 into liquid chemicals at the source of emission under ambient condition with minimal reliance on an industrial infrastructure2. Fundamentally, this catalytic conversion requires low activation energy and high reactivity, in order to accommodate the low thermal energy and partial pressure of CH4 substrate at ambient condition. Existing approaches of CH4 functionalization usually operated at high pressure and/or elevated temperature3,4,5,6,7,8,9,10,11,12,13,14,15,16, involve metal-catalyzed reactions3,4,5,6,7, superacid-based activation17, or catalysis based on peroxo species for a free-radical chain mechanism8,9,10. For the metal-catalyzed reactions reported by Periana and others (Fig. 1a)3,4,5,6, electrophilic activation of CH4 is followed by an oxidation process that regenerates the metal active sites, mostly precious metals including Pt and Pd. Yet, in ambient conditions, the reactivities of these electrophilic metal species seem insufficient to activate CH4. In the approaches based on radical chain mechanism (Fig. 1b)8,9,10, initiators including peroxo species yield oxygen radicals that activate CH4 at room temperature with low activation energies18. But it is uneasy to achieve a sustained, selective catalytic process that generates and replenishes radical species. The challenge of balancing the reactivity and regeneration of active species call for an alternative approach for ambient CH4 functionalization.

Fig. 1: Pathways to CH4 functionalization.
figure 1

a–c Representative approaches of CH4 functionalization based on electrophilic activation3,4,5,6 (a) and radical chain mechanism8,9,10 (b), in comparison with the proposed electrocatalytic method (c). d The proposed catalytic cycle of electrocatalytic CH4 activation with d0 vanadium (V)-oxo dimer (1). Mred and Mox, reduced and oxidized metal active sites, respectively; [O], chemical or electrochemical oxidation; Ea, activation energy; d0–M, d0 early transition metal species; F.E. Faradaic efficiency, TON turnover number, TLS turnover-limiting step.

Full size image

We propose that a controlled electrochemical generation of oxygen radicals is capable to address the above-mentioned challenge, as electrochemical redox process provides a sustained method of replenishing radical species at ambient conditions without sacrificing their high reactivities19,20. While d0 early transition metal centers are not known to be directly oxidizable, we accidentally found that electrochemical oxidation of d0 early transition metal-oxo species in 68–98% H2SO4 yields cation radicals on the sulfonic ligand that selectively activate CH4 at ambient pressure and room temperature (Fig. 1c). Here, we report d0 vanadium (V)-oxo dimer (V2V,V, 1)21 as the model catalyst for mechanistic understanding (Fig. 1d). As the tentative turnover-limiting step (TLS) uncommon for homogenous electrocatalysis, an electrochemical one-electron removal of 1 yields a cation radical (V2V,V •+, 2) reactive towards CH4, while the catalytic cycle is fulfilled by additional electrochemical oxidation and cation radical regeneration.

Results

Discovery of ambient CH4 activation with vanadium (V)-oxo catalyst

The d0 vanadium (V)-oxo catalyst (1) was prepared by dissolving V2O5 in 98% H2SO4. Cyclic voltammograms of 10 mM 1 in 98% H2SO4 under 1-bar nitrogen gas (N2) (blue), 1-bar CH4 (red), and a blank control (black) are displayed in Fig. 2a at 25 ºC on a platinum (Pt) working electrode. A quasi-reversible peak corresponding to VV/VIV redox couple was observed with a midpoint potential E1/2 = 0.644 V vs. Hg2SO4/Hg reference electrode, with a diffusion coefficient D = 2.18 × 10−11 m2 s−1 for 1 based on Randles–Savcik analysis (Supplementary Fig. 1a, b)22. Additional oxidation current of 1 was observed at the electrochemical potential E > 1.6 V vs. Hg2SO4/Hg, and such an oxidation current is larger in CH4 than in N2. This observation suggests that 1 can be further oxidized electrochemically and CH4 is likely to react with the resultant oxidized species. Bulk electrolysis in 98% H2SO4 under 1-bar CH4 was conducted at a E = 2.255 V vs. Hg2SO4/Hg for 6 h with an electrode of fluorine-doped tin oxide (Supplementary Fig. 1c). The liquid composition after electrolysis were analyzed by 1H and 13C nuclear magnetic resonance (NMR) spectroscopy. CH3OSO3H, which yields methanol after hydrolysis11, was observed at chemical shift δ = 3.34 ppm in 1H NMR after electrolysis with 10 mM 1 under CH4 (red in Fig. 2b). No gaseous or liquid products other than CH3OSO3H were observed as a product of CH4 oxidation within our detection limit via NMR spectroscopy (Fig. 2b), mass spectroscopy (Supplementary Fig. 2a, b) and gas chromatography (Supplementary Fig. 3). Organic products were not detected in the absence of 1 under CH4 (black), with 10 mM 1 under N2 (blue in Fig. 2b), or at a less anodic potential (E = 1.855 V vs. Hg2SO4/Hg) (Supplementary Fig. 2d). These data confirm that CH4 undergoes a two-electron oxidation into CH3OSO3H, initiated by the electrochemical oxidation of 1. The absence of a well-defined redox wave preceding the current onset in Fig. 2a suggests either a homogenous electrocatalysis limited by the rate of electron transfer or the occurrence of materials deposition during the scans of cyclic voltammetry. The measurement of X-ray photoelectron spectra on a FTO electrode after 6-h electrolysis with 10 mM 1 detected no residual V signal on the electrode (Supplementary Fig. 4) despite the observed CH3OSO3H formation (1-bar CH4, E = 2.255 V vs. Hg2SO4/Hg). Reusing this vanadium-exposed FTO electrode for 6-h electrolysis in neat 98% H2SO4 under CH4 yielded no CH4 activation. These results suggest a likely homogenous electrocatalysis limited by charge transfer. We also conducted isotope-labeling experiments by introducing 13CH4 as the substrate at 1-bar pressure. The introduction of 13CH4 in lieu of CH4 of natural abundance leads to the surge of 13CH3OSO3H signal at δ = 58.6 ppm in 13C NMR after electrolysis (Fig. 2c). An introduction of a 50% 13C-enriched CH4 yields not only the same peak in 13C NMR but also a triplet in 1H NMR with 13C–1H doublet and 12C–1H singlet at a 1:1 ratio of integration area (Supplementary Fig. 5). The optical absorption spectra of the solution before and after electrolysis were identical to each other (Supplementary Fig. 2e), implying that 1 as a catalyst was regenerated after electrolysis.

