Main

Traditional CH₄ oxidation catalysts are primarily based on solid materials in which the chemical environment of the active sites determines the catalyst’s activity and selectivity. Single-atom catalysts, due to their unique singular active sites, typically yield only C1 products, whereas clusters or particles with multiple atomic sites commonly produce C2 or longer-chain hydrocarbons1,2,3,4,5. However, these solid catalysts have an inherent flaw: their rigid solid structure restricts the ability to modify the chemical configuration of active sites during reactions. This notably limits the possibility of adjusting catalytic activity and selectivity during the reaction process.

Liquid metals, as emerging materials, present a novel approach for developing CH₄ oxidation catalysts. Their unique physical and chemical properties, such as low melting point, high thermal conductivity and excellent chemical stability, make them ideal catalyst carriers6,7, capable of dissolving specific metals and achieving stable dispersion at the atomic level8. In contrast to solid catalysts, liquid metals remain liquid at room temperature, with atoms not confined by a lattice, thus offering the possibility of dynamically adjustable structures. Unlike traditional catalytic adjustment methods, including the application of thermal, optical and electric fields, a magnetic field is easily generated by permanent magnets, and allows for external control of catalytic reactions under milder conditions. Recent studies on magnet-assisted catalysis have primarily focused on manipulating the electron spins at the catalytic centre to improve the adsorption and mass transfer of reactive molecules9,10,11.

Here we show a magnetic-field-responsive catalyst composed of an iron-embedded, catalytically active liquid metal solution (Fe–LMS), and a eutectic alloy, galinstan, as a weak-binding substrate. This Fe–LMS catalyst responds to changes in external magnetic field intensity, enabling a reversible and rapid rearrangement of iron atoms between isolated single atoms (Fe1–LMS) and agglomerated clusters (Fen–LMS) (Fig. 1, left). As a result, we achieve reversible conversion of the main liquid product of CH4 oxidation between CH3OOH and CH3COOH with high production rates (1,679.6 ({\rm{m}}{\rm{m}}{\rm{o}}{\rm{l}},{{\rm{g}}}_{{\rm{F}}{\rm{e}}}^{{-1},{{\rm{h}}}}{-1}) and 790.5 ({\rm{m}}{\rm{m}}{\rm{o}}{\rm{l}},{{\rm{g}}}_{{\rm{F}}{\rm{e}}}^{{-1},{{\rm{h}}}}{-1}), respectively) and high selectivities (99.9% and 91.7%, respectively) under a field of 0–500 G at room temperature (Fig. 1, right). Through in situ measurements and theoretical calculations, we find that the magnetic field switching influences the nanostructure, local coordination environment and electron spin of Fe–LMS, leading to the generation of intermediate species that determine the reaction pathways of C1 or C2. In particular, the change in the aggregation state and spin state of iron atoms in the presence of the external magnetic field breaks the adsorption energy scaling relationship with reactive molecules, altering reaction selectivity.

Fig. 1: Schematic of the mechanism.

Schematic diagram of macroscopic (left) and microstructural (right) changes in Fe–LMS under the control of a magnetic field switch.

Characterization of the catalyst with a sensitive magnetic field response

In this work, Fe–LMS catalyst supported on liquid galinstan alloy was used for the selective oxidation of CH4. Supplementary Fig. 1 and Supplementary Movie 1 provide a macroscopic perspective of the reversible morphological changes in a Fe–LMS droplet under the influence of a magnetic field. The transmission electron microscopy (TEM), scanning electron microscopy (SEM) and aberration-corrected scanning transmission electron microscopy (STEM) images in Supplementary Fig. 2 and Fig. 2a reveal that ultrasonically dispersed Fe–LMS microdroplets present a perfect sphere with a smooth edge. The energy-dispersive X-ray spectroscopy (EDS) image in Fig. 2b confirms the uniform distribution of gallium, indium and tin in galinstan, and an iron loading of around 1% (Supplementary Table 1) is observed to be evenly spread over the Fe–LMS microdroplet. In situ X-ray two-dimensional computed tomography (2D CT) slice (Fig. 2c) and 3D views (Supplementary Fig. 3) reveal a uniform, symmetric spherical distribution of iron under 0 G. The 2D CT slice (Fig. 2d) and 3D views (Supplementary Fig. 3) show a notable deviation under 500 G, with iron atoms aggregating towards one side, indicating the influence of the field on their spatial arrangement.

