Epoxidation of bio-derived plant oils is a sustainable route to manufacturing plasticizers, additives in lubricants, and other chemicals. The traditional synthetic approaches suffer from the employment of corrosive mineral acid or expensive peroxides (e.g., H2O2). In this work, we report the epoxidation of plant oils using O2 as the terminal oxidant catalyzed by Co-N-C/SiO2 single-atom catalyst. The single-atom dispersion of cobalt is confirmed by high-angle annular dark field-STEM and x-ray absorption fine structure techniques. In the epoxidation of methyl oleate under mild reaction conditions (35 °C, 0.1 MPa O2), 99% selectivity to the desired product is achieved at full conversion. Even for crude oils, Co-N-C/SiO2 is also effective and good yields of the corresponding epoxides are obtained. In addition, the catalyst is easily recovered and can be reused five times without obvious decay in catalytic activity/selectivity. A superoxide radical involved reaction mechanism is proposed on the basis of kinetic study and EPR experiment.
I. INTRODUCTION
Plant oils (such as palm oil, rapeseed oil, and sunflower oil) represent one of the cheapest and most abundant bio-derived renewable resources.1,2 Common plant oils contain a substantial amount of mono- or multiple unsaturated fatty esters (or acids), where the C=C offer themselves for further functionalization such as hydroformylation, hydroboration, silylation, alkoxycarbonylation, and olefin metathesis for various applications.3 In addition, the characteristic long-chain methylene sequences and bio-based nature (biodegradable, non-toxic, and CO2 neutral) make plant oils ideal feedstock for the manufacture of fine chemicals and synthetic intermediates. In this respect, epoxidized plant oils and the derivate methyl fatty esters are typical examples of the largest industrial applications with an output of hundreds of thousands of tons annually.4,5 They are widely employed directly as plasticizers and stabilizers for chlorine-containing resins (e.g., PVC), additives in lubricants, and components in thermosetting plastics, and they can be good substitutes for some phthalates that are forbidden in EU because of the negative impact on both health and safety.6 In addition, they can be further transformed into chemicals such as polyoils, alkanolamines, and polymers (e.g., polyesters).7,8
Traditionally, epoxidized plant oils are produced on an industrial scale mainly by the Prileshajev peracid process (H2O2/formic acid/mineral acid).9–11 In this approach, the reaction of H2O2 and formic acid catalyzed by mineral acid generates in situ performic acid, which then fulfills the epoxidation transformation subsequently. However, this process suffers from severe side-reactions of ring opening and oligomerization as well as environmental issues due to the use of mineral acid (e.g., H2SO4).11 Great efforts have been devoted to developing efficient and environmentally benign synthetic routes, of which the systems using heterogeneous catalysts and H2O2/TBHP (tert-butyl hydroperoxide) as oxidants have attracted much attention. Various catalysts, such as TS-1,12 TiO2-SiO2,13 and Ti-MCM41,14 have been developed, yet for vegetable oils and methyl esters with larger molecules, the efficiency was deteriorated by mass transfer limitations. In addition, the employment of expensive H2O2/TBHP greatly increases the cost for industrial applications. On the other hand, the epoxidation of plant oil with O2 (or air) as the terminal oxidant is undoubtedly the most ideal one. Swern and co-workers first investigated the epoxidation of oleic acid and methyl oleate using O2 as an oxidant under the irradiation of UV, yet the yield of the desired product was relatively low.15 Köckritz et al. reported that the epoxidation of methyl oleate could proceed smoothly under 0.5 MPa O2 atmosphere when using AIBN (azobisisobutyronitrile) as an initiator.16 In these catalytic systems, the reactions were speculated to follow a radical mechanism, where the superoxide radical was a possible intermediate. Besides UV irradiation and radical initiator, metal catalysts are also able to promote the generation of oxygen-containing radicals. In addition, metal catalysts will offer the opportunity for improving the epoxidation selectivity through surface-mediated reactions. For example, cobalt-containing MOFs were reported to achieve a high catalytic activity and selectivity in alkene epoxidation reactions.17 However, few catalysts dealing with the epoxidation of plant oils using O2 as an oxidant have been reported yet.