Fig. 2: Electrochemical functionalization of CH4 and natural gas mixture.
figure 2

a, b Cyclic voltammograms (a) and 1H NMR spectra of liquid samples after 6-h electrolysis (b) for 10 mM 1 in 1-bar CH4 (red), 10 mM 1 in 1-bar N2 (blue), and blanks without 1 (black). Dashed blue, current density (j) of blue trace magnified by a factor of 10. *, internal standard acetic acid. c 13C NMR spectra of samples before (black) and after electrolysis with 13CH4 (red) and CH4 of natural abundance (blue), respectively. d Calculated TONs (red) and electric charges passed (blue) versus the duration of electrolysis. e F.E. of CH4 functionalization in 10 mM 1 vs. electrode potential E under 1-bar N2 (green), 1-bar CH4 (blue) and 3-bar CH4 (red). f Cumulative TONs for C1 (red), C2 (green, multiplied by a factor of 5), and C3 products (black, multiplied by a factor of 50) as well as F.E. values of total liquid products (blue) are plotted against the duration of bulk electrolysis. TON values for CH3OSO3H (red), CH3COOH plus C2H5OSO3H (green), and CH3COCH3 (black) within 240 h are shown on the right, respectively. Natural gas mixture supplied by SoCalGas was used as the substrate at ambient pressure. Trace products beyond C3 were also observed. 100 mV s−1 and Pt working electrode in (a); FTO working electrode in (b) to (f), and E = 2.255 V vs. Hg2SO4/Hg in (b) to (d), and (f).

Full size image

Electrochemical CH4 and natural gas functionalization

The electrocatalysis with 1 is durable and selective for functionalization of CH4 with high turnover numbers (TONs) and turnover frequencies (TOFs). Bulk electrolysis was conducted with 0.7 mM 1 at 25 ºC under 1-bar pressure of CH4. Liquid aliquots at different time points were analyzed, and the electrochemical TONs were calculated based on the existing method19,23. Figure 2d displayed the amount of electric charge and the calculated TONs as a function of electrolysis duration. A linear correlation suggests a durable catalyst of TON up to 45,000 in 72 h without signs of catalyst degradation. We also investigated the Faradaic efficiency (F.E.), defined as the selectivity of converting CH4 into CH3OSO3H based on the amount of electric charge, as function of E at 25 ºC. In 10 mM 1 (Fig. 2e), the absence of CH4 leads to no product formation (green), and under 1-bar CH4 an optimal F.E. = 63.5% when E = 2.255 V vs. Hg2SO4/Hg (blue). We found that the reaction selectivity is limited by the mass transport and a limited solubility of CH4 in solvent (~1 mM)11. When CH4 pressure increased to 3 bar, the optimal F.E. = 84.5% at E = 2.205 V vs. Hg2SO4/Hg (red in Fig. 2e). The corresponding TOFs of 1 as an electrocatalyst are 483 and 1336 h−1 for CH4 at 1 and 3-bar pressures, respectively, which are conservative and underestimated given the nature of our analysis (Supplementary Note 1). The measured TOF values at room temperature compare quite favorably with other reported catalysts at elevated temperatures and high pressures (Supplementary Table 1).

We expanded the substrate scope from CH4 to ethane (C2H6), propane (C3H8), and ultimately natural gas mixture at 1-bar pressure. C2H6 was oxidized to a mixture of acetic acid (CH3COOH) and ethyl bisulfate (C2H5OSO3H) (Supplementary Fig. 6a), whose TOF values are 297 and 235 h−1, respectively. C3H8 was converted to mostly isopropyl bisulfate (i-C3H7OSO3H) with trace acetone (CH3COCH3) in a 6-h electrolysis (Supplementary Fig. 6b), with TOF values of 962 and 2 h−1, respectively. One challenge in designing a process of natural gas utilization is to balance the low reactivity of the major component CH4 with the high reactivity of minor light alkane components24, as in some cases the latter substrates can react 100-time faster than CH425. The similar TOF values among CH4, C2H6, and C3H8 reported here renders 1 a suitable candidate of direct natural gas utilization at ambient conditions without much upstream separation.

