Introduction

Supported metal catalysts are known as one of the most important candidates for heterogeneous catalysis1. In the vast majority of catalytic processes, the interaction between support and active metal site is of great importance in determining their catalytic performance2. When utilizing reducible metal oxides as supports3,4,5, taking advantage of their reducibility and capacity for metal-support bonding, the concept of strong metal-support interaction is proposed6,7,8. In terms of non-reducible supports of Al2O3 and SiO2, these interactions can hardly involve significant charge transfer or the involvement of support lattice oxygen in the catalytic cycle9,10. Hydroxyl groups on the surface of these supports are therewith emphasized as the bridge to build required interaction between metal and support, which is beneficial to acquiring highly dispersed and stable metal sites on the supports11,12.

The Al-OH sites on Al2O3 are revealed for anchoring metal species. Abundant hydroxyl groups on the γ-Al2O3 cause the single-atom dispersion of Ag species13. Nevertheless, coordinatively unsaturated aluminum atoms on the Al2O3 also shows an anchoring effect for metal sites14. From this point of view, silica-based support serves as an ideal candidate to independently study the interaction between hydroxyl groups and active metals. Vodyankina and Bao reported that the presence of surface OH groups on the SiO2 regulated the distribution and size of Ag nanoparticles12,15. More recently, Smith et al. found that local distribution of OH groups on SBA-15 stabilized V species and contributed to preferable catalytic behaviors16. Taken together, in silica-supported metal catalysts, surface hydroxyl plays a vital part in regulating the dispersion of the introduced metal sites. In general, the adsorbed water is commonly preserved in conventional synthesis process, and the participation of both water molecules and the derived hydroxyl groups in heterogeneous catalysis is fundamental and important. However, the special role of adsorbed water on the silica is generally ignored, resulting in the lack of a deeper understanding of synthesis mechanism.

In the meantime, with the growing demand for propylene and the explosion of the shale gas revolution17, propane dehydrogenation (PDH) process has become more commercially attractive18,19, which exhibits significant environmental and energy implications. However, commercial PDH processes using high cost of PtSn/Al2O3 and high toxicity of Cr/Al2O3 catalysts have restricted their further developments20,21. Recently, the well-defined and highly dispersed sites of Fe22, Ga23, Zn24, Co25 and Y26, obtained via grafting tailored metal precursors within hydroxyl on the silica supports, have shown the capability to break C-H bonds of alkanes. Among these catalysts, highly dispersed tetrahedral Co(II) sites (Td-cobalt(II)) have gained much interest, due to its superior PDH catalytic potential27,28. Hock et al. developed strong electrostatic absorption (SEA) method to synthesize Td-cobalt(II) contained Co/SiO2 catalyst27. Similarly, Sooknoi et al. investigated the influence of different cobalt precursors, including [Co(NH3)5Cl]Cl2, [Co(bipy)3](NO3)2 and [Co(en)2Cl2]Cl, on the reactivity of the Co/SiO2 catalyst29. Besides, surface organometallic chemistry (SOMC) was employed to obtain this highly dispersed Td-cobalt(II) through protonolysis of organometallic precursor within surface OH groups on the SiO2, and bulky ligands were devoted to facilitate sites isolation25,29. In these cases, although hydroxyl is fully utilized to realize the required interaction between metal and support, expensive metal precursors and complicated synthesis routes are necessary to achieve precise control over the surface structure of the catalysts30. Specifically, adsorbed water coexists normally with the surface hydroxyl groups of silica support, focuses towards the role of adsorbed water in regulating the generation of highly dispersed and stable metal sites, however, is yet underestimated.

Herein, we described the role of adsorbed water during the synthesis of highly dispersed Td-cobalt(II) sites onto the silica support. The results showed that the hydrates were critical to the tetrahedral Co site formation. It was proposed that under a simple direct reduction process, the adsorbed water could help to immobilize the Co precursor, and then further form Td-Co(II) sites. A systematic characterization and DFT calculation proved the existence of the adsorbed water and the importance of the intermediate of [Co(H2O)6]2+, respectively. More importantly, the resulting Td-cobalt(II)/SBA-15 catalyst was superefficient for PDH reaction, which exhibited better reactivity when compared with other reported Co based catalysts. The present work illustrates new understanding of adsorbed water on silica for inducing the formation of highly dispersed cobalt(II) sites, and provides simple and effective approach to design high reactivity of cobalt-based PDH catalyst.

