The growing availability of low-cost and abundantly sourced natural gas leads to increased interest in its conversion to value-added chemicals. Natural gas is composed of dominantly small hydrocarbons with methane taking typically a volumetric fraction of about 70 to 90% [1–3]. Nowadays, great efforts have been conducted to convert methane into more useful chemicals like syngas, methanol, light olefins, aromatic compounds, etc. via direct or indirect routs [4]. The indirect rout involving the methane reforming and Fischer–Tropsch processes plays a crucial role in industry, as it affords one of the most important classes of chemicals—olefins [5]. However, such a 2-step conversion sequence wastes considerable part of methane molecules by unavoidably producing useless CO₂ and H₂O. By contrast, direct methane conversion shortens the reaction paths and utilizes more proportion of methane [6–7]. In general, there are 2 major routes for the direct conversion of methane to light olefins, i.e., oxidative coupling of methane (OCM) and nonoxidative coupling of methane (NOCM) [4]. The OCM process uses oxidant to overcome the thermodynamic restrictions and make the reaction exothermic [8]. However, the by-products like CO₂ and H₂O are still unavoidable, decreasing the atom economy.

Since 1990s, numerous efforts have been made to produce hydrocarbons through NOCM processes and promising progress achieved recently [9–11]. For example, a Pt–Bi bimetallic catalyst was reported to selectively convert methane to ethane with high carbon selectivity (>90%) and typical methane conversion of ca. 2% at 600 to 700 °C [9]. The Fe/SiO₂ catalyst is capable of 48% methane conversion via the NOCM at 950 °C, producing ethylene, naphthalene, and benzene with a selectivity of 53%, 25%, and 22%, respectively [12]. Similar products over a Pt–Sn catalyst at 700 °C were observed, whereas the methane conversion was lower than 0.3% [13]. In addition, in NOCM, methane conversion may proceed at low temperatures: 6 wt% of Pt/SiO₂ catalyst was reported to convert methane to hydrogen and C₂H₆ continuously at 250 °C [14]. Here, obviously, more efforts are required in the production of more valuable hydrocarbons like olefins via NOCM at low temperatures.

In 1974, it was reported that tantalum can convert methane or acetylene to carbon and hydrogen at 1,500 to 2,300 °C [15], kicking off the story of methane activation at the tantalum center. Later in 1991, the gas-phase activation of methane by Ta²⁺ ions was reported [16], followed by a series of gas-phase studies on methane activation by tantalum-involved ions [17]. A series of ions like [TaO]⁺ [18], [TaO₂]⁺ [19], [TaO₃]⁺ [20], [TaN]⁺ [21], and [TaCO₄]⁺ [22] are able to break the H₃C–H bond in the gas phase under ambient conditions. Here, the relativistic effects induced strong Ta–C interaction matters [23–25]. The excellent performance of Ta in gas-phase methane activation systems was later applied in the condensed phase: By building a H–Ta–O₂ active site on silica, the catalyst converts methane to ethane with 98% selectivity at temperatures below 500 °C, although the methane conversion was below 0.5% [26]. The current work was also inspired by previous publications on gas-phase methane activation as mediated by Ta-involved species.

Notably, although Ta–O is an ideal center for methane activation, further conversion of the strong Ta–C bond may encounter high energy barrier [19–20]. To weaken the so-formed Ta–C bond, an electron-rich Ta center is required. This is supported by previous report on highly efficient thermal activation of methane by [TaN]⁺ in the gas phase [21]. Thus, in general, a Ta–N unit is preferable for methane conversion, and a Ta–N₄ center may possibly be built in the condensed phase. Here, we report a catalyst with Ta–N₄ center as anchored in phthalocyanine that is supported by graphitic carbon nitride (g-C₃N₄). Such a catalyst is able to convert methane selectively to propylene at a relatively low temperature (350 °C).

As the participation of oxygen in the transformation of methane is not wanted, we decided to construct an active Ta–N_x center on the carbon-based materials. Thus, inspired by the synthesis method to prepare single atom catalyst [27–33]. Quite a few organonitrogen species were used to prepare Ta-complex precursors for further pyrolysis on carbon-based materials. However, the attempts with ethylenediamine, 2-methylimidazole, and 1,10-phenanthroline failed to produce an active catalyst. This was attributed to the lack of regulated coordination pattern all through the preparation of the supported catalyst. Thus, we tried to construct a macrocyclic ligand anchored Ta center, which was further supported on carbon nitride.

