CoOx Nano-clusters anchored onto 10-membered ring of Silicate-1 boost propane dehydrogenation Authors: Ruiqiang Liu, Bing Ma, Linlong Zhang, Jingqing Tian, Chen Zhao Published: 2025 ABSTRACT AbstractCobalt-zeolites catalysts for propane dehydrogenation (PDH) suffer from low activities on the confined Co sites in zeolites or poor stabilities on the CoOx species in presence of hydrogen. In this work, we synthesize a unique Co structure with (≡Si-O)2Co(≡Si-O)2…CoO clusters (smaller than 2 nm) onto desilicated Silicate-1 via impregnation-dry gel transformation approach. This structure leads to a high C3H6 formation rate of 34 mmol gcat−1 h−1 as well as a low deactivation rate of 0.0116 h−1 at specific conditions (100% propane, 550 °C, 5.0 h−1, 72 h), outperforming currently reported Co-based catalysts in terms of maximum activity and stability. The integrated differential phase-contrast scanning transmission electron microscopy (iDPC-STEM) and extended X-ray absorption fine structure (EXAFS) measurements confirm that CoO clusters bond with the O atoms adjacent to the tetrahedrally coordinated Co single sites anchored on the 10-membered ring of Silicate-1. The electron-negativity property of ((≡Si-O)2Co(≡Si-O)2…CoO) species facilitates the adsorption of H* from C3H8 and the following dehydrogenation steps, and thus enhancing the overall activity, while its lower formation energy ensures the higher catalyst stability even in presence of hydrogen atmosphere. The designed Co-based catalyst offers great potential for taking the combined advantages of Co singles atoms as well as CoOx clusters for high-performance PDH.Subject terms: Heterogeneous catalysis, Chemical physics, Catalyst synthesis, Catalytic mechanisms Introduction Propylene is a significantly important product of the petrochemical industry. The direct propane dehydrogenation to propylene (PDH) has gained substantial interest owing to the high propylene selectivity1–4. The primary catalysts for PDH are focused on PtSn/Al2O3 (Oleflex) and Cr/Al2O3 (Catofin) catalysts5,6. However, the natural scarcity of Pt and the environmental toxicity associated with Cr centers have driven the search for alternative catalysts7–9. Recently, cobalt-based catalysts have shown significant potential for alkane dehydrogenation owing to their low cost, non-toxicity, and efficient C–H activation abilities10,11. Co species can be encapsulated in zeolite frameworks with an ordered microporous structure and high thermal stability. These Co species exist either as single-atom sites (≡Si-O)4Co within the zeolite framework, or as CoOx nanoclusters confined within the channels12,13. However, single-atom Co sites embedded in zeolite skeleton typically suffer from low PDH conversion rates, despite their excellent stability. This issue can be mainly attributed to the fact that Co (II) in the form of Si–O–Co is present in the TO4 tetrahedral structural units of the zeolite framework, and therefore, the surrounding four Si–O bonds stabilize single-atom Co sites, thereby preventing agglomeration at high temperatures. However, these strong interactions also weaken the capability of isolated Co to activate the C–H bond of C3H8. For example, Hu et al.14 synthesized Co-MFI catalysts using Co protected by ethylenediamine ligands. This catalyst consisted of two Co–O–Si bonds and pseudo-bridging hydroxyls (Co···OH–Si) located in MFI zeolites with a distorted tetrahedral structure [(≡SiO)2 Co (HO–Si≡)2] to provide PDH activity. This catalyst showed ca. 40% propane conversion and a propylene formation rate of 7.5 mmol gcat−1 h−1 at 580 °C, which remained stable for over six regeneration cycles. Wu et al.15 prepared a single-atom Co/SiO2 catalyst with coordinative unsaturated Co–O3 sites as the active sites of the PDH reaction, resulting in a low C3H8 conversion of 24%, with a 15 mmol gcat−1 h−1 propylene formation rate and a long lifetime of 800 h. Nanoclusters of CoO encapsulated in the micropore channel displayed a high C3H8 conversion rate, although with low C3H6 selectivity and high deactivation rates. These features were produced by the reduction of untrammeled CoO nanoclusters to Co0 in the presence of a hydrogen atmosphere, resulting in the breakage of C–C bonds and carbon deposition. Liu et al.16 synthesized CoOx@MFI catalysts via an in-situ hydrothermal method, with the CoO (0.4–0.6 nm) nanoclusters located in the sinusoidal channels of the MFI zeolite being highly active. Nevertheless, Co0 species were detected after 8 h of reduction in hydrogen at 600 °C, followed by a decrease in propane conversion from 59 to 36%. Having this in mind, the rational design and fabrication of zeolite-encapsulated cobalt species catalysts with high activity and propylene selectivity, as well as superior stability, remains highly desirable for academic and industrial applications. In this work, we synthesized