Results

Formation of tetrahedral Co(II)

We primarily attempted to compare the difference in the chemical states of cobalt species obtained from direct reduction (Dir-reduction/catalyst precursor was directly reduced by H2) and indirect reduction (H2-reduction/air-calcination sample was reduced by H2) processes. Firstly, XRD was examined to identify the crystal structure of Co species on different samples (Fig. 1a). The diffraction peaks of Air-calcination sample fit well with that of the standard Co3O4 at 2θ = 36.9°, 59.3° and 65.2° (JCPDS No. 42-1467). After H2 reduction (indirect reduction), the peaks of Co3O4 disappeared and a weak diffraction peak of metallic Co at 2θ = 41.2° was observed, which was further confirmed by TEM from Supplementary Fig. 1, where obvious aggregation of Co with interplanar spacing of d(103) = 0.453 nm were detected. In contrast, the Co species from the Dir-reduction catalyst were undiscerned (Fig. 1a). Additionally, from Fig. 1g and Supplementary Fig. 2–4 (TEM and STEM/EDS-mapping), the highly homogeneous dispersion of Co species was witnessed, and the SAED pattern indicated the amorphous Co states, suggesting the markedly improved dispersion of Co species via the H2-direct reduction method. Secondly, XPS was used to investigate the surface chemical states of catalysts. All the samples showed two broad and asymmetric main peaks at about 775–790 eV and 790–810 eV (Fig. 1b), which were corresponded to Co 2p3/2 and Co 2p1/2, respectively. For the Co 2p spectra of the Air-calcination catalyst, two spin-orbit doublet peaks of Co(II) and Co(III) and their broad satellite peaks were displayed. The dominating peaks of Co(II) at BE = 781.5 eV and Co(III) at BE = 779.4 eV were in good agreement with those reported for Co3O431. Moreover, metallic Co evidenced by the BE = 778.1 eV was found in the H2-reduction sample32, suggesting that Co3O4 was reduced by H2 treatment, and consistent with the XRD and TEM results. It is worth noting that all of the Co species in the Dir-reduction catalyst were existed as Co(II) (Co 2p3/2 at 781.8 eV and Co 2p1/2 at 797.8 eV), revealing that Co was neither reduced to Co0 nor oxidized to Co3O4 on the Dir-reduction catalyst. Moreover, quasi in-situ XPS measurement from Supplementary Fig. 5 confirmed again that only Co(II) species were found on the Dir-reduction sample.

Fig. 1: Functional characterization of Co/SBA-15 catalysts.
figure 1

a XRD patterns, b Co 2p XPS spectra, c H2-TPR profiles, d UV-vis spectra for Dir-reduction, Air-calcination and H2-reduction catalysts, e the normalized intensity of Co K-edge XANES spectra, f the corresponding Fourier transformation of k3-weighted EXAFS oscillation, g STEM image and EDS elements mapping of Dir-reduction catalyst.

Full size image

Following, H2-TPR was performed to characterize the reduction behavior of the cobalt species (Fig. 1c). Air-calcination catalyst presented three reduction peaks at about 300 °C, 350 °C, and 500 °C, ascribing to the subsequent reduction from Co3O4 to metal Co (Co3+→Co2+→Co0)33. No reduction peak was observed in the H2-reduction catalyst, since the catalyst was in-situ reduced in the TPR equipment. Namely, H2 treatment consumed the oxygen in Co3O4 to form metal Co, which can be supported by the XRD, XPS, and TEM results. Interestingly, very different reduction curve was exhibited on the Dir-reduction sample, and the main feature of the peaks in the TPR profiles was apparently shifted to high temperature region at approximately 800 °C, revealing that Co species were connected strongly within the silica support and hard to be reduced34. The UV-Vis spectra provided the configuration of various cobalt species. Adsorption bands at about 410 nm and 720 nm were observed on the Air-calcination catalyst, which were allocated to the ligand-metal charge-transfer of O2− → Co2+ and O2− → Co3+ in spinal Co3O435. By contrast, H2-reduction catalyst showed very weak peak due to Co3O4 was reduced into Co0 phase. For the Dir-reduction catalyst, a broad absorption band with maximum peaks at approximately 543 nm, 578 nm, and 642 nm were observed, which was associated with ν3 (4A2 → 4T1(P)) transition characteristic of the tetrahedral Co(II) (Td-Co(II))36,37.

To acquire more information about the coordination and structural features of the Co species in the Dir-reduction catalyst, EXAFS measurements were implemented. The XANES spectrum of the Dir-reduction catalyst (red line) from Fig. 1e showed an absorption edge located between Co foil and CoO (closer to CoO). Therefore, combining the EXAFS and XPS data, we inferred that the valence state of cobalt in the Dir-reduction catalyst was around +2. Consistently, the conclusion of Bader charge calculation from Supplementary Fig. 6 also demonstrated that the oxidation state of Co was very close to +2. Besides, the results of EXAFS fitting and corresponding Fourier transform were summarized in Fig. 1f, Supplementary Fig. 7 and Supplementary Table 1. The EXAFS spectra recorded Co-O and Co-Co shells with a distance of 2.07 Å and 3.12 Å, respectively. The corresponding coordination number (CN) were 4.0 and 4.8, which was lower than the average number of the CoO phase, indicating that the tetracoordinated Co(II) species dominated over the Dir-reduction catalyst. It was clarified that the presence of cationic cobalt bonded to the silica with Co-O-SiOn linkages at the Co-SiO2 interfaces stabilize the dispersed Co species38.