Natural gas supplied to UCLA via pipeline by SoCalGas26 was used as the substrate of electrolysis (E = 2.255 V vs. Hg2SO4/Hg) with 0.7 mM 1 in 98% H2SO4 at room temperature and ambient pressure. Powered by electricity, natural gas was oxidized into organic chemicals while dioxygen in air was reduced on the counter electrode, fulfilling a partial oxidation of natural gas with air with a net reaction of CH4 + 1/2O2 + H2SO4 → CH3OSO3H + H2O. The yielded products mainly consist of CH3OSO3H from CH4, CH3COOH, and C2H5OSO3H from C2H6, and CH3COCH3 from C3H8 in the natural gas. The TONs of C1–C3 products reached about 107,000, 9300, and 200 within 240 h, respectively (Fig. 2f) with the final concentration of CH3OSO3H approaching 10 mM (Supplementary Fig. 7). The observed one-order-of-magnitude difference between 1 and CH3OSO3H product again supports the catalytic feature of our observation. The total F.E. of all liquid products remain stable at around 90% during electrolysis after an initial induction period (Fig. 2f, Supplementary Table 2). The linear relationship between TONs and electrolysis durations suggest that 1 remains active and is tolerant to the impurities in natural gas mixture. Previous analysis suggests that a lower H2SO4 concentration in the electrolyte, ideally below 80%, is needed for industrial implementation11. This requirement is against the thermodynamic limit of reactions in H2SO4 with SO3 as the oxidant11. Yet we found that 1 remains active towards CH4 functionalization in aqueous solution with H2SO4 concentrations as low as 68% under 25 ºC and 1-bar CH4 (Supplementary Fig. 6c), yielding a mixture of methanol and CH3OSO3H (Supplementary Fig. 6d). The robustness under prolonged operation and the applicability in diluted H2SO4-H2O mixed solvent renders the catalyst potentially suitable to be employed to functionalize natural gas with minimal maintenance.

Scale-up potential over vanadium (V)-oxo electrocatalyst

The reported catalyst 1 is also capable to yield high product concentrations amenable for practical implementations. Due to the limited solubility of CH4 in the solvent within current batch reactor11, electrocatalytic experiments of higher product concentrations were conducted under room temperature at 11-bar CH4 pressure, in order to mitigate the mass transport issue of limited gas solubility (vide supra) (Supplementary Fig. 7). A 72-h electrolysis leads to a CH3OSO3H concentration of ~110 mM with F.E. = 81.2% (E = 2.376 V vs. Hg2SO4/Hg). Adding 1 M CH3OSO3H prior to the electrocatalysis under the same condition does not hinder the catalysis or decompose the pre-added CH3OSO3H. A single-pass conversion of 1% was observed in the mixed-flow electrochemical reactor, comparable to the results that lead to electrochemical reduction of CO2 and CO at near-industrial level27,28. The product concentrations reported here are higher than the ones in other electrocatalysis5,6, and suggest that high product concentration of electrocatalysis is attenable. There seems no observable fundamental limit for product concentrations exceeding one mole per liter, a threshold considered suitable for industrial applications11.

Kinetics investigation of electrochemical CH4 functionalization

The attractive feature of catalyst 1 led us to investigate the underlying mechanism during electrolysis with CH4 as the substrate. The current density corresponding CH3OSO3H formation (jCH4), a surrogate of CH4-activating rate, was investigated as a function of catalyst concentration [1] (Fig. 3a), the electrode potential E (Fig. 3b), the partial pressure of CH4 (pCH4) (Fig. 3c), and the temperature T (Fig. 3d). A linear relationship with a slope = 1.03 ± 0.08 between log10(jCH4) and log10([1]) suggests that CH4 activation is first-order on 1 (Fig. 3a). When log10(jCH4) was plotted against E (Fig. 3b), a Tafel slope of about 120 mV dec−1 was observed before jCH4 plateaus at larger E values as CH4 is depleted near electrode. This suggests that the first electron removal from 1 is the TLS, uncommon for homogeneous electrocatalysis (Supplementary Note 2). The overlapping points under 1 and 3-bar CH4 pressure when E < 2.1 V vs. Hg2SO4/Hg suggest that CH4 is not involved in the TLS or any pre-equilibrium steps. When E > 2.1 V vs. Hg2SO4/Hg, a linear relationship between log10(jCH4) and log10(pCH4) with a slope of 0.91 ± 0.07 (Fig. 3c) suggests that CH4 is activated in a first order after the TLS. We also found that the residual current density, the difference between total current density (jtotal) and jCH4, is independent of pCH4 (Supplementary Fig. 2c). It further corroborates that no gaseous or liquid products other than CH3OSO3H were generated from CH4 oxidation, and solvent oxidation into O2 and possibly trace persulfate29,30 is the only plausible side reaction (vide infra). The Arrhenius plot between ln(jCH4) and 1/T yields an apparent activation energy (Ea) as low as 10.8 ± 0.6 kcal mol−1 (Fig. 3d), consistent with the observed ambient reactivity. When conducting electrolysis of 1 in 98% D2SO4 with CH4 of natural isotope abundance, 2D and 1H NMR spectra showed no H/D exchange in the methyl group of product CH3OSO3H (Supplementary Fig. 6e). This excludes the possible mechanism induced by an electrochemically generated superacid17, which should yield significant H/D exchange in the methyl group11. In addition, when CH4 was exposed to 1 in 98% H2SO4 with added K2S2O8 or H2O2 in the absence of electricity, CH4 functionalization was not observed at ambient conditions (Supplementary Fig. 6f). This illustrates that it is difficult for chemical method to sustainably generate reactive radical intermediates at room temperature, which necessitates our use of electrochemistry as proposed before. It also shows that the possible formation of peroxo species including peroxoacids is not contributing to the observed reactivity. When the electrolyte was switched from 98% H2SO4 to oleum with 20% SO3, a 5.4:1 molar ratio between CH3SO3H and CH3OSO3H was observed after electrocatalysis (Supplementary Fig. 8). This indicates the formation of CH3• radical during the catalysis, which yields CH3SO3H in the presence of SO39. Overall, our experimental data support an electrochemical catalysis of low activation energy. After a turnover-limiting one-electron oxidation of 1, the oxidized species undergoes a first-order C–H activation in CH4 and a formation of CH3• radical (Fig. 1d).