distinctive PDH active sites containing ultra-small CoO nanoclusters anchored around the single-atom Co site onto a Silicate-1 10-membered ring. This catalyst showed ultra-high C3H6 selectivity (99%) while maintaining a high production rate of 34 mmol gcat−1 h−1 and a low deactivation rate (0.0116 h−1) under conditions (100% propane, 550 °C, 5.0 h−1, 72 h), significantly outperforming previously reported Co-based catalysts. This new structure and the intrinsic mechanism concerning the C3H8 dehydrogenation on the surface of the catalyst are investigated by advanced measurements such as X-ray absorption spectroscopy (XAS) measurements, high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and the integrated differential phase-contrast scanning transmission electron microscopy (iDPC-STEM). Results Synthesis and characterization of Co@S-1 For the synthesis of Co@S-1 catalyst, the Co species was encapsulated into the silicalite-1 (S-1) zeolite through a synthesis process involving alkaline treatment, Co2+ impregnation and dry-gel transformation, as shown in Fig. 1a. Initially, a rich Si–O− microenvironment was generated by treating S-1 seeds with an alkaline solution, followed by the adsorption of Co2+ onto Si–O− surface and dry-gel conversion. Subsequently, 1.6% Co@S-1 (the Co content determined by inductively coupled plasma, ICP) was formed after air-calcination. The X-ray diffraction (XRD) pattern of 1.6%Co@S-1 exhibited the typical characteristic diffraction peaks of the MFI framework at 8.2, 9.0, 23.3, 24.1, and 24.6°17 (JCPDS No. 44-0696) (Supplementary Fig. 1a). No diffraction peaks corresponding to Co species were observed, which was assigned to the highly dispersed Co species in the MFI zeolite. As illustrated in Supplementary Fig. 1b and Table S1, the Brunauer–Emmett–Teller (BET) surface areas decreased with increasing Co content in the Co@S-1 catalysts (353–308 m2 g−1). Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images revealed a regular hexagonal structure with a homogeneous dimension of approximately 270 nm (Fig. 1b, c and Supplementary Fig. 2). High-resolution transmission electron microscopy (HRTEM) images clearly indicated the absence of Co clusters or particles, with the lattice fringe spacing (11.0 Å) being consistent with that expected for MFI-type zeolites, thus verifying high crystallinity (Fig. 1d and Supplementary Fig. 3). High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and energy-dispersive X-ray spectroscopy (EDS) elemental mapping revealed that the Co species were highly dispersed (Fig. 1e and Supplementary Figs. 4 and 5). Structural characterizations by electron microscopy. a Schematic illustration showing the preparation process of CoOx@Silicalite-1; b SEM, c TEM, d HRTEM, and e HAADF-STEM images and corresponding O, Si, and Co elemental mappings images observations of 1.6%Co@S-1. Due to the difficulty in distinguishing Co from Si atoms in zeolite frameworks from HAADF-STEM images under low electron dose conditions, the integrated differential phase-contrast scanning transmission electron microscopy (iDPC-STEM) technique was employed for atomic-scale identification. As shown in Fig. 2a–c, no nanoparticles were observed along the [010] crystal plane of MFI zeolite. The enlarged iDPC-STEM images (Fig. 2d) revealed that the 10-member rings (10MR) of MFI zeolite exhibited a twisted appearance, indicative of the influence of cobalt doping on the zeolite lattice. The single-atom Co (Si–O–Co, pink circles) species located on the S-1 zeolite skeleton can be observed, while the presence of CoOx clusters (green circles) was also detected in the 10-member rings (10-MR) (Fig. 2e–j). Thus, the presence of both Si–O–Co species and CoOx clusters is confirmed in the 1.6% Co@S-1 catalyst. Microscopic structure characterizations by iDPC-STEM. a–e the corresponding iDPC-STEM images of the 1.6% Co@S-1 catalyst, f–k zoomed-in areas and the corresponding Co–O–Si and CoOx structural views in (e). Structure analysis of the active component in Co@S-1 The current active sites for the Co-zeolite catalysts in the PDH reaction included framework isolated Co–(O–Si)4 and CoOx nanoclusters encapsulated in the zeolite channels11,14,16,18,19. The 1.6% Co@S-1 sample was then treated with an acidic aqueous solution, resulting in the complete removal of ultra-small CoOx clusters, while retaining only single Co atoms in S-1 (0.3% Co@S-1, with Co content confirmed by ICP) as a comparison. To investigate the presence of specific properties of Co species in the zeolite before and after acid washing, the 1.6% Co@S-1 and 0.3% Co@S-1 catalysts were characterized by infrared (IR) spectroscopy of adsorbed CD3CN, Raman spectroscopy, hydrogen temperature-programmed reduction (H2-TPR), ultraviolet-visible spectroscopy (UV–vis), X-ray photoelectron spectroscopy (XPS), X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS). FT-IR