Correspondingly, the discussed characterizations illustrated collectively that the Co3O4 was mainly existed in the Air-calcination catalyst, and it was easily reduced into metallic Co after H2 reduction. Importantly, as for the Dir-reduction catalyst, highly dispersed and extremely stable Td-Co(II) reduced hardly by H2 flow below 800 °C were obtained. Besides, the formation of Td-Co(II) species was also evidenced on the Dir-reduction catalysts with different Co contents (4 and 6%), as revealed from Supplementary Fig. 8 (XRD), Supplementary Fig. 9 (XPS), Supplementary Fig. 10 (H2-TPR), and Supplementary Fig. 11 (UV-vis).

Verifying the effect of absorbed water on Td-Co(II) formation

It is universally accepted that surface OH groups on SiO2 are the anchoring sites to precisely acquire the isolated metals39. Nevertheless, the role of adsorbed water on silica is always neglected, since the stabilization of metal on supports occurs at high temperature. On account of this, following experiments were designed to demonstrate the importance of absorbed water in the formation of isolated Td-Co(II) (Fig. 2 and Supplementary Fig. 12, it is noted that the four contrast samples of PM, PM200, PM900 and IMP900 were all conducted by H2-direct reduction treatment, and the reduction temperature was 600 °C). Generally, in the process of catalyst synthesis, the source of the water was classified into two categories, one was the liquid water added during the impregnation process, and the other was the adsorbed water connected with OH groups on the support. Firstly, SBA-15 support was mixed physically with Co(NO3)2·6H2O and subjected to H2-direct reduction treatment (Dir-reduction (PM)), to exclude the effect of liquid water on the formation of Co species (Fig. 2a, none of liquid water was exposed). Secondly, the hydroxylated surface of SBA-15 with a predominance of silanol groups is hydrophilic in nature, which presented a large amount of adsorbed water on the support. To eliminate the influence of absorbed water (Fig. 2b, neither liquid water nor adsorbed water was introduced), SBA-15 was heated at 200 °C (TG data in Supplementary Fig. 13 illustrated the complete removal of absorbed water), and then mixed physically with Co(NO3)2·6H2O in glove box before conducting H2-direct reduction (the sample was reduced directly by H2, Dir-reduction (PM200)). Thirdly, most of the surface hydroxyl groups were removed irreversibly by calcining SBA-15 at 900 °C (Supplementary Fig. 14), although it still retained the characteristic structure of SBA-15 (Supplementary Fig. 15). Afterwards, 900 °C-calcined SBA-15 was mixed physically with Co(NO3)2·6H2O and reduced directly by H2 to obtain Dir-reduction (PM900) sample (Fig. 2c, without absorbed water and surface hydroxyl). Fourthly, the residual little amount of surface OH on the 900 °C-calcined SBA-15 was utilized to obtain small amount of adsorbed water during impregnating Co(NO3)2·6H2O onto the support (the sample was reduced directly by H2, Dir-reduction (IMP900), Fig. 2d, a small amount of absorbed water was presented). The details for the existence of adsorbed water and hydroxyl groups in the four design experiments were given in Supplementary Fig. 12.

Fig. 2: Schematics of four design experiments were devoted to verify the effect of absorbed water.
figure 2

a physically mix SBA-15 and cobalt precursor. b physically mix SBA-15(200 °C) and cobalt precursor. c physically mix SBA-15(900 °C) and cobalt precursor. d impregnate SBA-15(900 °C) within cobalt precursor.

