滕嘉楠，許光月，傅堯

1中國科學(xué)技術(shù)大學(xué)，中國科學(xué)院城市污染物轉(zhuǎn)化重點(diǎn)實(shí)驗(yàn)室，安徽省生物質(zhì)潔凈能源重點(diǎn)實(shí)驗(yàn)室，合肥 230026

2合肥綜合性國家科學(xué)中心，能源研究所，合肥 230031

1 Introduction

The continuously increasing demand for petroleum-based chemicals has led to the rapid depletion of non-renewable fossil resources, and the massive emission of greenhouse gas CO2has caused a series of environmental problems. Therefore,renewable biomass substitutes have become a hot research topic in recent years. Compared to fossil resources, the structures of biomass-based platform molecules were rich in oxygencontaining groups, and were beneficial for the production of oxygenated chemicals such as carboxylic acid/ester building blocks. Among which, 2,5-furandicarboxylic acid (FDCA),produced from the oxidation of 5-hydromethyl furfural (HMF),was a monomer for new potential renewable polymers alternative to terephthalic acid (TPA). The glass transition temperature, thermal stability, E-modulus and tensile strength of FDCA-based polyesters were similar to those of PTA-based polyesters. Moreover, FDCA-based polyester polyethylene furanoate (PEF) performed better CO2, O2and H2O gas barrier properties than TPA-based polyester polyethylene terephthalate(PET)1-3. Various catalytic systems have been used to catalyze the highly selective oxidation of HMF to FDCA, including supported metals (such as Au, Pd, Pt, Ru) and metal oxides(CuO, Co3O4, MnO2and mixed-metal oxide)4-7. Very recently,Huang et al.8used Au@Pd/C to catalyze the aerobic oxidation of HMF to FDCA, reaching 80 molFDCA·molmetal·h-1productivity under 3 MPa O2at 80 °C for 2 h in the presence of additive NaHCO3.

However, the melting point and boiling point of FDCA were imprecise, and FDCA hardly dissolved in water and most organic solvents9,10. Therefore, it was difficult to purify FDCA through traditional vacuum distillation, recrystallization or other methods, thus limiting the scale of practical application10.Meanwhile, in the process of solution polycondensation, FDCA had poor dispersibility and will be decarboxylated when the temperature reaches 195 °C, which was against the polymerization11. In addition, FDCA was usually synthesized under strong alkaline conditions and required further acidizing,which would generate over-stoichiometric inorganic salt byproducts that caused environmental pollution12.

Instead, dimethyl furan-2,5-dicarboxylate (DMFDCA) had a low boiling point and was soluble in most solvents, so it is easy to separate and purify DMFDCA from reactive mixtures by vacuum distillation10,13. Meanwhile, DMFDCA was more stable under polymerization conditions. For instance, in the transesterification and polycondensation reaction of DMFDCA and ethylene glycol to synthesize PEF, the by-product was CH3OH rather than water, which was more easily separated through volatilization, thus improving the polymerization reaction rate12. Therefore, DMFDCA was an ideal substitute for FDCA and the one-step oxidation synthesis of DMFDCA from HMF was of practical significance. It not only shortened the reaction path, but also avoided the separation process of intermediates, thus saving cost12.

By now, there were a few studies on the direct synthesis of DMFDCA from HMF. The noble metal catalysts were able to activate oxygen and were commonly used in this conversion14,15.Christensen et al.16used TiO2-supported Au to catalyze the oxidation of HMF to DMFDCA, reaching 98% (x, molar fraction) yield of DMFDCA under 4 MPa O2at 130 °C for 3 h.Nakajima et al.17used Au/CeO2to catalyze the oxidation of HMF in the presence of additive Na2CO3, and 95% DMFDCA was obtained under 0.5 MPa O2at 130 °C for 15 h. Corma et al.18reported Au-CeO2nanoparticles (NPs) for catalyzing the conversion of HMF to DMFDCA, and obtained 99% (x)DMFDCA at 130 °C under flowing 10 MPa O2at a speed of 0.33 mL·s-1for 5 h. Kim et al.19used Fe3O4-supported AuPd alloy NPs to catalyze the oxidation of HMF, and the yield of DMFDCA reached 92%. Grassi et al.20catalyzed the oxidation of HMF with Au NPs supported by syndiotactic poly(styrene)-cis-1,4-poly(butadiene) (sPSB), and 80% DMFDCA was obtained under 1.5 MPa O2at 110 °C for 16 h. Our group also used Pd/C + Co(NO3)2+ Bi(NO3)3and PdCoBi/C catalysts to catalyze the oxidation of HMF to DMFDCA under atmospheric oxygen, and the yields of DMFDCA in the two systems were 93% and 96%, respectively10. Besides these high-cost noble metal, some non-noble metal catalysts were also reported on this reaction. Ananikov et al.12used simple MnO2/NaCN to catalyze the oxidation of HMF and obtained DMFDCA with a separation yield of 83%. Kantam et al.13used CuO/m-Al2O3to catalyze the oxidation of HMF to DMFDCA using tert-butylhydrogen peroxide (TBHP) as oxidant and methylation reagent, and obtained 92% DMFDCA. However, these catalytic systems used either toxic co-catalyst or equivalent peroxide, and were not suitable for application.