spectroscopy with CD3CN adsorption represents valuable information about the Co species present in the S-1 zeolite20. As a result, the adsorbed CD3CN molecule exhibited no diffusion restriction in the MFI micropore channel, enabling sufficient interaction with the hydroxyl groups and altering the stretching pattern to form a well-defined symmetry band10. The IR spectra of adsorbed CD3CN on 1.6% Co@S-1 showed three bands at 2308, 2273, and 2327 cm−1, which were ascribed to adsorption on the Lewis acid sites induced by the atomically dispersed framework Co, silanols and CoOx nanoclusters, respectively (Fig. 3a and Supplementary Fig. 6). The band at 2308 cm−1 is assigned to closed Lewis acid site, corresponding to the Co(–O–Si)4 in the MFI framework; while the weak absorption peak at 2327 cm−1 is ascribed to the open Co site, i.e., the nanoclusters of CoOx14,21. However, the 0.3% Co@S-1 sample did not show the open Co Lewis sites (i.e., band at 2327 cm−1 attributed to the sub-nanoclusters of CoO, essentially disappeared, see Fig. 3a and Supplementary Fig. 7), while a significant reduction in the twisting and stretching vibrations signals of the Co–O bonding in the CoO clusters was observed at 292 and 479 cm−1 in the Raman spectra (Fig. 3b). The spectroscopic characterizations of the Co@S-1 catalyst structure. a FTIR spectra of CNCD3 adsorption, b Raman, and c Co 2p XPS spectra of 0.3% Co@S-1 and 1.6%Co@S-1; d Normalized Co K-edge XANES spectra, e k3-weighted Fourier transform EXAFS spectra in the R-space of the 1.6% Co@S-1, 0.3% Co@S-1, and reference samples; f Fourier transform EXAFS spectra and the corresponding curve-fitting results of the 0.3% Co@S-1 and 1.6% Co@S-1 samples; g–i wavelet transform plots of the Co foil, Co3O4 and 1.6% Co@S-1 catalysts, respectively. The reduction peak in the H2-TPR profile for the 1.6% Co@S-1 catalyst was observed at 713 °C, with a hydrogen consumption of 198.78 μmol g−1 (Supplementary Fig. 8a, b and Table S2). This was probably attributed to the reduction of the CoOx species embedded in S-1 zeolite with strong metal–support interaction. However, the 0.3% Co@S-1 catalyst did not show any virtual reduction peak in the tested temperature range up to 850 °C. This suggests that different kinds of Co species exist in the two Co-based catalysts19,22. The coordination environment of the Co species in the zeolites was investigated by UV–vis spectroscopy in a subsequent step. Three prominent peaks at wavelengths of 653, 582, and 529 nm were observed for the 1.6% Co@S-1 and 0.3% Co@S-1 catalysts, corresponding to the 4A2 → 4T2, 4A2 → 4T1(4F), and 4A2 → 4T1(4P) transitions of the Co2+ species, respectively (Supplementary Fig. 8c). These results were consistent with the zeolite tetrahedral geometry (Td) Co species reported in the literature11,14. However, the band at 378 nm caused by Td CoO distortions confined in the zeolite framework was observed exclusively for the 1.6% Co@S-1 sample13,23. In addition, according to HAADF-STEM and EDS elemental mapping, the Co concentration decreased after acid washing, which was dispersed uniformly in the S-1 zeolite matrix, while the crystallinity was slightly reduced, in line with the XRD patterns (Supplementary Figs. 9 and 10). Therefore, we concluded that the 0.3% Co@S-1 catalyst primarily exhibited a skeletal tetra-coordinated Co–(O–Si)4 structure. Conversely, Co–(O–Si)4 and CoOx configurations were both observed in the 1.6% Co@S-1 catalyst with certain interactions. This is also consistent with the iDPC-STEM results. XPS was employed to analyze the surface components and their chemical states in the prepared catalysts (Fig. 3c). Core-level binding energy signals corresponding to Co 2p3/2 and Co 2p1/2 were observed at 781.6 and 797.5 eV, respectively, indicating the presence of Co2+ species in the 0.3% Co@S-1 and 1.6% Co@S-1 catalysts12,14,24. In contrast, when Co was directly loaded onto the external surface of the S-1 zeolite to obtain the 1.9% Co/S-1 catalyst, Co3+ was detected at 780.0 and 796.4 eV and Co2+ was detected at 786.4 and 803.5 eV (Supplementary Fig. 11a), indicating the presence of bulk Co3O4 on the external surface of the S-1 zeolite13,25. These results were also consistent with H2-TPR, Raman and TEM results (Supplementary Figs. 12 and 13), implying that Co2+ species in 0.3% Co@S-1 and 1.6% Co@S-1 are quite stable for resistance to oxidation upon calcination at 550 °C in air. To identify the coordination environment and atomic structures of the CoO clusters and Si–O–Co sites in the S-1 zeolite, XANES and EXAFS were performed on the 0.3% Co@S-1 and 1.6% Co@S-1 catalysts. Reference samples of Co3O4, CoO, and Co foils, representing the Co–O–Co and Co–Co bonds, respectively, are shown in Fig. 3d–f. The Co absorption edge positions of 0.3% Co@S-1 and 1.6% Co@S-1 were identified between those of CoO and Co₃O₄, with