Full size image

Specifically, in Fig. 3a (XRD), no diffraction peak was seen in all the catalysts, implying uniform distribution of Co species. In Fig. 3b (in-situ TPR), Dir-reduction (PM) catalyst exhibited one significant peak higher than 800 °C, representing the reduction of Co(II) species that were strongly interacted with silica support. In contrast, almost no remarkable reduction peak was observed in the TPR curves of the Dir-reduction (PM200) and the Dir-reduction (PM900) catalysts, suggesting the absence of Co(II) species that were intensely interacted with SBA-15 support. Significantly, in the Dir-reduction (IMP900) catalyst, the characteristic reduction peak of embedded Co(II) sites on the support was observed, but the temperature and the intensity of the reduction peak at around 750 °C was downshifted when compared to the Dir-reduction (PM) sample, revealing that Co species did not interact strongly with the support when the surface OH was insufficient. Furthermore, quasi in-situ XPS of the Dir-reduction (PM) catalyst from Fig. 3c showed the BE of typical high-spin Co(II) species. On the contrary, the rest three catalysts indicated the peaks of metallic Co. In addition, samples were analyzed by the ex-situ UV-vis and recorded in Fig. 3d. The absorbance at 543 nm, 578 nm, and 642 nm, as the characteristic band for Td-Co(II), were observed in the Dir-reduction (PM) and Dir-reduction (IMP900) catalysts. But the intensity of Td-Co(II) over the Dir-reduction (IMP900) was lower than that on the Dir-reduction (PM). Over the Dir-reduction (PM200) and Dir-reduction (PM900) catalysts, the characteristic peak representing for cobalt oxide at approximately 410 nm and 720 nm were detected. It is worth noting that in Fig. 3b there was no reduction peak below 600 °C can be found because these four catalysts were subjected to H2-direct reduction treatment at 600 °C. However, since metallic cobalt was easily re-oxidized in air, ex-situ UV-vis spectra captured the diffraction peak of cobalt oxide (Fig. 3d). Therefore, it can be drawn that Td-Co(II) can only be obtained in the Dir-reduction (PM) and Dir-reduction (IMP900) catalysts that contained adsorbed water. Namely, adsorbed water on the SBA-15 support was essential for the acquisition of required Td-Co(II).

Fig. 3: Characterizations of verifying the effect of absorbed water on Td-Co(II) formation.
figure 3

a XRD pattens, b in-situ H2-TPR profiles, c quasi in-situ Co 2p XPS spectra, d ex-situ UV-vis spectra of the four designed catalysts; e UV-vis spectra of the dried catalyst precursors (without reduction treatment), f 1H-NMR of the dried catalyst precursors with different Co loadings; g the normalized intensity of Co K-edge XANES spectra, h the corresponding Fourier transformation of k3-weighted EXAFS oscillation of the dried precursor for Dir-reduction catalyst; i UV-vis spectra of the Dir-reduction catalysts by using different reduction temperature.

Full size image

A detailed analysis of how Co species work with absorbed water will be discussed in the following. UV-vis spectra of the samples dried at 100 °C (without reduction treatment) were depicted in Fig. 3e. As for the catalyst precursors of Dir-reduction and Dir-reduction (PM) that contained sufficient adsorbed water, the existence of [Co(H2O)6]2+ was supported by the typical triplet at 525 nm, 575 nm, and 650 nm40,41. Whereas, this peak was not found in the precursor of Dir-reduction (PM200) catalyst, since the absorbed water was completely removed. Inversely, one peak consistent with Co(NO3)2·6H2O was appeared. Figure 3f presented the 1H-NMR of the dried precursor for Dir-reduction catalysts (after drying at 100 °C). Two resonances at around 1.2 ppm and 3.7 ppm were detected on the pure SBA-15 support. Normally, the peak at 3.7 ppm was assigned to the hydrogen-bonded OH42, while the sharp peak at 1.2 ppm was assigned to the isolated OH43. Besides, a peak that emerged at around 4.9 ppm was related to the adsorbed water on the SBA-15 surface44. Notably, with the increase of Co loading, the peak of adsorbed water was widened and shifted to the higher value position, which, together with the results of UV-vis from Fig. 3e, can further illustrate that this part of absorbed water was existed in the form of [Co(H2O)6]2+. Namely, the structure of [Co(H2O)6]2+ was well maintained after drying at 100 °C. As shown in Fig. 3g, the Co K-edge XANES spectra of the dried precursor of Dir-reduction catalyst was similar to that of Co(NO3)2·6H2O, but the changes in the electronic environment of Co was evidenced by the slight shift of absorption edge. In Fig. 3h, the Fourier transformation of the k3-weighted EXAFS showed that Co-O bonds within the CN of both Co-O1 and Co-O2 around 6 were observed, where their bond lengths were 2.089 Å and 3.514 Å, respectively, which displayed slight difference when comapred with the referenced sample of Co(NO3)2·6H2O. As a result, it means that the cobalt species were no longer in the form of cobalt nitrate at this time. The above experimental results lead to the conclusion that the existence of sufficient adsorbed water induced the transformation of precursor of Co(NO3)2 into intermediate of [Co(H2O)6]2+, and finally, Td-Co(II) was gained after H2-direct reduction.