Fig. 2: Characterization of Fe–LMS.

a,b, STEM (a) and EDS elemental mapping (b) images of Fe–LMS. c, 2D CT reconstruction images for Fe–LMS under 0 G. d, 2D CT reconstruction images for Fe–LMS under 500 G. e, HAADF-STEM images for Fe–LMS under 0 G. f, Partially enlarged image of e. g, The corresponding iron EDS mappings for f. h, The corresponding gallium EDS mappings for f. i, HAADF-STEM images for Fe–LMS under 500 G. j, Partially enlarged image of i. k, The corresponding iron EDS mappings for j. l, The corresponding gallium EDS mappings for j. m, Iron K-edge EXAFS spectra in R space for the iron foil, Fe–LMS–1 wt% (0 G) and Fe–LMS–1 wt% (0 G). n, Mössbauer spectra for iron standard sample under 0 G, Fe–LMS–1 wt% under 0 G and Fe–LMS–1 wt% under 500 G. o, ESR absorption versus magnetic field for iron standard sample, Fe–LMS–1 wt% under 0 G and Fe–LMS–1 wt% under 500 G.

Source Data

We used atomic-level-resolution high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and corresponding high-resolution energy-dispersive X-ray spectroscopy (EDS) mapping. Without a magnetic field, the sample shows amorphous dispersion of iron/gallium atoms (Fig. 2e–h). In contrast, a 500-G magnetic field induces α-Fe(100) cluster formation (0.206-nm lattice spacing, Fig. 2i–l), while gallium remains uniformly distributed. Quantitative EDS confirms magnetic-field-driven iron aggregation without altering the gallium distribution (Supplementary Fig. 4).

X-ray absorption fine structure (XAFS) analysis reveals magnetic-field-induced structural reorganization in Fe–LMS. At 1 wt% iron loading and zero-field conditions, the R-space spectra shows a single coordination peak, while 500-G exposure generates additional peaks at 3.7 Å and 4.5 Å (second and third nearest-neighbour coordination, Fig. 2m), similar to those seen in iron foil and indicating reversible iron aggregation (Supplementary Fig. 5). Higher iron loading (10 wt%) inherently exhibits these features (Supplementary Fig. 6). X-ray absorption near-edge spectra (Supplementary Fig. 7) and k-space data (Supplementary Fig. 8) further confirm distinct coordination environments between dispersed and aggregated state samples. Curve fitting (Supplementary Figs. 9–11 and Supplementary Table 2) demonstrates gallium as the primary coordination shell for iron in 1 wt% Fe–LMS, with Fe–Fe bonds emerging via magnetic-field application. Self-absorption effects were systematically excluded through fluorescence-mode XAFS measurements at a range of incident angles (Supplementary Fig. 12), validating data reliability.

In situ Mössbauer spectroscopy was used to characterize the catalyst’s structure. Due to the measurement principle, free atoms are excited to emit photons, experiencing recoil momentum and preventing resonance absorption detection, whereas lattice-embedded clusters show negligible recoil, enabling measurable signals12,13,14,15. Under zero-field conditions, the absence of detectable signals (Fig. 2n) confirms atomically dispersed iron. Under 500 G, a signal peak with a shape remarkably similar to that of the iron reference sample is observed (see fitting data in Supplementary Fig. 13 and Supplementary Table 3), demonstrating magnetic-field-driven cluster formation. Electron spin resonance (ESR) measurements further corroborate this transition, showing standard-like ferromagnetic signatures only under magnetic fields, while zero-field conditions exhibit weak signals due to gallium-mediated spin relaxation (Fig. 2o)16,17,18. The system shows near-instantaneous magnetic response (Supplementary Fig. 14), with magnetization scaling linearly with field intensity and complete reversibility upon field removal.

Switching of methane oxidation reaction pathway regulated by magnetic field

The CH4 oxidation reaction was conducted in a high-pressure autoclave (2 MPa CH4) using Fe–LMS droplet catalysts and diluted H2O2. The reactor was placed between adjustable magnets (0–500 G). Products were analysed by gas chromatography (GC) and 1H NMR. Gallium-based substrates required indium/tin additives to reduce melting points (see phase diagrams in Supplementary Fig. 16). Ga–In and Ga–Sn exhibited catalytic activity comparable to that of galinstan, confirming their primary role as melting-point depressants (Supplementary Fig. 17). Given that the galinstan alloy formulation demonstrates the most notable melting point depression and is widely used in applications19,20,21, it was selected as the liquid metal substrate in this work. The Fe–LMS-catalysed CH4 oxidation reaction generates two main kinds of liquid (CH3OOH and CH3COOH) and gas (CO and CO2) products (Fig. 3b,c and Supplementary Figs. 18 and 19), respectively.