a closer alignment to CoO. Additionally, a pre-edge feature of Co (II) was observed at 7709.8 eV, corresponding to the 1 s → 3 d excitation of Co (low spin), in line with the results reported for the isolated Co (II) atoms in Co/SiO2 and also supported by the XPS results11,26. In addition, the absorption edge of the 1.6% Co@S-1 sample exhibited a slight shift compared to the 0.3% Co@S-1 sample. This shift was attributed to the interaction of CoO clusters with Si–O–Co27. The Fourier-transformed (FT) k3-weighted K-space and R-space EXAFS spectra of the catalysts are shown in Supplementary Fig. 14 and Fig. 2e, respectively. In comparison to 0.3% Co@S-1 with only the Co–O band locating at 1.6 Å, referring to the existence of single Co atoms (Co–O–Si), 1.6% Co@S-1 shows two peak positions at 1.6 and 2.7 Å, corresponding to typical first shell Co–O and second shell Co–O–Co coordination environments14,16. This suggested that the 1.6% Co@S-1 catalyst contained both CoOx sub-nanoclusters and isolated Co–(O–Si)4 structures. The structural parameters and coordination numbers of the Co species in 0.3% Co@S-1 and 1.6% Co@S-1 were further analyzed by the Artemis software package fitting, and the results are compiled in Fig. 3f and Supplementary Table S3. The EXAFS fitting curves in R-space revealed a high degree of correlation with the theoretically predicted structure (Fig. 3f and Supplementary Fig. 15). The 0.3% Co@S-1 catalyst contained Co–O bonds with a coordination number of 4.04 ± 0.23 and a low bond distance of 1.80 ± 0.04 Å (corrected distance), and no Co–O–Co apparent peak was observed. This indicated that single Co atoms are coordinated with four O atoms to form the Co–O4 unit of the zeolite frameworks15. However, the coordination numbers observed for the Co–O (first shell) and Co–O–Co (second shell) scattering paths in the 1.6% Co@S-1 sample were 4.0 and 4.1, respectively, with corresponding bond distances being 2.00 ± 0.01 and 2.85 ± 0.01 Å (corrected distances). These results directly corroborated the coexistence of ultra-small CoOx nanoclusters and single-atomic Co–(O–Si)4 sites, along with the robust coordination interactions observed in the 1.6% Co@S-1 sample28. To obtain further insight into the coordination environment of the Co atoms, EXAFS wavelet transform (WT) analysis was conducted to determine the backscattered atoms in the R-space, thus providing the radial distance and K-space resolution (Fig. 3g–i and Supplementary Fig. 16). As shown in Fig. 3i, two intensity maxima were identified in the 1.6% Co@S-1 contour plot, which could be ascribed to the Co–O and Co–O–Co bonds, respectively. In contrast, only one intensity maximum ascribed to the Co–O bond was observed in the EXAFS WT contour plots of 0.3% Co@S-1. There are five different sites for the CoOx cluster rivets onto the 10-MR of MFI zeolite in the 1.6% Co@S-1 sample, but XAFS cannot distinguish the accurate structure with the location of the CoOx cluster onto the MFI 10-MR. Thus, the DFT approach is employed to explore the possible positions. Five symmetric models are supposed to explore CoOx–MFI cluster configurations (Fig. 4). The proposed M5 model exhibits the lowest formation energy (−502.34 eV), and the calculated distances of Co–O and Co–O–Co layers (1.96 and 2.80 Å) in the M5 model well match those data in the experimental EXAFS (2.00 ± 0.01 and 2.85 ± 0.01 Å). This model involves CoOx sub-nanoclusters within the 10-MR of the MFI zeolite and tetrahedrally coordinated single sites Co on the 10-MR to form a ((≡Si-O)2Co(≡Si-O)2…CoO) structure in the 1.6% Co@S-1 catalyst. Proposed models with corresponding structures of 1.6%Co@S-1 catalyst by DFT calculations. where the Co–O and Co–O–Co layers in the M1–M5 models represent the average distance. The 0.3% Co@S-1 catalyst was observed to contain only Co–O bond (bond length of 1.80 Å) in general conformity with the DFT calculation model (average distances of 1.76 Å) (Supplementary Fig. 17 and Table S3). Additionally, the formation energy of the (≡Si–O)2Co(≡Si–O)2…CoO clusters structure in the 1.6% Co@S-1 catalyst was −502.3 eV, lower than the −444.8 eV value for the tetrahedral Co–(O–Si)4 structure in the 0.3% Co@S-1 catalyst, indicating the stability of this structure (Supplementary Table S4). In conclusion, CoO sub-nanoclusters were anchored within the 10-MR of the MFI zeolite, together with tetrahedrally coordinated single sites Co on the 10-MR to form ((≡Si–O)2Co(≡Si–O)2…CoO) structures in the 1.6% Co@S-1 catalyst. PDH performance over Co@S-1 To increase the content of Si–O− in the S-1 seeds, a range of organic bases was used for desilication. As shown in Supplementary Figs. 18 and 19, catalysts with a tetrapropylammonium hydroxide (TPAOH) treatment exhibited optimal propane conversion and relatively higher propylene selectivity. Further optimization of the treatment conditions demonstrated