Following, we carried out UV-vis characterization of the catalysts precursors reduced directly at different temperatures (Fig. 3i). Interestingly, it is found that the characteristic peaks of Td-Co(II) appeared at 150 °C, and this structure was well preserved at higher reduction temperature of 250 °C, 400 °C, and 600 °C. It is reported that dehydroxylation of SiO2 started at least higher than 190 °C45, and from this point of view, it is a concern in our study to reveal how the Co(II) replace hydroxyl group to form the stable Td-Co(II) structure when dehydration condensation was performed below 190 °C. We hypothesized, based on the above results, the formation of [Co(H2O)6]2+ within the presence of adsorbed water may promote the dehydration condensation of surface hydroxyl sites.

Mediating mechanism of adsorbed water

In-situ DRIFTS of NH3 adsorption-desorption and in-situ FT-IR were applied to reveal the changes of catalyst surface groups during H2-direct reduction process, so as to illustrate the role of [Co(H2O)6]2+ in the formation of Td-Co(II). In view of the fact that NH3 can interact with hydroxyl groups46, which was regarded as a probe molecule to quantitatively analyze the concentration of surface OH on the SBA-15. If dehydration condensation can be facilitated by the presence of [Co(H2O)6]2+, the residual amount of hydroxyl on Co/SBA-15 was supposed to be less than that of SBA-15. Consequently, in-situ DRIFTS of NH3 adsorption-desorption was performed in Figs. 4e–3g. Obviously, the peak areas of both Co/SBA-15 and SBA-15 decreased with increasing reduction temperature, which was due to the gradual initiation of dehydroxylation during the heating process. More importantly, it is seen from Fig. 4e and Fig. 4f that the intensity of the peak over Co/SBA-15 was lower than that of SBA-15 at the same reduction temperature, and the peak area of Co/SBA-15 catalyst was smaller than that of the SBA-15. This difference was more distinct at high temperatures of 400 °C and 600 °C (Fig. 4g), which supported our suspicion that [Co(H2O)6]2+ exactly accelerated the condensation of hydroxyl sites.

Fig. 4: Analysis of mediating mechanism of adsorbed water.
figure 4

In-situ FT-IR spectra of a Dir-reduction, b Dir-reduction (PM), c Dir-reduction (IMP900), d Dir-reduction (PM900) during H2-direct reduction process; In-situ NH3-DRIFS of e SBA-15 and f Dir-reduction catalyst in the heating process, g peak area of hydroxyl group over SBA-15 and Dir-reduction catalyst from NH3-DRIFS; h DFT calculations of the free energy of dehydroxylation.

Full size image

Meanwhile, in Fig. 4a, as for the Dir-reduction sample, split peaks of free-NO3− at 1340 cm−1 and 1410 cm−1 were discovered47. When increasing temperature, the disappearance of these two split peaks was accompanied by the display of peak at 1530 cm−1, illustrating that free-NO3− was changed into the monodentate nitrate48. Notably, the nitrate was completely removed at high temperature of 600 °C, suggesting the transformation of cobalt nitrates into the stable Co-O-Si species49. Besides, the peak of adsorbed water at 1630 cm−1 also disappeared gradually with the increase of temperature50, indicating the consumption and transformation of adsorbed water. As for the Dir-reduction (PM) catalyst (Fig. 4b), it showed similar peak pattern to that of the Dir-reduction catalyst (Fig. 4a), illustrating that the adsorbed water substituted NO3− in the precursor of Co(NO3)2 to form hydrate of [Co(H2O)6]2+, resulting in the generation of free-NO3−. The difference is that the intensity of the diffraction peak in Fig. 4c was lower than that in Fig. 4b, which was caused by the fact that lower amount of adsorbed water in the Dir-reduction (IMP900) catalyst was not sufficient to obtain a higher amount of [Co(H2O)6]2+. By contrast, similar processes were not observed on the Dir-reduction (PM900) catalyst (Fig. 4d). In particular, only one peak at 1512 cm−1 attributed to Co(NO3)2·6H2O was emerged49, which is in agreement with the result of the UV-vis from Fig. 3e. In conclusion, it is precisely because the presence of absorbed water which was in favor of the formation of [Co(H2O)6]2+ made Co species easier to be transferred into Td-Co(II).

DFT calculations were carried out to investigate the importance of the intermediate of hydrated cobalt species. The catalyst was orderly treated by the impregnation and drying processes, then Co was existed in the form of [Co(H2O)6]2+(the coordination number and bond distance were calculated from XANES data). Notably, electrostatic interaction between fully coordinated [Co(H2O)6]2+ and the hydroxyls on the support was important in terms of stabilizing the precursors. Namely, [Co(H2O)6]2+ was supposed to be connected by the hydrogen bond (electrostatic interaction) between the oxygen in the silicon hydroxyl group and the hydrogen in the water of the [Co(H2O)6]2+ (Fig. 4h, Supplementary Fig. 18, Supplementary Fig. 19), then the calculated BE of −8.03 kcal/mol (−0.348 eV) from BE = E([Co(H2O)6]2+-SiO2)—E(SiO2)—E([Co(H2O)6]2+) indicated that the [Co(H2O)6]2+-SiO2 formed by the hydrogen bond between SiO2 and [Co(H2O)6]2+ was stable in the catalyst.