Single-atom catalysts have demonstrated excellent catalytic performances in various organic transformations,18–21 including oxidation reactions. On metal single atoms, O2 molecules activated via an end-on mode are intrinsically superoxide radicals and can abstract H atom from substrates such as alcohols. We previously reported the oxidative coupling of primary and secondary alcohols using O2 as an oxidant catalyzed by Co-N-C single-atom catalysts.22 The active intermediates were confirmed to be superoxide radicals according to EPR characterization. In this Communication, we report the epoxidation of plant oils using atomically dispersed Co-N-C/SiO2 as the catalyst and oxygen molecules as the oxidant. Good to excellent yields of the desired products are obtained in the epoxidation of the model compound methyl oleate as well as crude oils. In addition, the catalyst can be reused at least five times without obvious decay in catalytic activity/selectivity.
II. RESULTS AND DISCUSSION
A. Catalyst synthesis and characterizations
The Co-N-C/SiO2 sample was prepared with a similar method to that for the Co-N-C catalyst we previously reported,23 except for using SiO2 gel as the support [Fig. 1(a)]. During the pyrolysis process, the Co(phen)x (phen = 1,10-phenanthroline) complex underwent polymerization and condensation reactions to form Co-N-C carbonaceous materials loaded on SiO2. For comparison, other transition metal catalysts including Fe-N-C, Mn-N-C, and Cu-N-C were also prepared with the same method as for Co-N-C. All these catalysts exhibit similar N2 adsorption–desorption isotherms to that of the silica support (Fig. S1), with BET surface areas in the range of 300–350 m2/g (Table S1), which are just slightly lower than that of pristine SiO2, indicating the little contribution of the carbonaceous M–N–C materials to the textural properties, which is mostly likely due to the few-layered structure of the resultant M–N–C strongly attached to the surface of the SiO2 support in the form of SiO2-C composites. ICP-AES (inductively coupled plasma atomic emission spectrometry) analysis (Table S1) reveals that the metal contents of these samples are very close to 1.0 wt. %.
(a) Schematic illustration for the synthesis of the Co-N-C/SiO2 catalyst, (b) XRD pattern, (c) TEM image, (d) EDS mapping, and (e) AC-HAADF-STEM image of the Co-N-C/SiO2 catalyst.
(a) Schematic illustration for the synthesis of the Co-N-C/SiO2 catalyst, (b) XRD pattern, (c) TEM image, (d) EDS mapping, and (e) AC-HAADF-STEM image of the Co-N-C/SiO2 catalyst.
XRD patterns [Fig. 1(b) and Fig. S2] of the M–N–C catalysts do not show any diffraction peaks assigning to metal-containing species, except for the amorphous SiO2 and/or carbon matrix. Consistently, no M/MOx (M: Co, Fe, Mn, and Cu) nanoparticles are observed either in low-magnification STEM [Fig. 1(c) and Fig. S3] or in SEM images (Fig. S4). The above results indicate that the metal species of the M–N–C/SiO2 samples should be dispersed as tiny clusters or single atoms. In line with these results, the elemental mapping of the Co-N-C sample [Fig. 1(d)] reveals that Co, N, and C are uniformly dispersed in the sample and Co and N signals are overlapped with each other, suggesting that Co species are likely coordinated with N atoms. To identify the dispersion of Co species, the aberration-corrected high-angle annular dark field-STEM technique is employed and the result shows that a high density of cobalt single atoms [bright dots in Fig. 1(e)] is observed to be highly dispersed. When examining different regions of the sample, we also occasionally observe loose aggregates of Co single atoms (Fig. S5). Since no crystalline NPs are formed, the loose aggregates can tentatively be considered as pseudo-single atoms, similar to our previous reports.24