Given H2O2’s strong oxidative nature, potential metal oxide formation from Fe–LMS was evaluated. Catalytic testing (Supplementary Fig. 20) confirmed negligible contributions of all possible metal oxides to liquid product yields, irrespective of magnetic field application. Supplementary Fig. 21 reveals that mild acidity (pH ≈ 4) achieves relatively high yields of both CH3OOH and CH3COOH, whereas extreme pH values degrade catalytic activity. This pH condition was selected as optimal for dual-product generation under magnetic field modulation and adopted as the standard for subsequent experiments. Supplementary Fig. 22 confirms negligible Cl− influence from HCl using NaCl concentration gradients. Optimal activity occurs with 6% H2O2 (Supplementary Fig. 23), demonstrating the necessity of HCl/H2O2 synergy to maintain surface oxidation states. Visual evidence (Supplementary Fig. 24) shows that Fe–LMS retains spherical morphology only at pH 4 with 6% H2O2, while other conditions (pH 7 or 15% H2O2) induce oxidation.

Without magnetic fields, Fe–LMS primarily produces CH3OOH (471.1 ({\rm{m}}{\rm{m}}{\rm{o}}{\rm{l}},{{\rm{g}}}_{{\rm{F}}{\rm{e}}}^{{-1},{{\rm{h}}}}{-1}), 94.5%). Under 500 G, product selectivity reverses, yielding CH3COOH as the main product (230.6 ({\rm{m}}{\rm{m}}{\rm{o}}{\rm{l}},{{\rm{g}}}_{{\rm{F}}{\rm{e}}}^{{-1},{{\rm{h}}}}{-1}), 91.9%). Production rates scale linearly with iron loading (Supplementary Fig. 25), while long-term tests (48 h, Supplementary Fig. 26) show stable C1 yields (zero-field) or steady CH3COOH production (∼90% selectivity, magnetic field). Ultrasonic dispersion yields 100-nm Fe–LMS@H2O droplets (Supplementary Fig. 27), maintaining size postreaction. This enhances catalytic surface area, tripling both CH3OOH (1,679.6 ({\rm{m}}{\rm{m}}{\rm{o}}{\rm{l}},{{\rm{g}}}_{{\rm{F}}{\rm{e}}}^{{-1},{{\rm{h}}}}{-1}), 99.9%) and CH3COOH (790.5 ({\rm{m}}{\rm{m}}{\rm{o}}{\rm{l}},{{\rm{g}}}_{{\rm{F}}{\rm{e}}}^{{-1},{{\rm{h}}}}{-1}), 91.7%) production rates (Fig. 3b). Such high generation rates of CH3OOH and CH3COOH both attain excellent levels in the thermocatalytic field of CH4 oxidation as displayed in Supplementary Table 4.

Fig. 3: Catalytic performance in CH4 oxidation.

a, Schematic diagram of the reaction device. b, Production rates of liquid products (CH3OOH and CH3COOH) for Fe–LMS controlled by an external magnetic field. Data are represented as the mean ± s.d. from five parallel experiments. c, Production rates of gaseous products (CO and CO2) for Fe–LMS controlled by an external magnetic field. Data are represented as the mean ± s.d. from five parallel experiments. d, Correlation between the production rate of CH3COOH and additionally applied CO partial pressure for Fe–LMS@H2O under a 500-G magnetic field. e, 13C NMR spectra of liquid product for Fe–LMS without a magnetic field. Experimental conditions: 2 MPa 13CH4, 25 ml H2O2 (6%), pH 4, r.t., 1 h (red line); 2 MPa 12CH4, 25 ml H2O2 (6%), pH 4, r.t., 1 h (grey line). f, 13C NMR spectra of liquid product for Fe–LMS with a 500-G magnetic field. Experimental conditions: 2 MPa 13CH4, 25 ml H2O2 (6%), pH 4, r.t., 1 h (red line); 2 MPa 13CH4 + 0.5 MPa 12CO, 25 ml H2O2 (6%), pH 4, r.t., 1 h (green line); 2 MPa 12CH4, 25 ml H2O2 (6%), pH 4, r.t., 1 h (grey line). g, Correlation between selectivity of liquid products and magnetic induction intensity. Data are represented as the mean ± s.d. from five parallel experiments. h, Selectivity of CH3COOH for Fe–LMS when applying multiple off–on operations of a magnetic field. Each operation lasts for 1 h and CH4 is refilled after each operation.