Subsequently, free energy for dehydroxylation of SiO2 and Co/SiO2 were calculated, and the dehydroxylation process of SiO2 was exhibited in Fig. 4h and Supplementary Fig. 20. Since the energy barrier (1.264 eV) from IS to TS was the highest among all steps, it was considered as the rate-determining step. With the dehydroxylation carried out, the hydroxyl was removed from silanol groups. The reaction between the surface silanol groups (condensation) lead to the formation of Si-O-Si bonds and molecular water ((≡Si-OH) + ( ≡ Si-OH) → (≡Si-O-Si ≡ ) + H2O), as demonstrated in TS and FS from Fig. 4h and Supplementary Fig. 20, with the free energy of 0.449 eV and −0.796 eV, respectively. By contrast, the dehydroxylation process of Co/SiO2 was compared to that of SiO2 (Fig. 4h and Supplementary Fig. 19). Two processes were simulated over Co/SiO2. The first process was [Co(H2O)6]2+ binding to SiO2 through hydrogen bond, and the second process was dehydroxylation. It is seen that [Co(H2O)6]2+ was stabilized by Si-OH via hydrogen bond with a free energy of −0.291 eV, which illustrated that the transformation from [Co(H2O)6]2+ and SiO2 into [Co(H2O)6]2+-SiO2 was spontaneous. Following, the dehydroxylation stage was proceeded, and [Co(H2O)6]2+ occupied the site of dehydroxylation, with a free energy of 0.069 eV. Then, Co-O-Si bonds were gradually formed at high temperature (TS, Fig. 4h). It is obviously that the rate-determining step in the dehydroxylation over Co/SiO2 was from IM1 to TS, with the energy barrier of 0.634 eV (Fig. 4h), which was lower than that of SiO2. After that, the free energy of IM2 and FS was 0.176 eV and −1.129 eV (Fig. 4h), respectively, suggesting that the Td-Co(II) was acquired spontaneously due to the decrease of free energy. As a result, DFT calculations from Fig. 4h suggested that the energy barrier for dehydroxylation over Co/SiO2 was lower than that over SiO2, confirming that the existence of [Co(H2O)6]2+ promoted the dehydroxylation on the silica support.

As indicated in previous studies, the dehydroxylation temperature on silica started from 190 °C, while the decomposition temperature of cobalt nitrate was about 240 °C51. As a result, in such a close range of temperature, it was difficult for Co(II) to interact with oxygen in the hydroxyl group to form a stable Co-O-Si structure. Inversely, Co oxides were more readily available. Surprisingly, the formation of [Co(H2O)6]2+ reduced the energy barrier of the condensation process on the silica, which made it easier for Co(II) to be immobilized by the hydroxyl groups. This result well interpreted why the UV-Vis spectra in Fig. 3i found the characteristic peak of Td-Co(II) at low temperature of 150 °C, and also in accord with the result of NH3-DRIFS in Fig. 4e–g that the presence of [Co(H2O)6]2+ promoted the condensation of the higher amount of hydroxyl groups.

In conclusion, it is evidenced that the cobalt was in the form of [Co(H2O)6]2+ over Dir-reduction and Dir-reduction(PM) catalysts, which contained abundant absorbed water. In contrast, cobalt nitrate was presented over these Dir-reduction(PM200) and Dir-reduction(PM900) samples, which were lacking in absorbed water. The above results indicated that cobalt nitrate can be converted into [Co(H2O)6]2+ in the presence of adsorbed water, and [Co(H2O)6]2+ made the highly dispersed Td-cobalt(II) sites to be available during direct H2-reduction process. Subsequently, a systematic in-situ characterizations suggested that the presence of [Co(H2O)6]2+ promoted the process of dehydroxylation, and the following DFT calculation also proved the importance of the intermediate of [Co(H2O)6]2+. Besides, as for the specifical role of direct H2-reduction process, it was intended as a way to avoid the oxidation of Co species and facilitate the acquisition of tetrahedral cobalt(II). When adsorbed water was absent, there was no way to obtain highly dispersed Td-Co(II). Hence, the necessary condition for the availability of highly dispersed Td-cobalt(II) sites was the presence of adsorbed water, rather than the direct H2-reduction treatment.