Taking Co-N-C as an example, we further employ x-ray absorption fine structure (XAFS) characterizations to probe into the electronic and coordinative structures of the central Co atoms. Figure 2(a) shows the X-ray adsorption near-edge spectra (XANES) at the Co K edge of the Co-N-C/SiO2 as well as two reference samples. A comparison of the E0 value (the first inflection point) shows that the Co-N-C/SiO2 sample has almost the same E0 value (7718 eV) as Co(OAc)2 does, suggesting an oxidation state of Co(II). As shown in the FT-transformed EXAFS spectra [Fig. 2(b)], the Co-N-C/SiO2 presents a prominent peak at 1.6 Å (not corrected in the phase), which is assigned to the Co-N/O contribution according to our previous work.23 Meanwhile, although a small shoulder peak appears at 2.2 Å (not corrected in the phase), their assignment to the Co-Co scattering path can be safely precluded by the wavelet transform (WT) technique. As shown in Fig. 2(c), Co foil affords a lobe at (2.0 Å, 7.0 Å−1), which is attributed to Co-Co coordination. On the contrary, no lobe at high k value but one at low k values (1.5 Å, 3.5 Å−1) appears in the Co-N-C/SiO2 sample [Fig. 2(d)], suggesting that the central Co atoms are bonded to N/O rather than heavy Co atoms. Taken together, it is concluded that cobalt is dispersed as single atoms in our Co–N-C/SiO2 catalyst. The coordinative structure of Co is determined by EXAFS data fitting. As shown in Table I and Fig. S6, the coordination structure of Co single atoms can be depicted as CoN4-1-2-O2, where the Co-N shell has a coordination number (CN) of 3.6 at a distance of 1.96 Å and the Co-O shell has a CN of 1.9 at a distance of 2.04 Å, respectively, similar to our previous report.23
(a) The normalized XANES spectra at the Co K-edge and (b) the k2-weighted Fourier transform EXAFS spectra in r-space. The wavelet transformation (WT) for the k2-weighted EXAFS signal of (c) Co foil and (d) Co-N-C/SiO2 sample.
(a) The normalized XANES spectra at the Co K-edge and (b) the k2-weighted Fourier transform EXAFS spectra in r-space. The wavelet transformation (WT) for the k2-weighted EXAFS signal of (c) Co foil and (d) Co-N-C/SiO2 sample.
The best-fitted EXAFS results for the Co-N-C/SiO2 catalyst.a
| σ2 | ∆E0 | r-factor | ||||
|---|---|---|---|---|---|---|
| Sample | Shell | CN | R (Å) | (10−2 Å2) | (eV) | (%) |
| Co foil | Co-Co | 12 | 2.48 | 0.6 | 5.0 | 1 |
| Co-N-C/SiO2 | Co-N | 3.6 | 1.96 | 2 | −2.3 | 0.2 |
| Co-O | 1.9 | 2.04 | 0.7 |
| σ2 | ∆E0 | r-factor | ||||
|---|---|---|---|---|---|---|
| Sample | Shell | CN | R (Å) | (10−2 Å2) | (eV) | (%) |
| Co foil | Co-Co | 12 | 2.48 | 0.6 | 5.0 | 1 |
| Co-N-C/SiO2 | Co-N | 3.6 | 1.96 | 2 | −2.3 | 0.2 |
| Co-O | 1.9 | 2.04 | 0.7 |
CN is the coordination number for the absorber–backscatterer pair, R is the average absorber–backscatterer distance, σ2 is the Debye–Waller factor, and ∆E0 is the inner potential correction. The accuracies of the above parameters are estimated as CN, ±20%; R, ±1%; σ2, ±20%; and ∆E0, ±20%. The data range used for data fitting in k-space (∆k) and R-space (∆R) are 3.0–12 Å−1 and 1.2–2.0 Å, respectively.