Source Data

Under 500 G, CO production increases sharply whereas CO2 yield remains stable (Fig. 3c). Previous studies reported that mixing CO into CH4 as a coreactant leads to the formation of CH3COOH22,23. There is a linear increase in CH3COOH yield with higher CO/CH4 ratios under magnetic fields (Fig. 3d). In the absence of a magnetic field, the introduction of CO only slightly increased the yield of CH3COOH (Supplementary Fig. 28), with the main product on the dispersed-state Fe–LMS remaining CH3OOH. 13C NMR analysis confirmed distinct carbon sources for CH3OOH and CH3COOH (Fig. 3e,f). CH3OOH was derived exclusively from 13CH4, while CH3COOH showed characteristic peaks at 176 ppm (carbonyl) and 20 ppm (methyl)24. Introducing 12CO into 13CH4 (4:1) reduced the carbonyl signal, demonstrating CO’s role as a key intermediate in CH3COOH formation. Product selectivity is magnetically tunable, with CH3OOH:CH3COOH ratios shifting progressively from 95:5 (0 G) to 8:92 (500 G) (Fig. 3g). The system exhibits robust reversibility, maintaining CH3OOH (>90%) and CH3COOH ( > 80%) selectivity over 11 magnetic switching cycles (Fig. 3h), demonstrating precise control over product distribution via field intensity.

In situ experiment and mechanism exploration

Ambient-pressure X-ray photoelectron spectroscopy (APXPS) was performed to investigate the real chemical environment during the reaction process (Supplementary Fig. 29). In a vacuum environment, an Fe(0) signal was detected (Fig. 4a). Upon the introduction of oxidants, the iron signal intensified, and the oxidation state increased (Fig. 4b)25. However, no signal of signal was detected on the Fe–LMS immersed in 30% H2O2 solution for 1 h (Fig. 4c). Gallium element signals revealed that, with the introduction of oxidants, a portion of the initially zero-valent gallium undergoes oxidation to form Ga2O3 (1,117.0 eV to 1,118.2 eV, Supplementary Fig. 30)26. These results indicate that iron atoms migrating to the surface are anchored by oxygen atoms, forming Fe–O–Ga structures, which are identified as the active sites. When the catalyst surface is completely oxidized by Ga2O3, iron atoms are encapsulated inside and lose the ability to contact with the reactants. Intermediate species of CHx (285.6 eV) and *OCH3 (286.4 eV) are simultaneously detected on the dispersed catalyst (Fig. 4d)27,28. This is probably due to the C–H bond cleavage of CH4 to generate CHx, followed by being adsorbed on metal atoms or bonding with oxyen atoms to form *OCH3. *COx (289.8 eV) is found as the main intermediate species on the aggregated catalyst (lower spectra in Fig. 4d). A new peak appearing at 293 eV is attributable to the physical adsorption of CO. The gaseous CO is probably generated by the further desorption of *COx, which has been identified as the main gas product in Fig. 3c.

Fig. 4: In situ experiments and mechanism exploration.

a, NAP-XPS spectra for Fe–LMS (1 wt% iron loading) in a vacuum environment. cps, counts per second. b, NAP-XPS spectra for Fe–LMS (1 wt% iron loading) in a 0.13-mbar H2O2 solution (30%) vapour atmosphere. c, NAP-XPS spectra for Fe–LMS (1 wt% iron loading) following a 24-h exposure to a 0.13-mbar H2O2 (30%) vapour atmosphere. d, NAP-XPS spectra for dispersed-state and aggregated-state catalysts. e, Iron K-edge EXAFS spectra in R space recorded without a magnetic field for pristine Fe–LMS, Fe–LMS immersed in H2O2, and Fe–LMS under reaction conditions. f, Partially enlarged image of c (dashed line indicates the peak position). g, Iron K-edge EXAFS spectra in R space recorded under a 500-G magnetic field for pristine Fe–LMS, Fe–LMS–B immersed in H2O2, and Fe–LMS–B under reaction conditions. h, Partially enlarged image of d (dashed line indicates the peak position). i, Iron K-edge EXAFS spectra in R space recorded under a 500-G magnetic field for Fe–LMS–B with H2O2, Fe–LMS–B under reaction conditions, and Fe–LMS–B with an additional 1 bar and 5 bar CO beyond the reaction conditions. j, Schematic diagram of reversible regulation of structure changes by magnetic field switches leading to changes in reaction pathways.

Source Data

In situ XAFS (Supplementary Fig. 29c) revealed dynamic changes in the local environment of iron atoms during the reaction. Fe1–LMS in the absence of a magnetic field showed a 1.5-Å Fe–O peak upon H2O2 exposure, signifying formation of the Fe–O bond. With the simultaneous introduction of H2O2 and CH4, this peak shifted +0.1 Å (1.9 → 2.0 Å, Fig. 4f and Supplementary Table 2), indicating Fe–O bond elongation, consistent with in situ XPS data showing bond weakening due to *OCH3 formation. Under a 500-G field, the spectrum of Fe–LMS provided with H2O2 lacks a legible peak with respect to the Fe–C/O bond (Fig. 4g,h); however, an enhanced peak at 1.6 Å emerges in the spectrum of Fe–LMS provided with H2O2 and CH4, illustrating the existence of an Fe–C/O bond. Considering that the intermediate species *COx was detected in the in situ XPS results for Fe–LMS under similar conditions, it is rational to speculate that the enhanced peak concerning the Fe–C/O bond derives from the formation of intermediate species *COx on iron atoms during the reaction period. Here, CO as the main gas product for the CH3COOH path was additionally introduced into the reaction system at gradually increasing partial pressures (0, 1 and 5 bar). Figure 4i shows ever-heightening peaks of Fe–C/O bonds with gradually increasing CO partial pressures, clarifying that the enhancement of these peaks is associated with the adsorption of *CO. The above results explicitly revealed an apparent distinction of coordination environments for iron atoms under magnetic field switching. The mechanism of the reversible reaction pathway is depicted in Fig. 4g. The magnetic field switching induces reversible conversion of the states of iron atoms between dispersion and agglomeration, resulting in the formation of different intermediate species and thus finally realizing the alteration of the main liquid products.