The structure of tantalum phthalocyanine (TaPc) with a Ta–N₄ center was first reported in 1988 [34], and it has shown promising in electrocatalysis, particularly the oxygen reduction reaction [24,35–38]. TaPc was thus selected as the precursor complex, which was prepared according to previous reports [34]. In fact, we first tried thermocatalytic conversion of methane with TaPc only (confirmed by matrix-assisted laser desorption ionization mass spectrometry method; Fig. S1). As a result, trace product was observed at a temperature up to 450 °C. Considering that the TaPc molecules may stack upon crystallization [39–42], the unsatisfying performance of TaPc was attributed to low dispersity. A proper support is thus necessary to anchor each Ta–N₄ center separately.

g-C₃N₄ has proven to be an ideal catalytic material for various chemical processes [43]. With the following advantages, g-C₃N₄ was selected as the active center carrier in NOCM: (a) It is usually prepared above 500 °C through tubular furnace with high thermostability; (b) g-C₃N₄ is of 2-dimensional structure and high electron mobility, and it can support a structure like TaPc stably via π–π stacking [44–45]; and (c) the nitrogen atoms of heptazine rings and the structural cavities may serve as ideal coordination sites to stabilize a metal center like Ta [46–47].

After a series of attempts to support TaPc well on g-C₃N₄, the controlled sequential impregnation and pyrolysis procedures were found optimal. Through activation under hydrogen flow, the so-prepared TaPc/C₃N₄ catalysts exhibit surprisingly high activity toward methane conversion (Fig. 1). Moreover, highly selective generation of propylene was observed, and ethylene was produced as well. The performances of different TaPc/C₃N₄ catalysts are shown in Fig. 2. As shown in Fig. 2B, the higher the Ta content is, the higher activity the TaPc/C₃N₄ catalysts give. On the other hand, however, the yield of propylene is not increasing proportionally with the content of Ta, resulting in lower turnover frequency (TOF) values with over high Ta content. Thus, 0.08 wt% of Ta sample gives the highest TOF with the valve of 0.99 s⁻¹ at 350 °C (Fig. 2B and Fig. S9B). At this condition, the selectivity of propylene is up to 96.0%, corresponding to 4.0% ethylene (Fig. 2A). Further, in the lifetime test, 0.08 wt% of Ta sample affords a long single-run lifetime of >300 h at 350 °C with stable conversion of methane (Fig. 2C); after reactivation, it still lasts for >120 h. Although higher TOF was obtained when performing the reaction at 450 °C, the lifetime of the catalyst is shortened dramatically (Fig. 2C). More technical details are provided in the Supplementary Materials (Fig. S9).

As there is no previous report on the production of propylene via NOCM, the products have been verified carefully: The produced hydrogen was detected through online gas chromatography (Fig. S2); the produced propylene and ethylene was first confirmed using gas chromatography upon comparison with standard samples, and further evidence was provided by gas chromatography-mass spectrometry (Fig. S3).

Further, the in situ diffuse reflectance infrared Fourier transform spectroscopy analysis was performed to obtain even more information regarding the methane conversion processes (Fig. 3). The activated catalyst with adsorbed methane was used as the background. Upon elevating the temperature to 300 °C, the intensified negative peaks of 3,016 and 1,304 cm⁻¹ represent the stretching vibration and in-plane bending vibration of methyl C–H bond, indicative for the desorption of methane. When achieving 350 °C, the positive peak of 3,003 cm⁻¹ arises, corresponding to the methyl stretching vibration. In addition, the positive peak of 1,541 cm⁻¹ is indicative for the stretching vibration of the –C=C– fragment. Without any signal assigned to methylene, combining with the gas chromatography-mass spectrometry result, the product was verified as propylene. After running at 400 °C for 2 h, the temperature is decreased to 30 °C, and only negative peak of methane remains, indicating no deposition of heavy products. More details are shown in Fig. S4.

According to the Gibbs–Helmholtz formula, the thermodynamic limit of this work is evaluated. As shown in Fig. 4, considering as coupling consecutive reaction, the NOCM are wildly influenced by the temperature, and it seems that high conversion of methane can be obtained at high temperature, although it should be noted that looping and deposition are not considered. At lower temperature, it shows that the experimentally observed conversion almost approaches its thermodynamic limit (Fig. 4B). Experiments under a temperature of >500 °C was not performed as the catalyst already loses its stability at 500 °C. In addition, Fig. 4 clarified that the presence of inert gas does not affect the equilibrium constant but the equilibrium composition. With the same total pressure, inert gas actually plays a dilution role in the system. According to the Le Chatelier’s principle, with positive stoichiometric coefficient, the NOCM process is favored with higher conversion. In addition, the presence of the inert gas allows for deeper condensation of the system, thus facilitating the production of propylene. Alternatively, the possibility for NOCM proceeding as parallel processes was also considered, while inconsistence was found in kinetic modeling (see Figs. S7 and S8 and Tables S1 to S5 for more details).