Fig. 3: Electrocatalytic kinetics.
figure 3

a The logarithmic of partial current density for CH4 functionalization, log(jCH4), versus the logarithmic of 1’s concentration, log([1]). b Log(jCH4) vs. E under CH4 pressures of 1 bar (blue) and 3 bar (red) with the fitted Tafel slopes displayed. c Log(jCH4) vs. the logarithmic of CH4 pressure, log(pCH4). d The natural logarithmic of partial current density for CH4 functionalization, ln(jCH4), vs. inverse of temperature, T−1, at 1.955 V (black), 2.005 V (blue), and 2.055 V (red) vs. Hg2SO4/Hg, respectively. The corresponding apparent activation energies (Ea) are displayed. Unless noted specifically, 25 °C, 10 mM 1 in 98% H2SO4, E = 2.255 V vs. Hg2SO4/Hg, pCH4 = 1 bar, data recorded from 6-h bulk electrolysis.

Full size image

Unveiling dimer structure of vanadium (V)-oxo catalyst

Despite its ease of preparation, the structural information of 1 is not well understood. It was hypothesized to be a V2V,V dimer with two terminal VV≡O moieties connected by a bridging oxo21. We measured 1’s optical absorption (Supplementary Fig. 9a) and the 51V NMR spectrum (Supplementary Fig. 9b), which confirmed that 1 is different from monometallic VO2+ species in an aqueous medium Density functional theory (DFT) calculations suggest that 1 may exist as two isomers, 1a and 1b (Fig. 4a), with a calculated energy difference of 1.2 kcal mol−1. In an attempt to obtain the correct structure of 1 and real conformation in solution, X-ray absorption spectroscopy of V atom was conducted for 10 mM 1 in 98% H2SO4, solid V2O5, and metallic V foil (solid red, dashed blue, and dashed yellow in Fig. 4b, c, respectively). We carried out a least-square-regression analysis on X-ray absorption near-edge structure (XANES) for the threshold positions, the first peak in the derivative spectra, of VO, V2O3, VO2, and V2O5 to determine the electronic structure and oxidation state of vanadium in 1 (Supplementary Fig. 9c)31. The electronic structure of vanadium in 1 remains similar to that of vanadium in V2O5, confirming the d0 electronic structure of vanadium. The extended X-ray absorption fine structure (EXAFS) can offer coordination information of absorbing atoms by extracting the structural parameters. As shown in Fig. 4c, the absence of noticeable peaks in the region beyond 4 Å (solid red), compared with those of V2O5 and V foil (dashed blue and dashed yellow, respectively), indicates that 1 is a complex homogenously dispersed in the solvent. The peak around 1.56 Å in 1’s EXAFS spectrum (gray area) is attributed to the V–O bonds, following the assignment of V–O bonds in the V2O5 sample. While this comparison provides some information, the general low symmetries of the vanadium-based species prevent us from gaining detailed structural information of 1 solely based on EXAFS data32. To this end, we conducted the fitting of 1’s V K-edge EXAFS spectrum combining the structure suggested by DFT calculations (shown in Fig. 4d). It reveals that the central V atoms are penta-coordinated by O atoms with three types of V–O bond lengths (1.58, 1.68, and 1.96 Å) in the first coordinated shell, with a bridging oxo with a V–O bond length of 1.68 Å. We note that the EXAFS of 1 suggests a unique coordination environment 2.0–3.5 Å away from V atom (blue area), which is different from the monometallic VO2+ species in aqueous medium (Supplementary Fig. 9d). The fitting results of second shell (blue area) indicate that consistent with the predicted structure 1a, there are not only three S atoms in the second shell (2.73 and 3.13 Å) but also one V atom at the distance of 3.27 Å away from the central V atom (Supplementary Table 3). Detailed analysis is provided in the Supplementary information (Supplementary Fig. 9e, f). These results reveal the existence of a hypothesized structure of µ-oxo bridged V2V,V dimer33, and suggest that 1a is the structure of 1 in the solution.

Fig. 4: Structural information of the catalyst and a proposed mechanism.
figure 4

a Possible isomeric structures of 1 and their relative energetics based on DFT calculations. b, c Normalized intensity of V K-edge X-ray absorption near-edge structure (XANES) (b) and extended X-ray absorption fine structure (EXAFS) (c) for 1 (solid red), V2O5 (dashed blue), and metallic V (dashed yellow). d Calculated coordination number (C.N.) and the distance (R) away from V atom based on EXAFS results. e Calculated frontier orbitals involved in the TLS and the proposed transition state of C–H activation step. HOMO highest occupied molecular orbital, LUMO lowest unoccupied molecular orbital. *Designated when considering spin–orbital coupling, equivalent to singly occupied molecular orbital (SOMO) in restricted formalism.