Theoretical works

To understand the selectivity of reaction pathways due to the structural change of catalysts, we performed ab initio molecular dynamics (AIMD) simulations and density functional theory (DFT) calculations. In the absence of a magnetic field, it is found that two iron atoms in the liquid gallium alloy are inclined to disperse into two iron single atoms due to the entropy effect (Fig. 5a and Supplementary Fig. 31). However, under a magnetic field, the two iron atoms are agglomerated into an iron dimer (Fig. 5a and Supplementary Fig. 32); single iron atoms tend to align their spins uniformly and move along the direction of the field, leading to agglomeration under a magnetic field. In addition, it is found that an iron atom surrounded by gallium metal in the droplet configuration can be pulled out and anchored by the oxygen atoms on the surface (Supplementary Fig. 33). This result is consistent with that obtained from previous in situ experiments, and is based on a comprehensive analysis combining XAFS results (Supplementary Table 2) and AIMD stability assessments (Supplementary Figs. 34 and 35). The structure of the two O-coordinated iron single atoms is relatively stable without a magnetic field. For the agglomerated iron clusters with an external magnetic field, we chose the tetrahedral Fe4 with the coordination number of Fe–Fe being 3 for calculations.

Fig. 5: Theoretical calculations.

a, AIMD simulation trajectories of the Fe–Fe distance as a function of time. The initial distance between the two iron atoms is 3.5 Å, with the spin moment direction unfixed and fixed, respectively. b, Calculated Gibbs free energy profile and schematic diagram of CH4 to CH3 on Fe1–LMS. c, Calculated Gibbs free energy profile of partial steps of CH3O to CH2O on Fe1–LMS and Fe4–LMS. d, PDOS of the C of CH3O in Fe1–LMS and Fe4–LMS. e, PDOS of the O of CH3OH and CH3O in Fe1–LMS (dashed line indicates the peak position). f, PDOS of the O of for CH3OH and CH3O in Fe4–LMS (dashed line indicates the peak position).

Source Data

Next, we investigated the generation of the Ga–O–Fe structure from H2O2 decomposition. We found that H2O2 can be easily decomposed into two OH or into O and H2O on the iron active site without oxygen anchoring (Supplementary Fig. 36a), and this process is barrierless. Once the iron active site is fixed by the released O, H2O2 can be formed through the interaction of a hydrogen bond with an adjacent O (Supplementary Fig. 36b,c), and can then dissociate to OOH and H via O–H rupture with a barrier of 0.33 eV (TS1). The O–O bond-cleavage barrier is 0.27 eV through the transition state TS2. Finally, the Ga–O–Fe structure eventually forms on the catalyst surface.

We further researched the process of CH4 oxidation on Fe1–LMS and Fe4–LMS. Figure 5b shows the calculated Gibbs free energy profile of partial steps of CH4 oxidation to CH3OH on Fe1–LMS (see Supplementary Figs. 37–39 for complete steps). CH4 is initially located on Fe1–LMS through the iron atom. CH4 dehydrogenation occurs via TS3 with an energy barrier of 1.05 eV. Then, the generated OOH combines with a CH3 radical to form CH3OOH through TS4 (0.44 eV). Apparently, methyl hydroperoxide (CH3OOH) is the sole first product of peroxide-assisted CH4 oxidation, which is consistent with a previous study29. The CH3OOH is decomposed into CH3O and OH via TS5 (0.18 eV) by cleavage of the O–O bond. Finally, CH3O is readily reduced to methanol (TS6, 0.32 eV). Clearly, the CH4 dehydrogenation is the rate-determining step with a barrier of 1.05 eV. In addition, the formed CH3O continues the dehydrogenation with a large barrier of 3.03 eV via TS7 on Fe1–LMS, which is unfavourable to forming oxides with higher oxidation state (CO, CH3COOH, etc.) (Fig. 5c and Supplementary Fig. 40).