Mechanism of propane dehydrogenation

Figure 5a displayed the reactivity of H2-reduction and Dir-reduction catalysts with Co content of 2% toward PDH at 600 °C. A high C3H8 conversion (37%) and C3H6 selectivity (96%) were achieved over the Dir-reduction catalyst. In contrast, the conversion and selectivity over the H2-reduction catalyst were 15 and 92%, respectively. More importantly, when comparing Dir-reduction sample with other reported cobalt-based catalysts27,32,34,35,52,53,54, it performed apparently the maximum reaction rate (Fig. 5b), suggesting the developed catalyst gives a potential application for PDH. Meanwhile, PDH reactivity comparing Dir-reduction sample with other reported cobalt-based catalysts27,32,34,35,52,53,54, it performed apparently the maximum reaction rate (Fig. 5b), suggesting the developed catalyst gives a potential application for PDH. Meanwhile, PDH reactivity of the Dir-reduction catalysts with Co loading of 4 and 6% was recorded in Supplementary Fig. 21. Similarly, the performance of the Dir-reduction catalysts was better than that of the H2-reduction catalysts. Besides, the reactivity results of the designed samples from Fig. 2 were recorded in Supplementary Fig. 22, and C3H8 conversion was ranked as follows: Dir-reduction > Dir-reduction (PM) > Dir-reduction (IMP900) > Dir-reduction (PM900), which was consistent with our expectations that the content of highly dispersed and stable Td-Co(II) was correlated positively with the PDH reactivity. It is known that the structure of Co3O4 and metallic Co were most likely led to the cracking of C-C bonds, causing the formation of CH4 and some coke precursors34,55, while highly dispersed Td-Co(II) were supposed to be efficient in C-H activation for hydrocarbons52. This clearly explained why the Dir-reduction catalyst performed better catalytic reactivity for the PDH.

Fig. 5: Mechanism of propane dehydrogenation.
figure 5

a C3H8 conversion, C3H6 selectivity, b comparative study of reaction rate for the reported cobalt-based catalysts; C3H8 adsorption in-situ FT-IR of c Dir-reduction catalyst and d H2-reduction catalyst; C3H6 adsorption in-situ FT-IR of e Dir-reduction catalyst and (f) H2-reduction catalyst; g Free energy diagrams for PDH.

Full size image

In-situ FT-IR was developed to further investigate the difference in the activation of propane and the desorption of propene between Dir-reduction and H2-reduction catalysts. In the C3H8 adsorption FT-IR spectra (Figs. 5c, 4d), the bands at 2967 cm−1, 1460 cm−1 and 1370 cm−1 assigned to C3H8 were detected20. It is seen from Fig. 5c that new bands at 1626 cm−1, 1875 cm−1, and 2004 cm−1 representing for the formation of C3H6 were emerged when the Dir-reduction catalyst was heated to 300 °C54, suggesting C3H8 was participated in the reaction to produce C3H6 over the catalyst, while no new peak was detected on the H2-reduction catalyst, which illustrated the better C3H8 activation reactivity over the Dir-reduction sample. Furthermore, C3H6 adsorption FT-IR measurements were given in Fig. 5e, f. Three peaks located at 1626 cm−1, 1875 cm−1 and 2004 cm−1 were also found on the Dir-reduction catalyst (Fig. 5e), which was attributed to the characteristic adsorption bands of C3H6. By contrast, the peak located at 1626 cm−1, assigned to the adsorbed C3H6 on the Td-Co(II), can hardly be observed on the H2-reduction catalyst (Fig. 5f), suggesting strong interaction between the adsorbed C3H6 and the H2-reduction catalyst. It is noted from Fig. 5f that after desorption at 300 °C, a broad peak observed from 1530 cm−1 to 1670 cm−1 resulted from the formed carbonaceous products28, and the peak intensity increased with the desorption time, which indicated that C3H6 was interacted strongly with the H2-reduction catalyst, thus causing the formation of carbonaceous species. Subsequently, in-situ FT-IR of propene hydrogenation process (the reverse of dehydrogenation) was performed in Supplementary Fig. 23. It is observed that the peak at 1626 cm−1 attributed to the adsorbed C3H6 on the Td-Co(II) was absent the H2-reduction catalyst. Moreover, carbonaceous species disappeared on H2-reduction catalyst during the propylene hydrogenation process. When combining the results of Supplementary Fig. 23 and Fig. 5f, it is implied that C3H6 was easily cleaved to carbonaceous species, which caused the poor selectivity of the H2-reduction catalyst. Following, density functional theory (DFT) calculations were applied to illustrate the mechanism of the PDH reactivity on the Dir-reduction and H2-reduction catalysts, and the geometries of the transition state (TS) were exhibited in Supplementary Fig. 24 and Supplementary Fig. 25. It is obtained from Fig. 5g that energy barriers for the activation of first and second C-H bonds on the Dir-reduction catalyst were 0.40 eV and 0.33 eV, respectively, while H2-reduction catalyst possessed higher energy barrier of 0.52 eV and 1.02 eV for the reaction. Regarding the above results, our calculations are consistent with the experimental results of a higher PDH reactivity for the Dir-reduction Co/SBA-15 catalyst.