that catalysts prepared with 0.3 mol L−1 of TPAOH at 70 °C for 4 h exhibited excellent C3H8 conversion and C3H6 selectivity (Supplementary Fig. 20). Additionally, an increase in Co loading from 0.45 to 1.6% resulted in higher in propane conversion, propylene selectivity, and formation rates (Supplementary Fig. 21). This indicated that the single atomic Co and CoO configurations encapsulated in the S-1 zeolite framework potentially demonstrated enhanced catalytic performance in the PDH reaction. However, the formation rates of C3H6 showed a marked decline with an increase in Co loading beyond 1.6%. This was probably because the Co species with elevated contents, predominantly in the form of CoOx nanoparticles (Supplementary Figs. 22 and 23), showed weak interaction with the S-1 zeolite, resulting in a decrease in activity at the highly reductive hydrogen conditions. Using the 5% C3H8/N2 mixture as feed, the initial C3H8 conversion and C3H6 selectivity were 75.5 and 92.7%, respectively, at the weight hourly space velocity (WHSV) of 1.2 h−1 (Fig. 5a and Supplementary Fig. 24). Then increasing the propane concentration to 50% and 100% resulted in initial conversion of 49.0% and 38.7%, respectively. Although the initial C3H6 selectivity decreased to 84.8% (50% C3H8/N2) and 77.6% (pure C3H8), both were recovered to above 90% after 5 h reaction time. At conditions with 50% C3H8/N2 mixture and WHSV of 1.2 h−1, the propane conversion (23.0–59.8%) and propylene formation rates (6.1–12.4 mmol gcat−1 h−1) on 1.6% Co@S-1 catalyst increased with working temperatures from 500 to 600 °C, reaching the equilibrium concentration (Fig. 5b). The propylene selectivity was slightly decreased with increasing working temperatures, and this was mainly caused by propane/propylene cracking at higher temperatures (Supplementary Figs. 25 and 26). In addition, WHSV was further increased to 5 and 10 h−1 at 50% C3H8 and 600 °C, resulting in initial C3H6 formation rates of 53.1 and 97.5 mmol gcat−1 h−1, respectively, while the propylene selectivity was maintained at 96.0% after reaction stabilization (Fig. 5c and Supplementary Fig. 27). The calculated carbon balance attained 99.6–103.1% at tested conditions (Supplementary Table S5). Obvious deactivation can be observed at all conditions of 1.2, 5, and 10 h−¹ at 600 °C, which are mainly attributed to the cracking reaction of C₃H₈ at 600 °C, with generated substantial carbon deposits that cover the Co active sites (refer to TG data in Supplementary Fig. 28). Catalytic performance of 1.6%Co@S-1 in PDH. a Effect of C3H8 content on C3H8 conversion and initial C3H6 formation rates, reaction conditions: T = 580 °C, WHSV = 1.2 h−1; b effect of temperature on C3H8 conversion and initial C3H6 formation rates, reaction conditions: feed gas of 50 mol% C3H8 in N2, WHSV = 1.2 h−1; c C3H6 formation rates at different WHSV values as a function of time on the stream and reaction conditions: feed gas of 50 mol% C3H8 in N2, T = 600 °C; d reaction-regeneration cycles, reaction conditions: feed gas of pure C3H8, T = 550 °C, WHSV = 5 h−1, regeneration conditions: 550 °C, air (20 mL min−1), 1 h; e catalyst stability during 240 h with a feed gas composition of 50 mol% C3H8 in N2. f Comparison of the catalytic performance of the 1.6% Co@S-1 catalyst tested herein with reported Cr, Zn, V, and Co-based PDH catalysts presented in Supplementary Table S6 with related details. The stability of the 1.6% Co@S-1 catalyst in the PDH was also investigated. Eight regeneration cycles were conducted at 550 °C, WHSV = 5 h−1 and pure C3H8 (Fig. 5d). Initially, the propane conversion rate decreased from 31.8 to 29.9% within 8 h. However, this rate was restored to 31.8% after regeneration by air-calcination, and the propylene selectivity remained constant at 95.0%. It was concluded that ((≡Si-O)2Co(≡Si-O)2…CoO) structures were stably present in the S-1 zeolite, resulting in mitigated degradation of the catalytic performance. The results of 240 h operation performance using 50% C3H8 feed gas at different reaction temperatures and space velocities are shown in Fig. 5e. The conversion of C3H8 decreased from 20.4 to 17.0%, while C3H6 selectivity stabilized at approximately 98.0% (C3H8 equilibrium conversion at 23.3%, Supplementary Fig. 29). Such PDH catalytic performance is much superior to that previously reported Co-based catalysts (Fig. 5f, and Supplementary Table S6). The maximum C3H6 formation rate reached 96.02 mmol gcat−1 h−1 at 50 vol% C3H8 feed and 600 °C, while a relatively low temperature of 500°C resulted in a propene formation rate of 9.17 mmol gcat−1 h−1. Therefore, we concluded that the ((≡Si-O)2Co(≡Si-O)2…CoO) clusters likely resulted in highly active sites for the PDH reaction. A comparative study was conducted to investigate the catalytic performance of the PDH reaction over the 0.3% Co@S-1 and 1.6% Co@S-1 catalysts. The