When the iron single atoms directionally move to form the iron cluster catalyst (Fe4–LMS) induced by the magnetic field, the CH4 oxidation pathway is completely altered and the main product is changed into CH3COOH. Figure 5b shows the calculated Gibbs free energy profile of partial steps of CH4 oxidation to CH3COOH on Fe4–LMS (see Supplementary Figs. 37, 41 and 42 for complete steps). CH4 dehydrogenation produces CH3 via TS8 with a barrier of 0.67 eV. Then, the formed CH3 combines with OOH to generate CH3O and OH (exothermic by 4.11 eV). The CH3O is changed into CO via three continuous dehydrogenation steps with barriers of 0.25, 0.62 and 0.79 eV through TS9–TS11, respectively. By comparison, this reaction is more difficult to achieve on Fe1–LMS with a high barrier of 3.03 eV (Fig. 5c and Supplementary Fig. 40). The CO couples with CH3 to generate CH3CO via TS12 (0.47 eV). CH3CO continues to combine with OH to produce CH3COOH through TS13 (0.95 eV). In terms of kinetics, the reaction CH3CO + OH → CH3COOH is the rate-determining step with a barrier of 0.95 eV. Furthermore, the formation of CH3OH (CH3O + H → CH3OH) on Fe4–LMS is difficult due to the high barrier of 1.47 eV via TS14 (Supplementary Fig. 42)

We further probed the intrinsic reasons for the selectivity of CH4 oxidation on Fe1–LMS and Fe4–LMS. We determined that the intermediate CH3O is the crossover point between the CH3OH and CH3COOH pathways. The projected density of states (PDOS) of C–H bonding of CH3O in Fe1–LMS and Fe4–LMS (Fig. 5d) indicate that the DOS peak (–0.15 eV) of carbon is closer to the Fermi level in Fe4–LMS than in Fe1–LMS (–0.63 eV), which makes it more likely to lose electrons and become oxidized. Therefore, the CH3O easily undergoes dehydrogenation (oxidation) in Fe4–LMS. We also calculated the orbitals and PDOS of the O–H bonding of CH3OH and the O of CH3O in Fe1–LMS and Fe4–LMS (Fig. 5e,f). We found that the DOS peak position of the O of CH3OH is more similar to that of the O of CH3O near the Fermi level in Fe1–LMS (Fig. 5e) compared with in Fe4–LMS (Fig. 5f). Therefore, the CH3O easily undergoes hydrogenation in Fe1–LMS. In addition, the d-orbital centre (−2.00 eV) of iron in Fe4–LMS is closer to the Fermi level than that in Fe1–LMS (−2.25 eV), and thus exhibits a stronger adsorption energy for CH3O (−2.94 eV) than that of Fe1–LMS (−2.47 eV) (Supplementary Fig. 44a). Therefore, CH3O is suitable for continuous dehydrogenation reactions to generate CO in Fe4–LMS, and CH3O is suitable for hydrogenation and desorption (desorption energy, 0.55 eV) in Fe1–LMS. We also found that O-2p and Fe-3d form antibonding σ1*, π1* and π2* bonds between CH3O with Fe1–LMS (Supplementary Fig. 44a,c) and form an antibonding σ1* bond between CH3O with Fe4–LMS (Supplementary Fig. 44a,f) through crystal orbital Hamiltonian population calculations of iron and oxygen (Supplementary Fig. 44b,e).

In addition, CO is the key intermediate for producing C2 products through C–C coupling. In Fe4–LMS, the barrier of C–C coupling on Fe–Fe active sites is 0.47 eV through TS12 (Supplementary Figs. 41 and 42). In Fe1–LMS, the barrier of C–C coupling on Fe–Ga active sites is as high as 0.94 eV via TS15 (Supplementary Fig. 45). In addition, CO only forms with difficulty due to the high barrier of 3.03 eV for CH3O dehydrogenation (Supplementary Fig. 40). Therefore, the C–C coupling for the C2 product (CH3COOH) requires the iron cluster to simultaneously anchor the CO and CH3, which is attributed to the enhanced CO adsorption energy on the iron cluster (1.71 versus 0.70 eV).

Conclusions

We have prepared a catalyst of Fe–LMS with a sensitive magnetic field response, which enables reversible conversion of the main liquid products of CH4 oxidation (that is, CH3OOH and CH3COOH) under the external control of a magnetic field at room temperature. We attain high production rates and selectivities and steady reproducibility. The external magnetic field manipulates the spin orientation of the iron single atoms, guiding the directional movement and agglomeration of the iron atoms from a dispersion state. Our work may provide a guide for the precise control of reaction pathway of a catalyst via a magnetic field, showcasing potential benefits of magnetic-based catalysis adjustment in improving reaction selectivity, streamlining reaction apparatus and conditions, and reducing both energy consumption and operational costs for industrial synthesis.