Nanocarbon has been widely used as heterogeneous catalyst and catalyst carriers due to its high specific surface area,adjustable pore structure and easily modified surface. N-doped carbon modified by metal showed excellent catalytic performance in hydrogenation/hydrogen transfer21,22, redox23,24,oxidative esterification25-30and environmental repair reactions31,32, relating to the uniquely electronic interaction between metal, nitrogen and carbon matrix. In particular, N-doped carbon modified by cobalt showed excellent catalytic performance in oxidation reactions30,33-40. Our group used CoxOy-N@C to catalyze the oxidative esterification of HMF and furfural in the presence of additive K-OMS-29. DMFDCA and methyl 2-furoate were obtained with high selectivity. Cai et al.41reported the use of a hollow yolk shell Co@CN for catalyzing the conversion of HMF to DMFDCA, and obtained 89% DMFDCA at 80 °C for 12 h.

In this work, a series of N-doped porous carbon with bimetallic cobalt and manganese catalyst CoMn@NC were synthesized via one-step co-pyrolysis and were used to catalyze the oxidation of HMF at low temperature under atmospheric pressure. DMFDCA was obtained with high yield by using molecular oxygen as a clean oxidant and methanol as esterification agent. The physical and chemical properties of the catalysts were analyzed by X-Ray diffraction (XRD), N2adsorption-desorption experiment, scanning electron microscopy (SEM), transmission electron microscopy (TEM),high angle ring annular dark-field scanning transmission electron microscopy (HAADF-STEM), high resolution transmission electron microscopy (HRTEM), energy dispersive X-ray spectroscopy (EDX) element mapping, aberrationcorrected HAADF-STEM, inductively coupled plasma optical emission spectrometry (ICP-OES), X-ray photoelectron spectroscopy (XPS), elemental analysis and Raman spectroscopy. The conditions for catalyst synthesize as well as oxidation reaction were optimized. The reaction mechanism was proposed through a series of control experiments combined with literatures. The tolerance, stability, and substrate expansibility of the catalytic system were tested to evaluate the potential for application.

2 Experimental

2.1 Chemicals and materials

All the chemicals involved in the preparation of catalysts were commercially available. Melamine, CoCl2·6H2O (AR),MnCl2·4H2O (AR), HNO3(AR, 65% to 68%), HCl (AR, 36% to 38%) and C2H5OH (AR) were purchased from Sinopharm Chemical Reagent Co. Ltd. PEG-PPG-PEG Pluronic? P123(P123, Mn= 5800) were purchased from Sigma-Aldrich.Materials used in reactions including HMF (99%), 2,5-diformylfuran (DFF, 98%), FDCA (98%), DMFDCA (98%) and formyl-2-methyl-furoate (FMF, 95%) were generously gifted by Hefei Leaf Energy Biotechnology Co., Ltd. Monomethyl furan-2,5-dicarboxylate (MFDCA, 95%) was purchased from Bide Pharmatech Ltd. 5-hydroxymethyl-2-methyl-furoate (HMMF)was synthesized as shown in the Supporting Information.CH3OH (AR), KOH (AR), NaOH (AR), K2CO3(AR), KHCO3(AR) and CH3COONa (AR) were purchased from Sinopharm Chemical Reagent Co. Ltd. Deionized water was purchased from Hangzhou Wahaha Group Co. Ltd.

2.2 Preparation of catalysts

In a typical procedure, 30 mL of C2H5OH was mixed with 4.0 g P123 and 6.0 g melamine with magnetic stirring, named as mixture A. 30 mL of C2H5OH was mixed with 11.0 g CoCl2·6H2O, 6.3 g MnCl2·4H2O (the molar ratios of Co/Mn was 3/2) and 7 mL HNO3with magnetic stirring, named as mixture B. Then the mixture B was added into the mixture A with robust stirring. The resulting mixture was kept stirring for 2 h at room temperature and then was heated to 80 °C in an oil bath (in air)to completely evaporate the C2H5OH. The obtained solid mixture was transferred into a quartz boat and pyrolyzed in a tube furnace under nitrogen atmosphere. The temperature was increased from room temperature to 180 °C and kept for 2 h,increased to 240 °C and kept for 2 h, increased to 800 °C and kept for 1 h, and cooled down to room temperature. The ramping rate was kept as 2 °C·min-1. The obtained black sample was further leached in HCl (1.0 mol·L-1) for 6 h, then was thoroughly washed with deionized water and dried under vacuum overnight.The as-prepared catalyst was denoted as Co3Mn2@NC-800.Similarly, other types of catalysts were prepared and denoted as CoxMny@NC-T, where x/y represented the molar ratios of Co/Mn and T represented the pyrolysis temperature.