C3H8 conversion and C3H6 selectivity of the 0.3% Co@S-1 catalyst remained constant at 12 and 98%, respectively. Under identical reaction conditions, the C3H8 conversion rate of the 1.6% Co@S-1 catalyst was 6-fold that of the 0.3% Co@S-1 catalyst (Supplementary Fig. 30). In addition, in order to evaluate the catalytic performance of different Co active species, we subjected the 1.6% Co@S-1 catalyst to varying degrees of acid etching and observed that the initial conversion of propane decreased significantly with the reduction of CoOx species (Supplementary Fig. 31). The C3H6 formation rates and turnover frequency (TOF) of the 0.3% Co@S-1 and 1.6% Co@S-1 catalysts were compared and shown in Supplementary Fig. 32. The formation rates of C3H6 and the TOF of the 1.6% Co@S-1 catalyst were significantly higher than those of the 0.3% Co@S-1 catalyst. The PDH reaction was also performed in an elevated temperature reducing atmosphere, which resulted in the deactivation of metal oxide species within the zeolite through reduction. However, the conversion of C3H8 and the selectivity of C3H6 remained nearly unchanged for the 1.6% Co@S-1 catalyst when treated with H2 at 580 °C for 1 h (Supplementary Figs. 33 and 34). Additionally, the 1.6% Co@S-1 catalyst was subjected to XPS analysis following the reaction, and we observed the presence of predominantly Co2+ species, with H2-TPR, XRD and UV–vis spectroscopy showing no significant changes after reaction (Supplementary Fig. 35). The catalytic performance of the 1.9% impregnated Co/S-1 catalyst is shown in Supplementary Figs. 36–38. In summary, the ((≡Si-O)2Co(≡Si–O)2…CoO) species in the 1.6% Co@S-1 catalyst were stabilized and remained non-oxidized during catalyst preparation (air, 550 °C) as well as unreduced in the PDH reaction. This indicated that the co-existence of isolated Co–(O–Si)4 sites and CoOx clusters in the 1.6% Co@S-1 catalyst provided efficient PDH activity, which was stabilized during the reaction conditions. Simultaneously, the (≡Si–O)2Co(≡Si–O)2…CoO structure of the 1.6% Co@S-1 sample exhibited distinctly superior PDH performance compared to the tetrahedrally coordinated Co–(O–Si)4 in the 0.3% Co@S-1 catalyst. PDH mechanism investigation To obtain further insight into the high activity and stability of the (≡Si–O)2Co(≡Si-O)2…CoO species encapsulated within the 10-membered ring of the MFI zeolites in the PDH reaction, computational models of 0.3% Co@S-1 and 1.6% Co@S-1 were constructed based on the EXAFS fitting results. The aforementioned spectroscopic and characterization analysis results revealed the coexistence of ultra-small CoOx nanocluster and tetra-coordinated Co–(O–Si)4 sites, which were connected together through the oxygen adjacent to the tetrahedrally coordinated Co in the 1.6% Co@S-1 catalysts. Bader charge analysis revealed the framework oxygen in proximity to isolated Co in the 0.3% Co@S-1 catalyst to have a charge of −1.2 e−, which was higher than the charge of −1.4 e− observed on the framework oxygen in proximity to the tetrahedral Co in the 1.6% Co@S-1 catalyst (Fig. 6a). This indicated that a greater number of electrons were transferred from the CoOx clusters to the framework oxygen adjacent to the tetrahedral Co in the zeolite, thereby enhancing the binding strength between the CoO clusters and framework Co. The differential density charge results also indicated that the CoO clusters transferred 0.98 e− to the isolated Co–O–Si sites, in line with this conclusion (Supplementary Fig. 39). Additionally, the charge density difference of the C3H8 molecules on the 1.6% Co@S-1 catalyst was higher than the 0.3% Co@S-1 catalyst, revealing an enhanced adsorption strength of the C3H8 intermediates by electron transfer or charge redistribution29 (Supplementary Fig. 40). The d band center was identified as a descriptor of the reactant adsorption energy30. Consequently, the density of states (DOSs) values for Co 3d in 0.3% Co@S-1 and 1.6% Co@S-1 catalysts were calculated (Fig. 6b). Investigation of the mechanism of the PDH reaction on the Co@S-1 catalysts. a DFT-optimized structures and bader charge of 1.6% Co@S-1 and 0.3% Co@S-1 catalysts; b the calculated partial density of states (PDOS) and d band center for the 0.3% Co@S-1 and 1.6% Co@S-1 catalysts; c Arrhenius plot on various reaction temperatures for the 0.3% Co@S-1 and 1.6% Co@S-1 catalysts, reaction conditions: pure C3H8, T = 520, 530, 540, 550 and 560 °C; d the energy barrier (Ea) of the first and second dehydrogenation steps of C3H8 with the corresponding intermediate structures on 0.3% Co@S-1 and 1.6% Co@S-1; e–g PDH reaction pathway and transition states with corresponding energy profiles, and mechanism investigation on the 0.3% Co@S-1 (cyan) and 1.6% Co@S-1 (violet) catalysts. Compared to the 0.3% Co@S-1 catalyst, the 1.6% Co@S-1 sample exhibited a closer proximity to the Fermi