Methods

Synthesis methods

Preparation of Fe–LMS, Fe@Ga–In, Fe@Ga–Sn and Fe@Ga

The synthesis Fe–LMS must be performed in a nitrogen atmosphere throughout to prevent oxygen oxidation.

Gallium (99.999%) was first melted at 50 °C in a beaker. Subsequently, 7 g molten gallium metal was mixed with 2 g indium powder (99.999%) and 1 g tin powder (99.999%), followed by vigorous stirring at 160 °C for 5 h. Different amounts of iron powder (0.1–2 g, 99.999%) were added into the liquid metal alloy followed by manual stirring until the iron powder was entirely dissolved. Then, 100 µl of 5% hydrochloric acid was added to keep the alloy in a liquid state. The obtained liquid metal catalyst Fe–LMS was stored in an oxygen-free container.

After adjusting the feeding ratio to Fe:Ga:In to 0.1:7:3 (w/w/w/), the aforementioned process was repeated to obtain Fe@Ga–In. After adjusting the feeding ratio to Fe:Ga:Sn to 0.09:7:1 (w/w/w/), the aforementioned process was repeated to obtain Fe@Ga–Sn. After adjusting the feeding ratio to Fe:Ga to 0.1:10 (w/w) and adjusting the temperature to 200 °C, the aforementioned process was repeated to obtain Fe@Ga.

Preparation of Fe–LMS@H2O

First, 10 mg of prepared Fe–LMS was added to 20 ml of deionized water, and the mixture was then placed in a room-temperature water bath and agitated ultrasonically for 30 min until the solution became turbid. The synthesis of Fe–LMS@H2O does not require a protective atmosphere. Fe–LMS@H2O is a suspension, and the prepared Fe–LMS@H2O must be used immediately to avoid sedimentation.

Catalytic performance evaluation

Methane oxidation by Fe–LMS with no magnetic field

Oxidation of CH4 was performed in a stainless-steel Teflon-lined autoclave with a volume of 100 ml. Typically, 10 mg catalyst Fe–LMS, 20 ml deionized water and 5 ml H2O2 (30%) were added to the autoclave. Hydrochloric acid was added dropwise until the pH of the solution approached 4. The autoclave was flushed three times with methane and then pressurized with methane to the desired pressures (0.5–3.0 MPa CH4, 99.999%). The reaction proceeded for 1 h at room temperature. The autoclave with obtained products was cooled in ice water for 20 min prior to analysis. Liquid products were quantified by 1H and 13C NMR spectroscopy. The gas-phase products are discharged through the reactor’s exhaust valve and collected in a gas bag, which is subsequently transferred to the GC for analysis.

Methane oxidation by Fe–LMS with a magnetic field

The CH4 oxidation reaction was carried out in a closed high-pressure autoclave. The high-pressure reactor was placed between two parallel permanent magnets, and the intensity of the magnetic field was controlled by an external distance-adjustment device. About 10 mg catalyst, 20 ml deionized water and 5 ml H2O2 (30%) were added to the autoclave. Hydrochloric acid was added dropwise until the solution pH approached 4. The autoclave was flushed three times and then pressurized with methane to the desired pressure (0.5–3.0 MPa CH4, 99.999%). The reaction mixture was left at room temperature for 1 h. The magnetic field can be regulated over the range 0–1,200 G. The autoclave with obtained products was cooled in ice water for 20 min prior to analysis. Liquid products were quantified by 1H and 13C NMR spectroscopy. The gas-phase products are discharged through the reactor’s exhaust valve and collected in a gas bag, which is subsequently transferred to the GC for analysis.

Methane oxidation by Fe–LMS@H2O with magnetic field switching

The CH4 oxidation reaction was carried out in a closed high-pressure autoclave. The high-pressure reactor was placed between two parallel permanent magnets, and the intensity of the magnetic field was controlled by an external distance-adjustment device. First, 2 ml Fe–LMS@H2O solution, 20 ml deionized water and 5 ml H2O2 (30%) were added to the autoclave. Hydrochloric acid was added dropwise until the solution pH approached 4. The autoclave was flushed three times and then pressurized with methane to the desired pressure (2.0 MPa CH4, 99.999%). The reaction mixture was left at room temperature for 1 h. Tests were conducted under magnetic fields of 0, 50, 100, 150, 175, 200, 210, 250, 300, 400 and 500 G. The autoclave with obtained products was cooled in ice water for 20 min prior to analysis. Liquid products were quantified by 1H and 13C NMR spectroscopy. The gas-phase products are discharged through the reactor’s exhaust valve and collected in a gas bag, which is subsequently transferred to the GC for analysis.