Full size image

Operando characterizations for mechanistic study

Additional operando characterizations were conducted to confirm a homogenous electrocatalysis and elucidate identities of immediate species. Operando Raman spectroscopy measures the vibrational spectra of chemical species near the FTO electrode at different values of E in CH4. No spectral changes were observed and the vibrational spectra evidently differs from solid V2O5 sample (Supplementary Fig. 10). This suggests that there is no detectable heterogeneous intermediate deposited on the FTO electrode during electrocatalysis. Operando XAS spectra of V atom was also measured at different values of E in CH4. Largely, the XANES and EXAFS spectra (Supplementary Figs. 11 and 12) differ from the ones of solid V2O5 sample and supports a homogenous electrocatalysis. Yet some information of reaction intermediates is available. The formal oxidation state of V species, indicative in the pre-edge region of XANES spectra (Supplementary Fig. 11), decreases from +5 to near +4 with increasing value of E, contrary to the typical trend observed in heterogeneous catalysts of electrochemical oxidation34. This reveals the presence of mixed-valence V2IV,V during catalysis (Supplementary Note 3). It also supports a homogenous, diffusion-controlled catalysis, since a hypothetical immobile V(IV) species deposited on the electrode may not have long enough lifetime to be detectable (Supplementary Note 3), given the large thermodynamic driving force of oxidizing V(IV) (>1.2 V from Fig. 2a). The pre-edge region also witnesses a broadening and intensity decrease of the pre-edge peak concurrent with the increase of E and the observation of electrocatalytic CH4 functionalization (Supplementary Fig. 11). This suggests an increase of coordination symmetry near V atom and possibly a loss of sulfonic ligand31. The operando EXAFS results (Supplementary Fig. 12) also displays a decrease of average coordination number of sulfonic ligands per V atom concurrent with increasing E values. Those results imply that the bisulfate group in CH3OSO3H likely originate from the vanadium catalyst, a plausible radical rebound mechanism35.

Proposed electrocatalytic cycle with vanadium (V)-oxo dimer

Combining experimental and computational results, we established a proposed catalytic cycle of 1 for CH4 functionalization (Fig. 1d) despite its uncommon assignment of TLS that warrants additional investigation (Supplementary Note 2). A turnover-limiting electrochemical oxidation of 1a removes one electron from O 2p orbitals in the sulfonic ligand, which is calculated as the highest occupied molecular orbital (HOMO) of 1a (Fig. 4e). The resultant cation radical 2 is predicted to possess an empty frontier spin-orbital on the same O 2p orbitals (lowest unoccupied molecular orbital (LUMO) of 2 in Fig. 4e), which is postulated to initiate H-atom abstraction from CH4. DFT calculations predict reaction trajectory between 2 and CH4 without significant energy barrier (Fig. 4e, Supplementary Fig. 13a). This is consistent with our experimental observation that the TLS is the one-electron oxidation of 1 other than the step of C–H activation (Supplementary Note 2). The calculated barrier-less C–H activation also helps explain the similar TOFs toward various light alkanes in the natural gas24. Currently, we were unable to experimentally characterize 2 and the H-atom abstraction step due to its transient nature, which will be of our research focus in the future. Yet the subsequent steps of CH4 functionalization seems to proceed with the formation of CH3• and a radical rebound process35. This leads to a two-electron oxidation and CH3OSO3H formation with a ligand loss on a V2IV,V dimer (3), which will be readily re-oxidized electrochemically to regenerate 1 (Fig. 1d, Supplementary Fig. 14).

Exploration of molecular and material variants for vanadium (V)-oxo catalyst

DFT calculations of the atomic charges36 suggest that cation radical 2 is stabilized thanks to orbital delocalization, in comparison to the scenario when one electron was removed from sulfuric acid (Supplementary Fig. 13b, c). This is consistent with the results that the calculated ionization energies of 1a is lower by 12–14 kcal mol−1 than that of sulfuric acid. This implies that the metal-oxo centers as carrier of sulfonic ligands stabilize the electrochemically generated cation radical, at the same time maintain a cation radical reactive enough for ambient CH4 functionalization. Other d0 early transition-metal-oxo species can possess similar reactivities. We found that d0 metal-oxo species, including TiIV-oxo and CrVI-oxo, are also electrochemically active towards functionalizing CH4 at ambient conditions (Supplementary Fig. 15). A more extensive survey over the first half of the Period 4 elements except Sc indicates that only Ti, V, Cr, and possibly Mn display similar reactivities (Supplementary Fig. 16). It suggests the strategy of employing d0 early transition-metal-oxo species is generally applicable for ambient electrochemical functionalization of natural gas. As such a general trend of reactivity was not observed before, we posit our electrochemical approach may offer new perspective towards the challenge of CH4 functionalization.

Practically, a heterogeneous catalyst variant may also be desirable. While 1 is characterized as a homogenous catalyst, we found a two-dimensional layered material, VOPO4∙2H2O (4) with exposed V-oxo edges37 (Supplementary Fig. 17a, b), acts as a heterogeneous variant of 1 in 98% H2SO4 (Supplementary Fig. 17c, d). This preliminary result suggests that even within the same metal-oxo system, the catalyst subsequently its reactivity can be tuned with additional materials design and engineering. This heterogenous variant also simplifes product separation in downstream process, thanks to the absence of vanadium in the liquid phase.

Discussion

Overall, the general tunability of catalyst composition may herald better catalysts with higher TOF, lower oxidation potential, as well as versatile design of the overall process. The ambient condition of reported catalysis facilitates the use of O2 in ambient air as the terminal electron acceptor, as well as the use of ambient natural gas feedstock for onsite functionalization without a centralized facility. Future research will focus on possible scale-up with the exploration of optimal reaction conditions. The employment of 98% H2SO4 or more diluted acids, other than oleum, mitigates the generation of excessive acid in product separation and is attractive for practical application38. Additional fundamental and engineering investigation, including the employment of gas diffusion electrode39 as well as ingenious design of electrochemical reactors28, will further explore the possible application of converting CH4 into commodity chemicals with minimal infrastructure support at remote locations. This will lead to the more efficient usage of green-house gases and reducing their emission into atmosphere.