2.3 Catalyst characterizations

XRD patterns of the powder catalysts were detected by the multi-function rotating target X-ray diffractometer (SmartLab,Japan), using Cu Kαradiation at 40 kV and 15 mA. The 2θ range was 10° to 70°, and the step length was 0.02°. N2adsorptiondesorption experiments were carried out using MicroActive of ASAP 2460 (United States) at 77.35 K. According to the N2adsorption isotherm, the specific surface areas and pore size distributions of the catalysts were calculated by BET and BJH methods, respectively. SEM images were obtained using GeminiSEM 500 electron microscope (Germany). TEM images were obtained using JEM-F200 electron microscopy (Japan).HAADF-STEM, HRTEM and EDX element mapping were obtained using Talos F200X electron microscopy (United States). The aberration-corrected HAADF-STEM images were obtained using JEM-ARM200F (Japan) equipped with a spherical aberration corrector. Element contents were analyzed by ICP-OES using Varian Vista-MPX (United States). The concentrations of cobalt and manganese were determined according to the characteristic peaks at 228.616 and 257.610 nm,respectively. XPS spectra of the catalysts were obtained by Axis Supra+(Japan) spectrometer under Al Kαirradiation (1486.6 eV). Raman spectra were acquired on a HORIBA Raman spectroscopy (Japan) equipped with CCD detector at 633 nm.

2.4 Atmospheric oxidation of HMF

In a typical reaction, 0.1 mmol of HMF, 10 mg of CoMn@NC, 20% (x) K2CO3and 1 mL of CH3OH were added to a 10 mL Schlenk tube. The tube was purged with O2for 3-4 times and the mixture was stirred at 40 °C for 12 h with an O2balloon. After reaction, the conversion of HMF and the yields of DMFDCA, DFF, HMMF and FMF were determined by GC using biphenyl as internal standard on a Shimadzu GC-2030 Series GC System equipped with a flame-ionization detector with HP-INNOWax column, and the yields of MFDCA and FDCA were determined by Waters e2695 HPLC System on Cosmosil C18 column at 30 °C at a wavelength of 265 nm. The mobile phase was 30% CH3OH and 0.1% phosphoric acid aqueous solution with a rate of 0.6 mL·min-1. The conversion of substrate and yields of products were calculated as the following formulas:

nHMF: mol of HMF before reaction; n 'HMF: mol of HMF after reaction.

2.5 Hot-filtration experiment

In a typical experiment, 0.1 mmol of HMF, 10 mg of Co3Mn2@NC-800, 20% (x) K2CO3and 1 mL of CH3OH were added to a 10 mL Schlenk tube. The tube was purged with O2for 3-4 times and the mixture was stirred at 40 °C with an O2balloon. Co3Mn2@NC-800 was filtered from the reaction solution after 0.5 h, and the product was analyzed by GC and HPLC as shown above. Then, the reaction was continued for 12 h. After reaction, the product was analyzed by GC and HPLC as shown above.

2.6 Metal leaching experiment

In a typical experiment, 0.1 mmol of HMF, 10 mg of Co3Mn2@NC-800, 20% (x) K2CO3and 1 mL of CH3OH were added to a 10 mL Schlenk tube. The tube was purged with O2for 3-4 times and the mixture was stirred at 40 °C for 12 h with an O2balloon. After reaction, the catalyst was collected by filtration, the contents of cobalt and manganese in remaining liquid mixture were determined by ICP-OES.

2.7 Recyclability of catalyst

After the first reaction, the catalyst was separated by filtration and washed with NaOH solution (0.1 mol·L-1) to remove possible protons on the catalyst. After washing with CH3OH to neutral, drying under vacuum overnight, the catalyst was reused for the next run under the same reaction conditions.