To probe the chemical state of Co single atoms, XPS characterization was performed. As shown in Fig. 3(a), the Co 2p3/2 XPS spectrum shows a prominent peak with a binding energy of 781.6 eV. This, together with the appearance of a satellite peak, is characteristic of Co(II) species.23 Meanwhile, there appear three chemical states of N at 398.6, 399.7, and 400.7 eV in N 1s XP spectrum [Fig. 3(b)], which is attributed to pyridinic N, Co-N, and pyrrolic N, respectively.25 In particular, in the Co-N moiety, the Co(II) species are strongly coordinated with N atoms doped in the carbonaceous matrix, and the as-formed structure is robust enough to resist aggregation during high-temperature pyrolysis.
B. Catalytic performances
The epoxidation of methyl oleate has important applications in producing eco-friendly plasticizers. Here, we for the first time apply the noble-metal-free M–N–C SACs for this reaction. As shown in Table II, the reaction could not proceed without any catalyst (Table II, entry 1). The NC/SiO2 sample without any metal species as well as the Co(OAc)2 sample did not catalyze the reaction either (Table II, entries 2 and 3). For the M–N–C/SiO2 catalysts with different active metals, Co-N-C/SiO2 afforded 49% conversion with 99% selectivity to the desired product at a reaction time of 1 h (Table II, entry 4), whereas the other M–N–C (M: Fe, Mn, Cu) catalysts gave either inferior conversion or selectivity (Table II, entries 5–7), demonstrating that the Co-N-C SAC is the best one among them. Indeed, when the reaction time was extended to 5 h (Table II, entry 8), the conversion over Co-N-C could reach 99% while the selectivity was still as high as 99%, manifesting the promising potential for practical applications. To show the unique catalytic function of the CoN4 moiety of the Co-N-C catalyst, we prepared several control samples including Co/NC/SiO2, Co/SiO2, and Co/NC nano-catalysts by incipient wetness impregnation followed by hydrogen reduction at 350 °C and tested their catalytic performances at the same Co/substrate ratio. The results (Table II, entries 9–11) show that the three nano-catalysts are all effective for the reaction, yet the conversion and selectivity are much lower than those of Co-N-C/SiO2, demonstrating that the single-atom dispersed CoN4 structure is superior to the Co NP counterparts in terms of both activity and selectivity. The effect of the support was also investigated (Table S2). When SiO2 was replaced by basic MgO, the conversion was greatly decreased, whereas the selectivity maintained high (Table S2, entry 2). By contrast, when Lewis acidic TiO2 or Al2O3 was used as the support, both conversion and selectivity declined to some extent (Table S2, entries 3 and 4). Interestingly, when the SiO2 support in Co-N-C/SiO2 was removed by acid etching, the atom-specific activity of Co-N-C/SiO2-leaching (mass loading of cobalt: 3.2 wt. %) was decreased (Table S2, entry 5), indicating that the interaction between cobalt single atoms and the SiO2 support is beneficial for the epoxidation reaction.26
Catalytic performances of M–N–C/SiO2 samples for the epoxidation of methyl oleate.a
 | |||
|---|---|---|---|
| Conv. (%) | Select. (%) | ||
| Entry | Catalysts | 1a | 2a |
| 1 | No catalyst | n.d. | n.d. |
| 2 | NC/SiO2 | n.d. | n.d. |
| 3 | Co(OAc)2 | n.d. | n.d. |
| 4 | Co-N-C/SiO2 | 49 | 99 |
| 5 | Fe-N-C/SiO2 | 38 | 82 |
| 6 | Mn-N-C/SiO2 | 55 | 76 |
| 7 | Cu-N-C/SiO2 | 27 | 97 |
| 8b | Co-N-C/SiO2 | 99 | 99 |
| 9 | Co/NC/SiO2 | 41 | 87 |
| 10 | Co/SiO2 | 29 | 71 |
| 11 | Co/NC | 32 | 84 |
| |||
|---|---|---|---|
| Conv. (%) | Select. (%) | ||
| Entry | Catalysts | 1a | 2a |
| 1 | No catalyst | n.d. | n.d. |
| 2 | NC/SiO2 | n.d. | n.d. |
| 3 | Co(OAc)2 | n.d. | n.d. |
| 4 | Co-N-C/SiO2 | 49 | 99 |
| 5 | Fe-N-C/SiO2 | 38 | 82 |
| 6 | Mn-N-C/SiO2 | 55 | 76 |
| 7 | Cu-N-C/SiO2 | 27 | 97 |
| 8b | Co-N-C/SiO2 | 99 | 99 |
| 9 | Co/NC/SiO2 | 41 | 87 |
| 10 | Co/SiO2 | 29 | 71 |
| 11 | Co/NC | 32 | 84 |
Reaction condition: 0.5 mmol methyl oleate, M–N–C/SiO2 catalyst (M/methyl oleate = 1:100, molar ratio), 2 mmol n-butyraldehyde, 4 ml acetonitrile, O2 bubbling, 35 °C, 1 h, and dodecane was used as an internal standard.