Methods

Chemicals and materials

All chemicals were purchased from Sigma-Aldrich, Thermo Fisher Chemical, or VWR International, unless otherwise stated. All chemicals were used as received unless specified. Dimethyl sulfoxide-d6 (DMSO-d6) was purchased from Cambridge Isotope Laboratories, Inc. All deionized (DI) water was obtained from a Millipore Milli-Q Water Purification System. Fluorine-doped Tin Oxide (FTO) glass was purchased from Hartford Glass Incorporation. CH4 (99.5%) was purchased from Airgas, C2H6 (99%), C3H8 (98%), and 13CH4 (99%; 99 atom% 13C) were purchased from the Sigma-Aldrich. Natural gas mixture was obtained from the outlet in Molecular Science Building 4211, Department of Chemistry and Biochemistry, UCLA, which is supplied via pipeline by SoCalGas. SRI multiple gas analyzer #5 gas chromatograph (GC), 8610C is used to analyze the natural gas mixture. The components are 91.78 mole% CH4, 4.31 mole% C2H6, 0.31 mole% C3H8, 0.04 mole% n-C4H10, 0.03 mole% i-C4H10, 0.01 mole% n-C5H12, 0.01 mole% i-C5H12, and 0.81 mole% CO2. Unless specifically noted, reagent-grade 98% H2SO4 (VWR BDH3074-3.8LP) was employed as the solvent, which contains 5 ppm of metal impurities. When needed, we also employ high-purity 98% H2SO4 (Sigma-Aldrich 339741), which contains 0.3 ppm of metal impurities as shown in the product certificate.

Chemical and material characterizations

Ultraviolet–visible (UV–vis) spectra was conducted on Hewlett-Packard 8453 UV–vis spectrophotometer. Proton NMR (1H-NMR) and carbon NMR (13C-NMR) were recorded on a Bruker AV400 (400 MHz) spectrometer. Deuterium NMR (2D-NMR) was recorded on a Bruker AV500 (500 MHz) spectrometer. Vanadium NMR (51V-NMR) was performed on an Agilent DD2 600 (600 MHz) spectrometer. Powder X-ray diffraction (XRD) patterns were measured on a Panalytical X’Pert Pro X-ray Powder Diffractometer with a Cu Kα source (λ = 1.54178 Å), The intensities were recorded within the 2θ range from 10° to 60° with a voltage of 45 kV, and a current of 40 mA. Scanning electron microscope image was measured with a JEOL JSM 6700F instrument. XANES and EXAFS were recorded at BL17C of National Synchrotron Radiation Research Center (NSRRC), Hsinchu, Taiwan. Gas chromatography–mass spectrometry (GC-MS) was performed on Agilent 6890-5975 GC-MS with Inert XL Selective Detector. The GC is equipped with a capillary HP-5MS column (Model No.: 19091S-433, 30.0 m × 250 μm × 0.25 μm). The instrument is operated with an oven temperature of 50 ºC, an inlet temperature of 280 ºC, and a flow rate of 1.2 mL min−1 with helium carrier gas. A split/splitless injector is applied with a split ratio of 5:1 and a split flow of 5 mL min‒1. The MS has a source temperature of 230 ºC and a quadrupole temperature of 150 ºC. The SRI multiple gas analyzer #5 GC is equipped with 3 S.S. columns including 18′′ Hayesep D, 3′MS 5A and 6′ Hayesep D. The instrument is operated with an oven temperature of 50 ºC, a temperature profile from 50 to 270 ºC, and a flow rate of 40 mL min−1 at 15 psi with argon carrier gas. X-ray photoelectron spectroscopy (XPS) was measured on a Kratos AXIS Ultra spectrometer (Kratos Analytical, Manchester, UK).

Catalyst preparation

Homogeneous bimetallic catalyst 1 was prepared by dissolving vanadium pentoxide (V2O5) in 98% H2SO4 with ultra-sound treatment for 6 h. Homogeneous titanium (IV)-oxo and chromium (VI)-oxo catalysts were prepared by dissolved titanyl sulfate (TiOSO4) and potassium chromate (K2CrO4) in 98% H2SO4, respectively. The heterogeneous variant 4 (VOPO4·2H2O) was prepared based on previous literature37,40. V2O5 (4.8 g), H3PO4 (85.5%, 26.6 mL), and H2O (115.4 mL) were refluxed at 110 ºC for 16 h. After gently cooling down to room temperature, the yellow precipitate in the mixture was collected by filtration and washed several times with water and acetone. The resulting sample was dried in an oven at 100 ºC for 3 h. When 4 was investigated for its electrochemical response, 4 was loaded onto a FTO electrode via a dip-coating procedure. A dispersion of 4 was prepared at a concentration of 6 mg mL−1 in 2-propanol. The yellow dispersion was ultrasonicated for 30 min until the color of the dispersion became faded. Afterwards, sodium carboxymethyl cellulose (CMC) was added into the dispersion (weight ratio of VOPO4·2H2O: CMC = 80: 5). The mixture was stirred at 600 revolution per minute (rpm) on the heating plate to remove excess 2-propanol and form a homogenous slurry, which was then dip-coated onto FTO at a loading amount of 1.9 mg cm−2 for 4.