3 Results and discussion

3.1 Catalyst screening

A series of CoMn@NC catalysts with various Co/Mn ratio and pyrolysis temperature were synthesized and employed in the oxidation of HMF. To avoid the influence of free or weakly supported metal clusters, all the catalysts were harshly washed by acid. The results were shown in Table 1. HMF couldn’t spontaneously convert to DMFDCA without catalyst (Table 1,entry 1). The metal composition of the catalyst had a great influence on the activity of the catalyst (Table 1, entries 2-6).Although Co@NC gave higher HMF conversion than Mn@NC did, the activities of monometallic catalysts were poor and gave less than 10% yield of DMFDCA (Table 1, entries 2-3). As for the bimetallic catalysts, the catalyst Co3Mn2@NC-800 performed best (Table 1, entries 4-6). HMF fully converted and obtained DMFDCA, HMMF, FMF and MFDCA with yields of 80%, 3%, 1% and 14%, respectively (Table 1, entry 5). It indicated that the synergistic effect of the two metals promoted the activity of catalyst. The pyrolysis temperature caused a volcano-like effect on the yield of DMFDCA (Table 1, entries 5,7-8). The activity of the catalyst slightly reduced in the air atmosphere but still reached 60% yield of DMFDCA (Table 1,entry 9). In addition, the yield of DMFDCA over the Co3Mn2@NC-800 (without P123) and Co3Mn2@NC-800(without HNO3) catalysts prepared without adding P123 and HNO3respectively were obviously poor (Table 1, entries 10-11).As a soft template, P123 in catalyst precursor introduced metals into carbon matrix during the evaporation of metal salt solutions to make the catalyst present ideal morphology. The decomposition of HNO3in catalyst precursors released oxidative gases (such as NO2and NO) at high temperature42, which might contribute to increasing the pore structure in catalyst and further increased the specific surface area of the catalyst.

Table 1 Screening of catalysts a,b.

3.2 Catalyst characterizations

The structural properties of the catalysts were analyzed by XRD analysis first (Fig. 1a). The strong diffraction peak at about 26.5° corresponded to the (002) crystal plane of graphite carbon,which indicated that there could be several layers of graphite carbon in the catalyst. Metallic diffraction peaks were barely observed, indicating that the metals dispersed well in the carbon matrix. According to the ICP-OES results, the actual contents of Co and Mn in Co3Mn2@NC-800 catalyst were 1.42% (w) and 0.33% (w) respectively. The extremely low metal content was also the reason that the metallic diffraction peaks could not be observed in the XRD pattern.

The specific surface area and pore size distribution of a series of catalysts were measured by N2adsorption-desorption experiment at 77.35 K. The nitrogen sorption isotherms of all catalysts showed type IV isotherms and H4 type hysteresis loops.There were considerable amounts of adsorptions in the lowP/P0regions of those bimetallic catalysts, which were caused by the filling adsorption of micropores. The BET specific surface areas(SBET) of Co3Mn2@NC-700, Co3Mn2@NC-800 and Co3Mn2@NC-900 were all very large (1135, 1681 and 1812 m2·g-1, respectively) (Fig. 1b). With the increase of pyrolysis temperature, theSBETincreased greatly, the pore size decreased slightly, and the pore volume increased slightly (Table S1). In contrast, theSBETof Co@NC-800 and Mn@NC-800 were much smaller (393 and 385 m2·g-1, respectively) (Fig. 1b).

Fig. 1 (a) XRD patterns of Co3Mn2@NC-700, Co3Mn2@NC-800, Co3Mn2@NC-900, Co@NC-800 and Mn@NC-800.(b) Nitrogen sorption isotherms of Co3Mn2@NC-700, Co3Mn2@NC-800, Co3Mn2@NC-900, Co@NC-800 and Mn@NC-800.

These results indicated that the adding of bimetallic CoMn was conducive to the pore forming process of carbon and significantly increased theSBETof the catalyst. CoCl2·6H2O and MnCl2·4H2O promoted the decomposition of melamine at high temperature to produce a large number of gas products (e.g.,cyanide and nitrogen), resulting in the formation of a large number of pores inside the catalyst, thus significantly increasing SBET. Co3Mn2@NC-800 had a large SBETand the largest pore volume (0.56 cm3·g-1) which contributed to its good activity43.The mesoporous structure with a large SBETwas conducive to expose more active sites and promote the diffusion of substrates and products in the channel of catalyst44. Typically, mesoporous channels facilitated substrate and/or products (e.g., O2) to enter into and depart from the active sites45,46. In this catalytic system,HMF and oxygen reacted on the metal surface through the porous structure of the N-doped carbon, and then diffused to the external surface of the catalyst through the porous structure29.

Then, the morphologies of catalysts were studied by SEM and TEM images. SEM images showed that the morphologies of catalysts were very different (Fig. 2a-f). It was difficult to find metals in SEM images because their sizes were ultra-small. The carbon matrix of Co3Mn2@NC-700 existed flake-like, and those of Co3Mn2@NC-800 and Co3Mn2@NC-900 existed block-like,and the size of the latter was larger (Fig. 2a-c). The three bimetallic catalysts synthesized with different pyrolysis temperatures all had porous structures. The carbon matrix of the Co3Mn2@NC-800 (without P123) was mostly sintered (Fig. 2f),indicating that P123 acted as a structural guide agent. TEM images of Co3Mn2@NC-700, Co3Mn2@NC-800 and Co3Mn2@NC-900 showed that the size of metal NPs in the catalyst increased with the increase of pyrolysis temperature(Fig. 2g-i). It agreed with the Ostwald ripening effect, in which high temperature induced metals to aggregate to form larger particles. The aggregation of metal NPs in Co3Mn2@NC-900 was the main reason for its low activity. In particular, there were more metal NPs at the edge of the carbon matrix of Co3Mn2@NC-700, while the metal NPs in Co3Mn2@NC-900 were encapsulated in the carbon matrix, which was due to the migration of metal NPs induced by high temperature.