Reaction time: 5 h; n.d.: not detected.
The Co-N-C/SiO2 SAC also shows an excellent catalytic performance for other plant oil substrates. As shown in Table III, for plant oils with different lengths of methylene sequences, the epoxidation reaction can proceed smoothly, and the desired products are obtained with good to excellent yields. In addition, the Co-N-C/SiO2 SAC could be facilely recovered by filtration or centrifugation. After being rinsed by ethanol, the catalyst was subjected to another batch of reactions and was reused five times without decay in conversion/selectivity [Fig. 4(a)]. Both XAFS [Figs. 4(b) and 4(c)] and XPS (Fig. S7) spectra of the used Co-N-C/SiO2 catalyst nearly overlapped with those of the fresh catalyst, implying that the coordinative structure and chemical state of Co single atoms remain unchanged after the reaction, demonstrating the high durability of the catalyst.
Catalytic performances of the Co-N-C/SiO2 sample for the epoxidation of various fatty ester and crude vegetable oils.a
| Conv. (%) | Select. (%) | ||
|---|---|---|---|
| Entry | Substrate | 1a | 2a |
| 1 | Camellia oil | 85 | 87 |
| 2 | Sunflower oil | 76 | 72 |
| 3 | Grape seed oil | 71 | 81 |
| 4 | Methyl ricinoleate | 81 | 83 |
| Conv. (%) | Select. (%) | ||
|---|---|---|---|
| Entry | Substrate | 1a | 2a |
| 1 | Camellia oil | 85 | 87 |
| 2 | Sunflower oil | 76 | 72 |
| 3 | Grape seed oil | 71 | 81 |
| 4 | Methyl ricinoleate | 81 | 83 |
Reaction condition: 0.5 mmol substrate, Co-N-C/SiO2 catalyst (M/substrate = 1:100, molar ratio), 2 mmol n-butyraldehyde, 4 ml acetonitrile, O2 bubbling, 35 °C, 5 h, and dodecane was used as an internal standard.
(a) The recycling experiments, (b) the normalized XANES spectra at the Co K-edge, and (c) the k2-weighted Fourier transform EXAFS spectra of fresh and used Co-N-C/SiO2 catalysts. Reaction condition: 0.5 mmol methyl oleate, Co-N-C/SiO2 catalyst (molar ratio of Co:methyl oleate = 1:100), 2 mmol n-butyraldehyde, 4 ml acetonitrile, O2 bubbling, 35 °C, 1 h, and dodecane was used as an internal standard.
(a) The recycling experiments, (b) the normalized XANES spectra at the Co K-edge, and (c) the k2-weighted Fourier transform EXAFS spectra of fresh and used Co-N-C/SiO2 catalysts. Reaction condition: 0.5 mmol methyl oleate, Co-N-C/SiO2 catalyst (molar ratio of Co:methyl oleate = 1:100), 2 mmol n-butyraldehyde, 4 ml acetonitrile, O2 bubbling, 35 °C, 1 h, and dodecane was used as an internal standard.