Electrochemical characterization

All electrochemical experiments were recorded using a Gamry Instruments Reference 600+ and Interface 1000 potentiostats. Unless mentioned specifically, a three-electrode setup was applied with a Pt wire pseudo-reference electrode and a Pt counter electrode. The Pt pseudo-reference electrode was calibrated to a Hg2SO4/Hg (saturated K2SO4) reference electrode (CH Instrument, Inc.) via the measurement of open-circuit potentials. The relationship is E(V vs. Hg2SO4/Hg) = E(V vs. Pt) + 0.755 V. The gas environment of the electrochemical cell was controlled either CH4 (Airgas, 99.5%) or N2 (Airgas, 99.999%), which were bubbled into the reactor at rates of 7.2 (CH4) and 10 (N2) standard cubic centimeters per minute (sccm) with the use of a mass flow controllers (Omega Engineering, Inc., FMA5510A). The data were reported after iR compensation. Unless noted specifically, the electrolyte is 98% H2SO4 with a certain vanadium concentration of 1.

Experiments of cyclic voltammetry were conducted in a single-chamber electrochemical cell with a 2-mm Pt working electrode (CH Instruments, Inc.). Bulk electrolysis with 5 mL homogeneous catalyst solution was typically conducted in a two-chamber electrochemical cell with a Nafion 117 membrane as the separator and a piece of commercial fluorine-doped tin-oxide (FTO) glass as the working electrode. Here, the FTO glass was encapsulated with Teflon tapes so that the exposed electrode is 1 × 1 cm in dimension. The solution was pre-saturated with CH4 or N2 for 20 min before the commencement of electrolysis. The background signal contribution of FTO glass was subtracted for the recorded data. Liquid aliquots were taken before, during and after the electrolysis for product analysis. Gaseous samples were taken manually from the outgas of the reactor, diluted with pure N2 (1:5 ratio) for transportation purpose, and manually injected into the GC–MS. The injected permanent gas was not well separated by the column, but the MS spectra of the eluted gas peak (t ~3.3 min) was capable to quantify permanent gases with a detection limit of ~10 ppm. Additional measurements of gas chromatography for the gaseous effluents was conducted by the SRI multiple gas analyzer #5. The experiments at elevated pressure was conducted into a custom-designed setup utilizing a pressure vessel with two gas feedthrough and three electric feedthroughs (Parr Instrument, Series 4600, 1000 mL). In this setup, the gas pressure was controlled between 1 and 11 bar and a constant gas flow of up to 28.7 sccm was maintained during the electrolysis. The temperature of the electrolysis was maintained by an oil bath at a range between 25 and 55 ºC. For experiments using substrates other than CH4, the same procedure is followed except the gas flow rate is set at 10 sccm calibrated against N2. In the 240-h electrolysis using natural gas mixture as the substrate, the electrolyte was refilled after each aliquot sampling in order to maintain a constant electrolyte volume. Experiment with 50% 13C-enriched CH4 was conducted by feeding regular CH4 and 13CH4 gases at equal flow rates controlled by two mass flow controller.

We found that typical reagent-grade 98% H2SO4 possesses nonzero background activity toward CH4 at high electrochemical potential (Supplementary Fig. 16). We argue that such background activity is due to the trace metal impurity, possibly V as H2SO4 is industrially synthesized by a V2O5 catalyst in the contact process. The reagent-grade 98% H2SO4 or oleum employed in most studies contains about 5 ppm of metal impurities (VWR BDH3074-3.8LP). When we employed high-purity 98% H2SO4 (Sigma-Aldrich 339741) that only contained 0.3 ppm of metal impurities as shown in the product certificate, the observed reactivity of neat 98% H2SO4 vanished (Supplementary Fig. 16).

When the heterogeneous variant 4 was investigated for its electrochemical response, a similar procedure was followed. However, owing to the difference of solution composition, the Pt pseudo-reference electrode has a different relationship with the Hg2SO4/Hg reference electrode: E(V vs. Hg2SO4/Hg) = E(V vs. Pt) + 0.268 V.

Attempts of using chemical oxidants at ambient conditions

When CH4 (7.2 sccm, 1 bar) was bubbled into a 98% H2SO4 solution with 10 mM 1 and 10 mM K2S2O8 for 6 h at ambient conditions, the formation of methyl bisulfate (CH3OSO3H) as a possible product of CH4 oxidation was not observed (Supplementary Fig. 6f). Similar experiment was conducted with 10 mM H2O2 in lieu of K2S2O8. The formation of methyl bisulfate (CH3OSO3H) as a possible product of CH4 oxidation was not observed (Supplementary Fig. 6f), either.

Quantification of liquid products

1H-NMR was applied to quantify product accumulation in DMSO-d6 using acetic acid (CH3COOH) as the internal standard. Totally, 0.4 mL liquid aliquots from electrolysis were mixed with 0.1 mL DMSO-d6 prior to the measurements. Chemical shifts are reported on a parts-per-million (ppm) scale. Methyl bisulfate (CH3OSO3H) exhibits a singlet at 3.34 ppm while the singlet from acetic acid (CH3COOH) peak resides at 1.96 ppm. A calibration curve was constructed by determining the relative ratio of integrated area between the NMR peaks of CH3OSO3H and CH3COOH. Product quantification of C2H6, C3H8, and natural gas mixture follows the same protocol, except for the quantification of CH3COOH as a C2 product from C2H6. The quantification of CH3COOH as a C2 product was fulfilled by adding a known concentration of CH3OSO3H as an internal standard in a separate 1H-NMR measurement.