Fig. 2 SEM images of (a) Co3Mn2@NC-700, (b) Co3Mn2@NC-800,(c) Co3Mn2@NC-900, (d) Co@NC-800, (e) Mn@NC-800 and(f) Co3Mn2@NC-800 (without P123); TEM images of(g) Co3Mn2@NC-700, (h) Co3Mn2@NC-800 and (i) Co3Mn2@NC-900.

The optimized catalyst Co3Mn2@NC-800 was further observed by HAADF-STEM and HRTEM, and its elemental composition was analyzed by EDX element mapping. HAADFSTEM image showed metal NPs were encapsulated in a carbon matrix (Fig. 3a). The crystal plane spacing of metal particles was 0.2 nm, corresponding to the (111) crystal plane of cobalt metal.The metal NP was encapsulated in about 15 layers of graphite carbon with layer spacing of 0.3 nm, corresponding to the (002)crystal plane of carbon nitride nanoribbon (Fig. 3b-d). The distribution of C, N, Co and Mn elements in Co3Mn2@NC-800 were analyzed by EDX element mapping. The NPs visible in the carbon matrix were Co, while Mn and N were uniformly dispersed in the carbon matrix (Fig. 3e-h). In addition,aberration-corrected HAADF-STEM images showed that Co existed as NPs (Fig. 3i) and Mn dispersed in the carbon matrix in the form of single-atoms (SAs) (Fig. 3j, highlighted by red circles).

XPS spectra were used to study the electronic interactions between metal, nitrogen and carbon matrix, as shown in Fig. 4.The C 1s spectrum of Co3Mn2@NC-800 could be deconvoluted into four different peaks with binding energies of 284.6, 285.0,286.1 and 288.2 eV, corresponding to C＝C (graphite C), C＝N/C-O, C＝O/C-N, and O-C＝O, respectively, indicating that nitrogen was successfully incorporated into the carbon matrix (Fig. 4a)47. The Co 2p spectrum of Co3Mn2@NC-800 had two main peaks at 780.62 and 796.21 eV, corresponding to Co 2p3/2 and Co 2p1/2, respectively. The corresponding satellite peaks (written as Sat) were at 785.63 and 803.45 eV (Fig. 4b).The peaks were identified as Co2+species doped by nitrogen atoms in the carbon matrix48,49. It proved that Co species coordinated with the doped nitrogen atoms in the carbon matrix and formed the Co-Nx, which was the key factor to improve catalytic performance. The two peaks with binding energies 641.78 and 645.92 eV in Mn 2p spectrum of Co3Mn2@NC-800 corresponded to Mn 2p3/2 of Mn-Nxand its satellite peak,respectively (Fig. 4c)50,51. It indicated that there existed strong coordination between Mn atom and nitrogen doped in carbon matrix to form uniformly dispersed Mn-Nx. The extremely low manganese content and its strong coordination with nitrogen together hindered the aggregation of Mn atoms, resulting in atomically dispersed Mn in N-doped carbon, as shown in EDX element mapping (Fig. 3h). According to the literature, Mn SAs modified N-doped carbon has good structural stability and high adsorption and decomposition activity for O2molecules52.

Fig. 3 (a) HAADF-STEM images of Co3Mn2@NC-800.(b-d) HRTEM images of Co3Mn2@NC-800. (e-h) EDX mapping of composition elements of Co3Mn2@NC-800, C, N,Co and Mn, respectively. (i-j) aberration-corrected HAADF-STEM images of Co3Mn2@NC-800. Mn SAs were highlighted by red circles (color online).

Fig. 4 (a) C 1s XPS spectrum of Co3Mn2@NC-800. (b) Co 2p XPS spectrum of Co3Mn2@NC-800. (c) Mn 2p XPS spectrum of Co3Mn2@NC-800.N 1s XPS spectra of (d) Co3Mn2@NC-700, (e) Co3Mn2@NC-800, (f) Co3Mn2@NC-900, (g) Co@NC-800 and (h) Mn@NC-800.(i) schematic demonstration of the structure and electronic interaction of CoMn@NC revealed by XPS spectra.

As nitrogen atom has one more extranuclear electrons than carbon atom and has a higher electron affinity, the electron cloud near carbon atoms connected to nitrogen atoms had a lower density. Based on the Mott-Schottky effect, there was a significant electron transfer process between metal and N-doped carbon, and metal NPs could give electrons to carbon atoms with higher work function to make their Fermi levels equal1,53. After electrons transferred to the surface of the carbon layer, the electron-deficient metal could easily grab the hydrogen of the substrate, and had a good activation ability for O2molecules to make the metal sites easy to be regenerated26.