To obtain more structural information of the active catalyst, the high-angle annular dark-field transmission electron microscope (HADDF-TEM) analysis was performed on the TaPc/C₃N₄ (0.08 wt%) sample. As shown in Fig. 5A, the bright points in the image indicate atomically dispersed Ta atoms on the surface. Further, the corresponding energy-dispersive spectroscopy maps of the TaPc/C₃N₄ heterojunction (Fig. 5B and Fig. S2) reveals the homogeneous distribution of Ta, C, and N across the nanosheet structure, which unambiguously demonstrates that TaPc are uniformly distributed on g-C₃N₄ surfaces. In addition, we also tried to perform the extended x-ray absorption fine structure or x-ray absorption near-edge structure analysis on the TaPc/C₃N₄ (0.08 wt%) sample; unfortunately, however, no useful information was obtained because of the rather low Ta content.

The thermal stabilities of different components of the catalysts were examined with thermogravimetric analyzer (TGA) analysis (Fig. 5C). The g-C₃N₄ material obtained via pyrolysis at 550 °C is thermally stable until 520 °C and completely decomposes at 750 °C. The TaPc complex undergoes a mass loss of 8.2% at 320 °C and is then kept stable until 550 °C; further decomposition occurs at 640 °C, leaving 11.52% ash eventually. The TaPc/C₃N₄ samples prepared at 400 and 500 °C have a good thermal stability below 500 °C, but the one prepared at 600 °C starts to lose mass at around 420 °C, probably because the backbone structure of the samples had already been destroyed during the preparation process. Considering the performances of different components of the TaPc/C₃N₄ sample under TGA, most likely, the impregnation–pyrolysis (<500°C) sequence retains the Ta–N₄ structure of TaPc on the g-C₃N₄ surface.

The x-rays diffraction patterns show that the typical g-C₃N₄ diffraction peak at 27.5° offsets to 21.3° upon stacking the carbon nitride plane to form the TaPc/C₃N₄ (0.08 wt%) catalyst [48]. This means that the interplanar spacing increases by inserting TaPc (Fig. 4D). After the catalyst is deactivated, the peak at 21.3° shifts to 22.3° and is less sharp, indicative for a decrease in the interplanar spacing of g-C₃N₄. Furthermore, there appear new diffraction peaks in the deactivated catalyst that are identified as C₃N₄ (35.6°, 57.0°, and 62.6°) (00-053-0671) [49] and TaN_0.83 (30.8° and 35.3°) (01-089-4765) [50], respectively. Thus, most likely, the deactivation of the catalyst results from the aggregation of dispersed Ta species to small TaN clusters.

In addition, the electronic morphology of the TaPc/C₃N₄ (0.08 wt%) catalyst was probed with x-ray photoelectron spectroscopy analysis. As shown in Fig. 5E, high-resolution C 1s scan shows the evidence of C–C bonding (284.2 eV) and C–N bonding (287.3 eV) and with proportions of 36.3% and 63.7%, respectively; high-resolution N 1s scan shows the pyridinic N (397.9 eV; 54.8%) and pyrrolic N (399.0 eV; 45.2%), respectively. Compared to TaPc, the TaPc/C₃N₄ (0.08 wt%) spectrum has a red shift for Ta, which means that the electron density of Ta is increased, indicative for unstacked TaPc species.

It should be noted that after the catalyst is reactivated, the initial activity and lifetime (Fig. 2B) cannot be recovered, implying the irreversible conversion of the Ta–N₄ units to the TaN clusters. Further, in the Raman spectrum of the deactivated catalyst, there is no peak at 1,300 and 1,600 cm⁻¹ indicative for the D₁ and G bands of aromatic carbon (Fig. 6A), indicating that coking does not occur. Identically, in the TEM images, the catalyst is still a thin slice with little crimp (Fig. 6B). However, further through the HADDF-TEM images, it is obvious that the metal atoms are gathered with each other (Fig. 6C and D), which bearing out the formation of TaN clusters.