energy level, revealing a greater electron occupation surrounding the (≡Si–O)2Co(≡Si–O)2…CoO structure. Meanwhile, the catalytic kinetic studies of PDH were performed in the temperature range of 520–560 °C. The calculated apparent activation energy of the 1.6% Co@S-1 catalyst (102.6 kJ mol−1) based on Arrhenius curves was significantly lower than that of the 0.3% Co@S-1 catalyst (120.0 kJ mol−1), as shown in Fig. 6c. Therefore, the (≡Si–O)2Co(≡Si–O)2…CoO structure of the 1.6% Co@S-1 sample catalytic performance improved, which was consistent with the experimental results. Subsequently, the free energies and transition states were calculated by DFT for both structures for C–H bond activation in the PDH reaction (Fig. 6d–g and Supplementary Table S7). The adsorption energies (Eads) of C3H8 on the 0.3% Co@S-1 and 1.6% Co@S-1 catalysts were −0.85 and −1.16 eV, respectively (Supplementary Tables S4). The combined differential charge results indicated that the (≡Si–O)2Co(≡Si–O)2…CoO structure was electron-rich, which facilitated the adsorption activation of C3H8 (Supplementary Fig. 39). In the initial dehydrogenation step of the 1.6% Co@S-1 catalyst, the adsorbed C3H8* underwent dehydrogenation to form an n-propyl intermediate (TS1: CH3CH2CH3* → CH3CH2CH2* + H*). During this process, C3H7* species coordinated with the Co atom, while the O atom of the CoOx clusters acted as a Lewis base, capturing the detached H*. The dissociation energy for this step was 1.84 eV, which was significantly lower than that of the 0.3% Co@S-1 catalyst (2.01 eV). Robust Co-C3H7* binding enabled the second dehydrogenation step (TS2: CH3CH2CH2* + H* → CH3CHCH2* + 2H*), with an estimated energy barrier of 0.43 eV. This was also significantly lower than the first dehydrogenation step, indicating that TS1 was the rate-determining step (RDS), in line with the reported results in literature31 (Fig. 6d). The subsequent coupled H–H desorption was achieved by the interaction of Co–H and O–H, with an energy barrier of 1.07 eV (TS3: CH3CHCH2* + 2H* → CH3CHCH2* + H2 (g)). Finally, the (≡Si–O)2Co(≡Si–O)2…CoO structure reverted to its original configuration after the desorption of C3H6. In the 0.3% Co@S-1 catalyst, the adsorbed C3H8* was activated by the tetrahedrally coordinated framework Co and adjacent oxygen atoms (Fig. 6e). The two isolated H* were bonded to the framework oxygen atoms after breaking the C–H bond, while C3H7* was adsorbed to another O atom close to the tetrahedrally coordinated framework Co. These two processes showed energy barriers of 2.01 and 0.63 eV, respectively, which were higher than the TS1 and TS2 energy barriers in the 1.6% Co@S-1 catalyst, suggesting that the (≡Si–O)2Co(≡Si–O)2…CoO structure was more favorable for C–H bond breakage. Subsequently, two O–H interactions combined to form H2 with a dissociation energy of 1.95 eV, which was significantly higher than the TS3 process in 1.6% Co@S-1 catalysts. In conclusion, DFT calculations demonstrated that the (≡Si–O)2Co(≡Si–O)2…CoO structure in the 1.6% Co@S-1 catalyst enhanced propane adsorption/activation, effectively lowering the energy barrier for the first dehydrogenation step and improving the PDH reaction catalytic performance. Discussion In summary, we synthesized a well-performed PDH catalyst with a unique structure of (≡Si–O)2Co(≡Si–O)2…CoO encapsulated in the 10-MR of Silicate-1, which showed a high propylene formation rate (34.0 mmol gcat−1 h−1) and a slow deactivation rate (0.0116 h−1) under a practical condition (100% propane, 550 °C, 5.0 h−1, 72 h). This catalyst was prepared by dry-gel transformation of zeolite seeds impregnated with Co2+. iDPC-STEM and EXAFS analyses convinced that CoO nanoclusters were embedded within 10-MR of MFI zeolite through interaction with tetrahedrally coordinated Co–(O–Si)₄ species. DFT calculations further indicated that the electron-negativity property of (≡Si–O)2Co(≡Si–O)2…CoO species facilitated the adsorption and activation of propane, as well as the subsequent dehydrogenation steps to selectively produce propylene. In addition, the low formation energy of the Co–CoOx active site led to the high stability of Co@S-1 catalyst even in the presence of a hydrogen atmosphere. Methods Synthesis of silicate-1 seeds 103.64 g of TEOS, 183.00 g of TPAOH (25 wt%) and 42.60 g of deionized water were mixed and stirred vigorously for 6 h at room temperature, and then heated in a Teflon-lined stainless-steel autoclave at 90 °C for 2 days under autogenous pressure. The solid in the resultant mixture was collected by centrifugation and dried overnight at 100 °C. Finally, the template was removed by calcination in static air at 550 °C for 5 h. Synthesis of desilicated S-1 It was prepared by treating the S-1 with TPAOH, PI, PI, and NaOH mixtures (C = 0.1, 0.2, 0.3 and 0.4) in the liquid-to-solid ratio of 40 mL g−1 of the S-1 at T °C (T = 30, 50, 