Finally, a series of characterization tests were also performed to investigate whether Td-Co(II) can be stably presented during the reaction. UV-vis and H2-TPR of the spent Dir-reduction catalyst during the first 30 min of the reaction were tested (Supplementary Fig. 26). It is obvious that the structure of Td-Co(II) remained intact during the reaction. Moreover, XPS, UV-vis and H2-TPR of the regenerated Dir-reduction catalyst from Supplementary Figs. 27 and  28 illustrated that Td-Co(II) was not oxidized by air after regeneration at 600 °C. In conclusion, highly dispersed tetrahedral Co(II) sites obtained from Dir-reduction catalyst shows the high ability to break C-H and maintains a well-defined structure during the reaction.

In summary, a new understanding of adsorbed water on the formation of highly dispersed Td-cobalt(II) sites from direct H2-reduction process was illustrated. It is indicated the cobalt interacted with the adsorbed water that were attached to the hydroxyl group on the silica support through hydrogen bond to form [Co(H2O)6]2+. Subsequently, the formed [Co(H2O)6]2+ was anchored by the oxygen in the hydroxyl group to form the highly dispersed Td-cobalt(II) sites. Significantly, the presence of [Co(H2O)6]2+ reduced the free energy of dehydroxylation process over the silica support of SBA-15, which created a beneficial condition for the formation of required Td-Co(II) sites. The obtained Co/SBA-15 catalyst from direct H2-reduction process displayed better reactivity than the reported cobalt-based catalysts toward PDH reaction. As a result, the present work provides a simple and effective approach to designing high reactivity of cobalt-based PDH catalysts.

Methods

Catalyst synthesis

The catalysts of Co/SBA-15 were prepared by incipient wetness impregnation method, using Co(NO3)2·6H2O as the precursor. The nominal amount of cobalt in the prepared catalysts was 2, 4, 6wt.%. Typically, the desired weight of Co(NO3)2·6H2O was dissolved in 8 mL of deionized water, 2 g of SBA-15 was added with continuous stirring. Subsequently, the obtained mixture was dried at 100 °С overnight. The drying sample was performed directly in 10% H2/Ar flow at 600 °С for 2 h (Dir-reduction catalyst). Meanwhile, the drying mixture was calcined in air at 600 °С for 5 h to achieve Air-calcination sample, and then was reduced in 10% H2/Ar flow at 600 °С for 2 h to obtain an H2-reduction catalyst.

Catalysts characterization and DFT calculations

Catalysts characterization, which includes powder X-ray diffraction (XRD), quasi in-situ X-ray photoelectron spectroscopy (XPS), H2-temperature-programmed reduction (H2-TPR), Ultraviolet-visible (UV-vis), Transmission electron microscopy (TEM), 1H NMR, in situ NH3-DRIFTS, in situ FT-IR and DFT computation were given in the Supporting Information (SI).

Catalytic tests

The catalytic reactivity of PDH was evaluated in a fixed-bed reactor under atmospheric pressure. In a typical test, 0.4 g of catalyst (sieved between 40 and 60 mesh) was loaded in the center of the reactor. Then, the reaction gas of C3H8 (5 mL/min) and N2 (30 mL/min) was introduced for reactivity evaluation. The reaction was carried out at 600 °С, and the feed and products were analyzed by an online gas chromatograph (Fuli 9790GC) equipped with TCD and FID detectors. The conversion of C3H8 and selectivity of C3H6 were calculated as follows:

$${{{{{\rm{C}}}}}}_{3}{{{{{\rm{H}}}}}}_{8}{{{{\rm{conversion}}}}}(\%)=\frac{{C}_{3}{H}_{8in}-{C}_{3}{H}_{8out}}{{C}_{3}{H}_{8in}}\times 100$$
(1)
$${{{{{\rm{C}}}}}}_{3}{{{{{\rm{H}}}}}}_{6}{{{{\rm{selectivity}}}}}(\%)=\frac{n{C}_{3}{H}_{6}}{n{C}_{3}{H}_{6}+(2/3)n{C}_{2}{H}_{6}+(2/3)nn{C}_{2}{H}_{4}+(1/3)nC{H}_{4}}$$
(2)

where n is the number of moles of hydrocarbons.