The N 1s spectrum of Co3Mn2@NC-800 could be deconvoluted into five different peaks with binding energies of 398.48, 399.76, 400.98, 402.44 and 403.82 eV, corresponding to pyridinic N, M-Nx(M = Co and/or Mn), pyrrolic N, graphitic N and N-Oxspecies, respectively54,55. In addition, there was a weak and wide peak with binding energy of 406.54 eV,corresponding to trace amounts of undecomposed nitrogencontaining precursors (Fig. 4e)56. The contents of various types of nitrogen in the three bimetallic catalysts synthesized with different pyrolysis temperatures were different (Fig. 4d,f, Table S2). As mentioned above, the nitrogen atom regulated the work function of the carbon matrix. In addition, nitrogen atoms,especially pyridinic N, could promote the adsorption of O242.The contents of total N and pyridinic N in Co3Mn2@NC-700,Co3Mn2@NC-800 and Co3Mn2@NC-900 decreased successively (Fig. 4d,f, Table S2), which were consistent with the results of elemental analysis (Table S3). Similarly, the content of M-Nxdecreased as the pyrolysis temperature increased (Table S2), which suggested that the M-N bond partly broke at high temperatures. Because the lone-pairs of pyridinic N atom were in sp2hybrid orbital and did not participate in the conjugation of the ring, so pyridinic N had a strong basicity which was beneficial to the dehydrogenation of alcohol (Fig. 4i). According to the literatures, the hydrogen storage performance of N-doped carbon was related to the amount of nitrogen atom, and it had the best hydrogen storage capacity when the amount of nitrogen atom was the optimal value57,58. The contents of pyrrolic N were in the order of Co3Mn2@NC-900 ＞ Co3Mn2@NC-700 ＞ Co3Mn2@NC-800. As lone-pairs of pyrrolic N participated in the conjugation system of the ring, the electron cloud density of N atom decreased and the electrons on the N-H bond shifted toward the N atom, so pyrrolic N had a certain acidity (Fig. 4i). Moderate acidity could facilitate the methylation of the 2-formyl group of HMF to react with CH3OH to form hemiacetal and in favor of the separation of acidic products from the catalyst. As mentioned above,nitrogen could grapple metals to improve the dispersion of metals and prevent them from aggregation59,60. Li et al.61demonstrated that the strong interaction between N-doped carbon and metal atoms could drive the transformation of nanocatalysts to stable single-atom catalysts (SACs). The monoatomic dispersion of Mn confirmed this conclusion. The contents of total N and pyridinic N of Co@NC-800 and Mn@NC-800 were low (Fig. 4g-h, Table S2), which might lead to poor catalytic activities. The contents of total N and pyridinic N in Co3Mn2@NC-800 (without HNO3) were lower than those in Co3Mn2@NC-800 (Fig. 4e, Fig. S1c and Table S2), indicating that HNO3promoted the formation of pyridinic N.

Fig. 5 Raman spectra of the Co3Mn2@NC-700, Co3Mn2@NC-800,Co3Mn2@NC-900, Co@NC-800, Mn@NC-800, Co3Mn2@NC-800(without HNO3) and Co3Mn2@NC-800 (without P123).

Doping nitrogen into the carbon matrix formed nitrogencontaining pentagon, which destroyed the hexagonal structure of carbon atoms in the graphite plane and resulted in plane bending and a large number of dislocations and defects in the cascade structure62. The defects of graphite structure increased with the increase of the amount of nitrogen63. The defects of the catalysts were evaluated by Raman spectroscopy. The three peaks at 1325,1600 and 2675 cm-1in the Raman spectra corresponded to the D, G and 2D bands of carbon material, respectively (Fig. 5)64,65.Among which, the D band was caused by the vibration of sp3defects and disordered sites in the hexagonal graphite layer, and the G band corresponded to the first order scattering E2gvibration mode of sp2hybrid carbon atoms in two-dimensional hexagonal graphite layer. The ratio of peak intensity of D band to G band (ID/IG) was used to evaluate the degree of graphitization and the number of defects of carbon materials.ID/IGratio close to 0 indicated high crystallinity of carbon materials. Both pyrolysis temperature and doping amount of heteroatom affected the number of defects44,64,66-70. The ID/IGvalues of Co3Mn2@NC-700, Co3Mn2@NC-800 and Co3Mn2@NC-900 were 1.22, 1.23, and 1.15, respectively,indicating that Co3Mn2@NC-800 had more defects but lower graphitization degree (Fig. 5). The disordered and defective carbons increased the active sites71. O2molecules could “cross”the carbon layer to reach the metal surface and be dissociated.The dissociated active species then “crossed” the carbon layer and overflowed to the outer surface. In contrast, Co@NC-800 and Mn@NC-800 had lower ID/IGvalues (1.07 and 1.12,respectively), suggesting that the introduction of heteroatoms and transition metals generally led to more defects, which was consistent with literature44. In particular, the ID/IGvalues of the of Co3Mn2@NC-800 (without HNO3) and Co3Mn2@NC-800(without P123) were also small (1.16 and 1.19, respectively),indicating that HNO3and P123 also promote the formation of defects to a certain extent.