70, and 90) under stirring for X h (X = 2, 4, 6, and 8). Typically, 1 g of S-1 seeds was added to 40 mL of 0.3 mol L−1 TPAOH and treated at 70 °C for 4 h. The solids were filtered and washed to pH = 7, and finally dried to obtain desilicated S-1 samples. Synthesis of Co/S-1 The Co/S-1 samples were synthesized by conventional impregnation using Co(NO3)2·6H2O and desilicated S-1 as support. Typically, 0.1 g of Co (NO3)2·6H2O was added to a 10 mL of deionized water, and 1 g desilicated S-1 was subsequently added and stirred at 40 °C to complete evaporation of water. After drying at 100 °C for 6 h, the 0.5% Co/S-1, 1% Co/S-1, 2% Co/S-1, 4% Co/S-1, and 6% Co/S-1 samples were obtained with different Co content loadings. Synthesis of Co@S-1 Co@S-1 was prepared by a solid crystallization method. Typically, 0.10 g of silica aerogel, 0.5 g of 2% Co/S-1, and 0.25 g of TPABr were mixed and ground for 20 min. The solid was treated at 180 °C for 3 days for crystallization using the dry gel conversion method and then thoroughly rinsed with water and calcined at 550 °C for 5 h in air to remove the template, and the sample was denoted as 1.6% Co@S-1. The catalysts were prepared similarly for 0.4% Co@S-1, 0.8% Co@S-1, 3.3% Co@S-1, and 5.2% Co@S-1 (Co content determined by ICP). In addition, the 1.6% Co@S-1 sample was washed with 1 mol/L HCl and 0.1 mol/L H2SO4 to remove ultra-small CoO clusters, resulting in the formation of 0.3% Co@S-1. Typically, 1 g 1.6% Co@S-1 catalyst was stirred at room temperature for 24 h in 1 mol L−1 HCl and then treated at 60 °C for 4 h in 0.1 mol L−1 H2SO4 (solid-liquid ratio of 1 g /40 mL). The final material was obtained after washing to neutral and drying, named 0.3% Co@S-1, which represents S-1 zeolite with only single Co atoms introduced. Simultaneously, a 2% Co/S-1 sample from the Synthesis of Co/S-1 section was directly calcined in air at 550 °C for 5 h without dry gel conversion, producing 1.9% Co/S-1, indicating the catalyst was synthesized by the direct impregnation method. Catalyst characterization X-ray diffraction (XRD), Transmission Electron Microscopic (TEM), X-ray photoelectron spectroscopy (XPS), Raman spectra, UV–vis spectra, DFT computation and other characterization techniques are detailed in the Supplementary Information. Catalytic reaction Catalytic PDH reaction was performed on a fixed-bed reactor under ambient pressure. Granular catalyst (0.2 g, 40–60 mesh) was loaded inside a quartz tube reactor (quartz tube OD 10 mm, ID 6 mm, length 400 mm) and adjusted to the constant temperature zone of the furnace. The catalyst was pretreated in N2 from room temperature to 500 °C and kept for 30 min to reach a steady state. Then, C3H8 gas mixture (5% mol C3H8/N2, 50% mol C3H8/N2, 100% mol C3H8/N2) was introduced into the reactor and the temperature was controlled at 500–580 °C. After 10 min on stream, the products were measured by an online gas chromatography (GC9790Plus, Fuli Instruments) equipped with a flame ionization detector (FID) and a thermal conductivity detector (TCD). The hydrocarbons of products were analyzed in FID with a capillary column (RB-Al2O3, 30 m × 0.53 mm × 15 μm). Based on the external standard and carbon balance method, the conversion and selectivity of the products were calculated in the Supplementary methods. Supplementary information Supplementary Information Transparent Peer Review file Source data Source Data Acknowledgements This research was supported by the National Natural Science Foundation of China (General Program, No. 52573241), Shanghai Municipal Science and Technology Commission Project (No. 22dz1209300) and the National Natural Science Youth Fund (No. 22205065). Author contributions C.Z., B.M. and J.T. conceived the project and directed the study. R.L. and B.M. carried out the synthesis, structural characterizations and catalytic measurements. R.L. and L.Z. conducted the STEM and XAS examinations. R.L., B.M. and C.Z. wrote the paper. All the authors discussed the results and commented on the paper. Peer review Peer review information Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available. Data availability All relevant data supporting the key findings of this study are available within the Supplementary Information files and Source Data or from the corresponding author upon request. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Bing Ma, Email: bma@chem.ecnu.edu.cn. Chen Zhao, Email: czhao@chem.ecnu.edu.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41467-025-66595-x. References Associated Data Supplementary Materials Supplementary Information Transparent Peer Review file Source Data Data Availability Statement All relevant data supporting the key findings of this study are available within the Supplementary Information files and Source Data or from the corresponding author upon request. Source data are provided with this paper.