Methane oxidation by Fe–LMS with additional CO and applying a magnetic field

The CH4 oxidation reaction was carried out in a closed high-pressure autoclave. The high-pressure reactor was placed between two parallel permanent magnets, and the intensity of the magnetic field was controlled by an external distance-adjustment device. About 100 µl catalyst, 19 ml deionized water and 1 ml H2O2 (30%) were added to the autoclave. Hydrochloric acid was added dropwise until the pH of the solution approached 4. The autoclave was flushed three times and then pressurized with CO to the desired pressure (0.1–1.0 MPa), and 2 MPa CH4 was then added into the reaction system. The reaction proceeded for 1 h at room temperature, under 0 or 500 G. The obtained products were cooled in ice water for 10 min prior to analysis. Liquid products were quantified by 1H and 13C NMR spectroscopy. Gas products were quantified by GC.

Reproducibility of liquid product conversion

First, 100 µl fresh catalyst, 20 ml deionized water and 5 ml H2O2 (30%) were added to the autoclave, which was flushed three times with deionized water and then pressurized with 2.0 MPa methane. Hydrochloric acid was added dropwise until the pH of the solution approached 4. The reaction was performed while switching the magnetic field on–off 11 times at 1-h intervals. The catalyst was washed with hydrochloric acid (pH 4) after each 1-h reaction to remove the oxidation film on the catalyst. The reactants and washed catalyst were then placed back into the autoclave for the next reaction. The obtained products were cooled in ice water for 10 min prior to analysis. Liquid products were quantified by 1H and 13C NMR spectroscopy.

Characterization methods

X-ray diffraction measurements were recorded on a Rigaku Miniflex-600 diffractometer using Cu Kα radiation (λ = 0.15406 nm) with a step size of 0.02° and a counting time of 0.5 s. Transmission electron microscopy images were recorded on a Hitachi H-7700 operated at 100 kV. Scanning electron microscopy images were recorded on a Supra 40. A Quantum Design MPMS3 was used for magnetic moment testing. Elemental analysis was performed by inductively coupled plasma atomic emission spectrometry using an Optima 7300 DV spectrometer. Liquid products were quantified by NMR spectroscopy. Measurements were conducted on a Bruker Avance-Ⅲ 400 spectrometer. 1H NMR spectra were recorded with a 2-s recycle delay, for 64 scans, using dimethyl sulfoxide as an internal standard. 13C NMR spectra were recorded with a 10-s recycle delay, for 2,048 scans. Gaseous products were quantified by a GC equipped with a 5-Å molecular sieve, a Porapak Q 80/100 mesh, and SE-30 and HP-Al2O3/S columns using helium (ultrahigh purity) as carrier gas.

In situ electron microscopy and corresponding atomic-level EDS mapping

Aberration-corrected HAADF-STEM images and corresponding EDS maps were recorded on a FEI-Titan Cubed Themis G2 300 STEM. Frozen sample rods were used to load samples, allowing for cooling with liquid nitrogen during testing. Before electron microscopy imaging, the samples were subjected to magnetic fields of 0 G and 500 G and frozen with liquid nitrogen for 10 min to fix the structure.

In situ X-ray 3D CT

In situ X-ray 3D CT was carried out at beamline BL07W of the National Synchrotron Radiation Laboratory. The sample holder, containing the nickel grid, was transferred to the chamber of a transmission soft X-ray microscope, where an elliptical capillary condenser focused the soft X-ray beam onto the cells for observation. In the sample chamber, the magnetic field is adjusted by controlling the distance between the natural magnet and the sample holder. For the generation of 3D volumes, the cells were rotated from −60° to +60°, capturing a continuous series of 121 projected images at 1° intervals with a 2-s exposure time. X-ray energies is 706 eV and 715 eV (covering the Fe L3 edge) were used. Alignment of the tilt series was performed using XMController, and 3D CT reconstruction was carried out using XMReconstruction.

In situ Mössbauer spectroscopy

Mössbauer spectroscopy measurements were conducted using a Wissel MR-2500 spectrometer. For the Fe–LMS sample under a 500-G magnetic field, the field strength at the sample location was adjusted by placing a natural magnet outside the measurement chamber. Each sample weighed 100 mg, and measurements were performed at room temperature in a vacuum environment. The spectral range was set to ±12 mm s−1, with a measurement duration of 24 h.

ESR

ESR was performed at the Steady High Magnetic Field Facilities, High Magnetic Field Laboratory, using the following parameters: temperature, 173 K; power, 0.01 mW; central field, 7,000 G; sweep width, 14,000 G; modulation frequency, 100 kHz; modulation amplitude, 2.00 G. The samples were frozen at 173 K for 10 min under magnetic fields of 0 G and 500 G, respectively, and the iron powder standard