3.3 Reaction conditions optimization

The effect of base strength on HMF oxidation was studied using Co3Mn2@NC-800 as catalyst (Fig. 6). When weak bases sodium acetate, potassium bicarbonate and potassium carbonate were used, the conversion of HMF and the yield of DMFDCA increased with the increase of basicity. It was mainly because the base promoted the dehydrogenation of 5-hydroxymethyl group of HMF. However, the conversion of HMF and the yield of DMFDCA decreased when strong bases KOH and NaOH were used. Therefore, the suitable strength of base was conducive to the oxidation reaction of HMF, and the optimized base was K2CO3.

Fig. 6 Effect of the strength of base.

Fig. 7 Effect of the amount of base.

The effect of base amount on product distribution was then studied (Fig. 7). With the base amount increased, the yield of DMFDCA decreased significantly, while the yield of MFDCA increased significantly at first and then decreased. It was because the higher base amount promoted the disproportionation reaction of 2-formyl group of HMF and intermediate FMF72-74. In addition, the yield of FDCA increased with the increase of amount of base, which indicated that the base not only promoted the oxidation of hydroxyl and aldehyde, but also promoted the saponification reaction10. Therefore, the appropriate amount of base was particularly important for the oxidation of HMF. The optimized amount of base was 20% (x).

The effect of temperature on product distribution was tested in the range of 25 °C to 70 °C (Fig. 8), the yield of DMFDCA first increased and then decreased with the temperature increasing. The yield of HMMF decreased first and then increased. Therefore, the suitable temperature was conducive to the oxidation of HMF, and the optimized temperature was 50 °C.

Fig. 8 Effect of the temperature.

3.4 Catalyst stability and tolerance

In order to confirm that the reaction was indeed catalyzed by heterogeneous CoMn@NC rather than homogeneous cobalt species, hot-filtration and metal leaching experiments were carried out. After reaction for 0.5 h under standard reaction conditions, the conversion of HMF and the yields of DMFDCA and HMMF were 22%, 5% and 9%, respectively. Once Co3Mn2@NC-800 was filtered from the reaction solution, no further increase of conversion of HMF and yields of products could be observed even after the reaction lasted for 12 h. After 12 h, the catalyst was filtered out and the contents of cobalt and manganese in the reaction solution were detected by ICP-OES.The results showed that the content of leached cobalt was lower than the detection limit, and the content of leached manganese was only 0.01% of the initial content. It proved that the catalyst was very stable under reaction conditions and the catalytic site was the heterogeneous catalyst.

Then, the stability of Co3Mn2@NC-800 catalyst was evaluated by recycle experiments. The catalyst could be reused after simple treatment. The results of the two cycles showed that the conversion of HMF was unchanged, while the selectivity of DMFDCA slightly reduced, and the selectivity of HMMF correspondingly increased (Fig. 9a). Furthermore, the composition and structure of the regenerated Co3Mn2@NC-800 catalyst was characterized by XPS (Fig. 9b,c). It was found that the binding energy corresponding to the characteristic peak in the Co 2p spectra and N 1s spectra of the fresh catalyst and the regenerated catalyst kept unchanged, indicating that the form and content of elements in the catalyst were basically unchanged.

Fig. 9 (a) Recycle experiment. Reaction conditions: 0.1 mmol HMF, 10 mg Co3Mn2@NC-800, 20% (x) K2CO3, 1 mL CH3OH, O2 balloon, 40 °C, 12 h.(b) Co 2p and (c) N 1s XPS spectra of fresh and regenerated Co3Mn2@NC-800.

In oxidation and reduction reactions, SCN-ions usually poisoned M-Nx/C catalysts. In order to investigate the tolerance of the catalytic system to SCN-ions, 40% (x) KSCN was added to the reaction system to carry out toxicity experiment. The result showed that the conversion of HMF and the yield of DMFDCA decreased to 86% and 36%, respectively,indicating that SCN-could not completely quench the oxidation of HMF. When the reaction time was extended to 24 h, the conversion of HMF was basically unchanged, but the yield of DMFDCA further increased to 44% (Table S4), indicating that the catalytic system was relatively stable and tolerant to SCN-ions. In addition, the results also indicated that compared with M-Nx, Co NPs encapsulated by carbon matrix contributed more to catalytic activity75.

3.5 Mechanism

In order to identify the reactive oxygen species in the reaction system and further predict the reaction mechanism, 40% (x) free radical scavenger butylhydroxytoluene (BHT) was added to the reaction system. The yield of DMFDCA greatly decreased to 38% at 50 °C, while the yields of HMMF and FMF increased to 15% and 46%, respectively. The result indicated that superoxide radical anion (·O-2) participated in the reaction system.