Further, quantum chemical calculations were performed to probe the origins of the excellent performance of the TaPc/C₃N₄ catalyst. Considering the preparation procedures and the characteristic results, the macrocyclic structure of TaPc is probably maintained, and a model with TaPc(C₃₂N₈H₁₆Ta) supported on C₃N₄ (C₉₀N₁₂₃H₁₅) via π–π stacking was thus built (Fig. 1 and Fig. S10). To perform feasible calculation on such a large model, the semiempirical extended tight-binding computation method [51,52] as developed by Grimme group was used in conjunction with the gau_xtb code [53]. Here, we focused on why ethylene and propylene correspond to the major product.

According to the calculation, the most stable structure for the dehydrogenation of the first CH₄ molecule correspond to a bridge N–CH₂–Ta structure (intermediate 1, Fig. 7A). Starting from this intermediate, the conversion of another 2 CH₄ molecules proceeds similarly. For example, as shown in Fig. 7B, activation of the second CH₄ occurs via insertion of Ta into the CH₃–H bond, generating 3; this is followed by migration of the methyl group from Ta to the CH₂ unit of the first CH₄ (3 → 4); next, the so-formed ethyl group delivers a hydrogen to the hydride ligand of Ta to extrude molecular hydrogen (4 → 5), affording the N–CH(CH₃)–Ta structure. The conversion of the third CH₄ precisely resembles the 2 → → 5 sequence, thus proceeding along the 6 → → 9 path. Meanwhile, the formation of the N–CH(CH₃)–Ta and N–C(CH₃)₂–Ta structures may compete with alternative channels. As shown in Fig. 7B and C, via a series of transformation, the C₂H₄, C₂H₆, C₃H₆, and C₃H₈ molecules can be released from the reaction system via the sequences 2 → → 10/11 and 6 → → 12/13, respectively (for intact processes, see Figs. S11 to S13); note that although some of these processes are energetically more favorable as compared to the generation of N–CH(CH₃)–Ta and N–C(CH₃)₂–Ta, all of them are entropically much less favorable as compared to the short paths 2 → → 5 and 6 → → 9. As a result, most likely, the N–CH(CH₃)–Ta and N–C(CH₃)₂–Ta structures are the major intermediates the system goes through. However, these 2 structures may undergo further isomerization to release neutral olefin molecules: By shifting a hydrogen from the methyl group to the methylene carbon, the intact C₂H₄ or C₃H₆ is generated (2 → → 14 and 6 → → 15). Obviously, such 2 isomerization processes are highly competitive. As to the activation of a fourth CH₄ molecule, although it is kinetically feasible at the structure 16, the lack of hydrogen in the bridging carbon prevents the conversion of the fourth methane from a sequential C–C coupling and H₂ elimination process. Consequently, the propagation of the carbon chain stops at a number 3, and C₃H₆ is released as a major product. More details are provided in the Supplementary Materials. Although we tried to calculate as many as possible pathways, there might exist other various routes; after all, the generation of C₂ and C₃ involves multiple substrates as well as a series of bond breaking and making processes.

In summary, a catalyst with single Ta–N₄ center anchored in phthalocyanine and supported on g-C₃N₄ was prepared. Such a catalyst is able to convert methane to ethylene and propylene at a relatively low temperature (350 °C) with the latter as the major product. A TOF value of 0.99 s⁻¹ and >300-h one-pass lifetime prove the robustness of the catalyst. The deactivation results from the transformation of the Ta–N₄ units to TaN clusters during the reaction process. Quantum chemical calculations indicate that the bridge N–CR₂–Ta (R = H and CH₃) structures serve as the key intermediates, which either enable the carbon-chain propagation or, alternatively, isomerize to release olefin molecules. Considering the vital role the Ta atom plays in the transformation that it serves as the courier to deliver both carbon and hydrogen, most likely, here, the relativistic effects induced strong Ta–C/H interaction functions once again. Future efforts may focus on how to uniformly load higher content of well-dispersed Ta on a carbon-based material and how to enhance the chemical stability of the Ta–N₄ structure. Further, the relativistic effects exert similar influence on the gas and condensed phases, which encourages us to continue the gas-phase guided construction of high-performance catalyst. Indeed, gas-phase studies allow us to correlate various structural/electronic features with the performance of the active center, while the major difficulty is still on how these favorable features are replicated in bulk systems. As to the direct conversion of methane to higher hydrocarbons, most likely, a 5d element-centered structure is necessary so as to bind the carbonide intermediates strongly for further propagation of the carbon chain.