Calculation of FE

The F.E. of bulk electrolysis was calculated based on the following equation:

$${\rm{F.E.}} = \frac{{nFC_{\rm{product}}V_{\rm{solution}}}}{{{\rm{Overall}}\,{\rm{charge}}}} \times 100\%,$$
(1)

Here, F is the Faraday’s constant, Cproduct is the concentration of product after bulk electrolysis, Vsolution is the total electrolyte volume, and the overall charge is the total electric charges passed through the working electrode. The variable n in the equation is the number of electrons required in order to generate one product molecule by electrochemistry. n = 2 for the formation of methyl bisulfate (CH3OSO3H) from CH4. n = 2 and 6 for the formation of ethyl bisulfate (C2H5OSO3H) and acetic acid (CH3COOH) from C2H6, respectively. n = 2 and 4 for the formation of isopropyl bisulfate (i-C3H7OSO3H) and acetone (CH3COCH3) from C3H8, respectively.

Calculation of TOF and TON

In the following we provide the protocols that we calculate the TOFs and TONs for the reported data, based on the methods established in prior literature19,23.

The diffusion coefficient for 1 (D) was determined from the cyclic voltammgrams based on the Randles–Sevcik equation22:

$$j_p = 0.4463nFC_{\rm{cat}}\left( {\frac{{nFvD}}{{RT}}} \right)^{\frac{1}{2}},$$
(2)

Here jp is the peak current density of quasi-reversible redox couple, n is the number of electrons transferred in the redox event, F is the Faraday’s constant, Ccat is 1’s concentration, v is the scan rate, R is the gas constant, and T is the temperature of experiment. As derived from Supplementary Fig. 1a, b, D = 2.18 × 10−7 cm2 s−1 for species 1 in the electrolyte.

The observed TOF of bulk electrolysis was determined based on the following equation19,23:

$${\rm{TOF}} = \left( {\frac{{j_{{\rm{product}}}}}{{nFC_{\rm{cat}}}}} \right)^2\frac{1}{D},$$
(3)

Here, jproduct is the partial current density of product formation in bulk electrolysis, n is the number of electrons required to generate one product molecule, F is the Faraday’s constant, Ccat is the concentration of catalyst 1, D is the diffusion coefficent of catalyst (2.18 × 10−7 cm2 s−1 for species 1 as determined above).

Similarly, the TON of bulk electrolysis was determined based on the following equation19,23:

$${\rm{TON}} = \frac{{C_{\rm{product}}V_{\rm{solution}}}}{{AC_{\rm{cat}}}}\sqrt {\frac{{\rm{TOF}}}{D}}.$$
(4)

Here, Cproduct is the product concentration after bulk electrolysis, Vsolution is the total electrolyte volume, A is the electrode area, Ccat is the concentration of catalyst 1, D is the diffusion coefficent of catalyst (2.18 × 10−7 cm2 s−1 for species 1 as determined above), and TOF is the TOF calculated based on above protocol.

On the condition of a homogenous process with molecular catalyst 1, the diffusion coefficient extracted from the V(V)/V(IV) redox couple is a reasonable approximate of the real catalytic redox couple, the vanadium(V)-oxo dimer and its one-electron-deficient cation radical, because we are unable to characterize the cation radical due to its transient nature. Here, we offer a rough estimate of the relative uncertainty of such an approximation based on our proposed mechanism. The Stokes–Einstein relationship41 predicts that the diffusion coefficient D of similar molecules follows: D ~MW−1/3, in which MW is the molecular weight. As MW = 538.16 g mol−1 for 1, the relative uncertainty of D in our procedure is about 6% or 12%, assuming the loss of one or even two of the sulfonic ligands, respectively. As the TOF ~D−123, at most about 10% relatively uncertainly will incur in our practice.

Computational methods

All calculations were performed with Turbomole42,43,44,45,46,47,48,49,50,51,52 using the M06 density functional53. The def2-SVP basis set was used for geometry optimizations and free energy corrections, and the def2-TZVP basis set was used for electronic energies54. Solvation was modeled with COSMO55 with the dielectric constant set to 10156. Images were rendered using Chemcraft57. The natural bond orbital (NBO) analysis was used for atomic charge calculations36.

Details of XAS experiments

XAS, including X-ray absorption near edge spectra (XANES) and extended X-ray absorption fine structure (EXAFS), at V K-edge were collected in total-fluorescence-yield mode at ambient conditions at BL17C of National Synchrotron Radiation Center (NSRRC), Hsinchu, Taiwan. Spectra were recorded in the energy range from −100 to 600 eV, relative to that of V K-edge absorption (5465.0 eV). The XAS spectra were processed by subtracting the baseline of pre-edge and normalizing that of post-edge. EXAFS analysis was carried out using Fourier transform on k3-weighted EXAFS oscillations to assess the contribution of each bond pair to Fourier transform peak. The curve fitting of EXAFS spectra was conducted using the software, REX2000, with FEFF program.

The operando XAS experiments at V K-edge were conducted under the same procedure at TPS beamline 44A of NSRRC, Hsinchu, Taiwan. A three-electrode arrangement was used during the operando measurements. The electrolyte was saturated with 1-bar CH4, and the measurements were performed using an Autolab PGSTAT204 potentiostat (Metrohm Autolab) in a customized reactor.

Operando Raman characterization

A three-electrode setup in a home-made cell was adopted for Operando Raman spectroscopy and the electrochemical measurements. The measurements in a CH4-saturated electrolyte were recorded using a Raman microscopy (UniNano UNIDRON) and an Autolab PGSTAT204 potentiostat (Metrohm Autolab). A laser of 633 nm with a spot size of ~1 µm2 served as the excitation source, and the output power was 2.5 mW. A 50× objective lens was employed for operando measurements during electrolysis, while all results were obtained under an exposure duration of 3 s with the accumulation number of 60 times.