C. Reaction mechanism
The control experiment result showed that the epoxidation reaction did not proceed in the absence of n-butyraldehyde, indicating that the aldehyde was indispensable for the reaction. To investigate the reaction mechanism, we performed the kinetic study and EPR (electron spin resonance spectroscopy) characterization. The conversion–time profile [Fig. 5(a)] showed that the yield of epoxidized methyl oleate was always accompanied by the consumption of n-butyraldehyde to give butyric acid, suggesting that n-butyraldehyde probably served as a mediator in the epoxidation reaction. In addition, EPR experiments using DMPO (5,5-dimethyl-1-pyrroline N-oxide) as the spin-trapping agent clearly detected the ROO peroxide radical when the Co-N-C/SiO2 catalyst was present in the catalytic system [Fig. 5(b)]. By contrast, no such radical was detected without the Co-N-C/SiO2 catalyst, suggesting that the Co-N-C/SiO2 catalyst was critical for the production of the ROO peroxide radical and the reaction might follow an aldehyde-mediated peracid reaction pathway. Based on the experimental results and literature report,27 a reaction mechanism is proposed in Scheme 1. The oxygen molecule is first adsorbed and activated on the Co single atom, which then reacts with n-butyraldehyde to give butyric acid peroxide, followed by the epoxidation of the C=C bond in plant oils to afford the desired product and butyric acid. The produced butyric acid peroxide might also be adsorbed and stabilized on the single-atom Co species rather than being released into the solvent, which is responsible for the high selectivity of the desired products.
(a) Time profiles of epoxidation of methyl oleate using Co-N-C/SiO2 as the catalyst; (b) DMPO spin-trapping EPR experiment. Reaction condition: 0.5 mmol methyl oleate, Co-N-C/SiO2 catalyst (Co/methyl oleate = 1:100, molar ratio), 2 mmol n-butyraldehyde, 4 ml acetonitrile, O2 bubbling, 35 °C, defined time, and dodecane was used as an internal standard. The EPR spectrum was recorded after 5 min upon introducing the reaction mixture to the DMPO solution (50 mg/ml in DMSO solvent).
(a) Time profiles of epoxidation of methyl oleate using Co-N-C/SiO2 as the catalyst; (b) DMPO spin-trapping EPR experiment. Reaction condition: 0.5 mmol methyl oleate, Co-N-C/SiO2 catalyst (Co/methyl oleate = 1:100, molar ratio), 2 mmol n-butyraldehyde, 4 ml acetonitrile, O2 bubbling, 35 °C, defined time, and dodecane was used as an internal standard. The EPR spectrum was recorded after 5 min upon introducing the reaction mixture to the DMPO solution (50 mg/ml in DMSO solvent).
Proposed reaction pathway for the epoxidation of methyl oleate over the Co-N-C/SiO2 catalyst.
Proposed reaction pathway for the epoxidation of methyl oleate over the Co-N-C/SiO2 catalyst.
III. CONCLUSION
In summary, we have developed Co-N-C/SiO2 single-atom catalysts for the epoxidation of renewable plant oils using O2 as the terminal oxidant and n-butyraldehyde as the reaction mediator. Good to excellent yields of the epoxidized plant oils are achieved under very mild reaction conditions. Compared to those supported Co NP catalysts, the Co-N-C/SiO2 SAC exhibits superior activity, selectivity, and stability, which can be attributed to the unique function of the Co(II)N4 structure for activating oxygen molecules. This single-atom strategy can be expected to find more applications in sustainable chemistry.
SUPPLEMENTARY MATERIAL
See the supplementary material for the detailed preparation of catalysts and additional experimental characterization data.
ACKNOWLEDGMENTS
This work was supported by the National Natural Science Foundation of China (Grant Nos. 21690080, 21690084, 21673228, 21721004, and 21878289), the Strategic Priority Research Program of the Chinese Academy of Sciences (Grant No. XDB17020100), and the Dalian National Laboratory for Clean Energy (DNL) Cooperation Fund, the CAS (Grant No. DNL 180303). We also thank the BL 14W beamline at the SSRF.
DATA AVAILABILITY
The data that support the findings of this study are available within the article.