When DFF and FMF were used as substrates to explore the intermediate, the yields of DMFDCA were 55% and 80% at 50 °C, respectively. Considering that the yield was 85% in HMF conversion at the same conditions, FMF might be the intermediate of the reaction, while DFF was not. When FMF was used as the substrate for experience in the presence of BHT, the conversion of FMF and the yield of DMFDCA decreased. It speculated that ·O-2involved in the conversion of FMF to DMFDCA step.

Based on the above experimental results and literatures, the reaction mechanism was proposed in Fig. 1026,50. The first stage was the conversion process of 2-formyl group of HMF (Fig. 10,stage 1). 2-Formyl group and hydroxyl group of CH3OH were adsorbed on the basic surface of CoMn@NC catalyst76and to form unstable intermediate hemiacetal A with the assistance of pyrrolic N acid group of the catalyst (Fig. 10, step 1). The electron-deficient Co NPs extracted proton of hydroxy from hemiacetal A and adsorbed O2. Hemiacetal A underwent further dehydrogenative oxidation to form HMMF, and O2was sufficiently activated into ·O-2for oxidative deprotonation of the H-metal intermediates to form H2O and regenerate the active site77. Pyridinic N basic group of the catalyst and the additive base facilitated such a deprotonation process (Fig. 10, steps 2-4). The second stage was the conversion process of 5-hydroxymethyl group of HMMF (Fig. 10, stage 2). C-H bond in the alcohol side chain of HMMF was activated over catalyst surface to form the aldehyde intermediate FMF (Fig. 10, steps 5-7). Further reactive processes were similar to the former stage 1 (Fig. 10, steps 8-10) and obtained the desired DMFDCA. In the whole reaction process, the reactive oxygen species ·O-2ensured the high selectivity for DMFDCA through promoting the dehydrogenative oxidation of the hemiacetal and HMMF26,78.

Fig. 10 Proposed reaction mechanism.

3.6 CoMn@NC catalyzed oxidation of substituted aromatic alcohols

In order to investigate the generality of catalyst, various benzyl alcohols with electron-absorbing (-Cl, -NO2) or electron-donating (-OCH3, -CH3) groups, furfuryl alcohols,heterocyclic alcohols were used as substrates and carried out reactions (Table 2). Benzyl alcohols with different substituents quantitatively converted into corresponding methyl esters (Table 2, 2a-2e). The properties of the substituents on the benzene ring seemed to have no obvious influence on the reactivity significantly. When furfuryl alcohol was used as substrate, the yield of methyl furoate was 73% after 72 h (Table 2, 2f), which indicated that the catalytic system was also suitable for other furan-based alcohols. Heterocyclic alcohols such as 2-thiophene methanol and 2-pyridine methanol also oxidized into corresponding methyl esters. The yields of thiophene-2-carboxylic acid methyl ester and pyridine-2-carboxylic acid methyl ester were 95% and 33%, respectively. After prolonging the reaction time to 72 h, the yield of the latter increased to 50%(Table 2, 2g-2h). This confirmed that Co3Mn2@NC-800 resisted N or S atomic poisoning, which was one of the significant advantages compared to noble metal-based catalysts. With diol 1,4-benzenedimethanol as substrate, the oxidation product was obtained with 57% yield (Table 2, 2i). In addition, the product type was further extended from methyl ester to ethyl ester. HMF and 1,4-benzenedimethanol could also react with C2H5OH in this catalytic system, and the yields of the corresponding diethyl ester products were 62% and 33% (72 h), respectively (Table 2,2j-2k).

Table 2 CoMn@NC catalyzed oxidation of substituted aromatic alcohols a.

4 Conclusions

In conclusion, a series of N-doped porous carbon with bimetallic cobalt and manganese catalysts CoMn@NC were synthesizedviaone-step co-pyrolysis, which showed excellent performance for aerobic oxidative of HMF to DMFDCA under atmospheric oxygen at low temperature. The Co/Mn ratio,pyrolysis temperature and synthesis additives were optimized to find Co3Mn2@NC-800 performed the best catalytic activity to reach 85% yield of DMFDCA at 50 °C and ambient oxygen pressure. The carbon nitride encapsulated Co nanoparticles and atomically dispersed Mn both coordinated with N-doped carbon to form M-Nx. The electron-deficient metal site on pyridinic N-rich carbon was beneficial for the activation of both HMF and oxygen. Oxygen formed superoxide radical anion to oxidize the aldehyde group and hydroxymethyl group of HMF. The catalyst was stable and could be expanded to various substituted aromatic alcohols. This catalytic system had applicable potential for the production of carboxylic esters for polymers.

Supporting Information:available free of chargeviathe internet at http://www.whxb.pku.edu.cn.