The rapid expansion in biodiesel production, of which glycerol is a by-product, has led to a serious surplus of glycerol [1, 2]. Approximately 950000 tons of glycerol are produced per year in the United States and Europe, and it is projected that production will be six times greater than demand by 2020. Some glycerol is burned for energy, but its heat value is low. Because of the glycerol glut, much effort is being made to convert glycerol to valuable products such as propanediol, acrolein, dihydroxyacetone (DHA), glyceric acid (GLYA), and glycerol carbonate [3-9]. Selective oxidation of glycerol is a hot topic and has been widely studied over Au-and Pt-based catalysts in the past decade.

Many Au-based catalysts for selective oxidation of glycerol in a strongly basic solution (NaOH/glycerol molar ratio > 2) have been reported. These catalysts have high activities and selectivities for GLYA [9-18]. However, the NaOH in the reaction mixture catalyzes cleavage of C-C in glycerol [17], and the oxygen atoms incorporated into GLYA over Au catalysts mainly originate from water used as the solvent for NaOH [18, 19].

For practical applications, selective oxidation of glycerol without addition of bases is preferable. The pioneering studies performed by Kimura et al. [20] and Gallezot [21] showed that glycerol can be oxidized by molecular oxygen over Pt-based catalysts in a base-free aqueous solution. It was later reported that H-mordenite-supported Au-Pt [7, 22] and Mg (OH)₂-supported Au-Pt and Au-Pd nanoparticles (NPs) [23] also catalyze glycerol oxidation to free GLYA, without addition of a base, and CuO-supported Au NPs are active in the selective oxidation of glycerol to DHA without NaOH [15]. More recent studies have shown that supported cuboctahedral Pt [24] and Pt₉Sn₁ [25] NPs are more active than tetrahedral Pt in the selective oxidation of glycerol. In addition, it has been suggested that the support properties significantly affect the product selectivity [26], and weaker basic sites can increase the selectivity for GLYA and prevent its deep oxidation [27]. Nitrogen-doped supports [28, 29] and mesoporous carbon nitride [30] further improve the performance and stability of Pt NPs as a result of electron donation from nitrogen to Pt.

Previous studies performed in our laboratory showed that Pt catalysts supported on micropore-free multi-walled carbon nanotubes (MWCNTs) and carbon nanofibers were more active than traditional Pt/carbon in the selective oxidation of glycerol because of easier access to Pt on the outer wall of the support [31, 32]. However, Pt catalysts suffer from deactivation because of weak metal-support interactions, over-oxidation of surface metal NPs to oxides, and/or decarboxylation of aldehyde intermediates to form CO, which strongly adsorbs on Pt [33].

Graphene is a two-dimensional single-layer material consisting of sp²-hybridized carbon. It has a large theoretical specific surface area (up to 2600 m²/g) [34] and use of the surface area is highly efficient because both sides of the graphene are accessible. Functional groups such as nitrogen dopants in the graphene framework can interact electronically with anchored metal particles and strengthen the metal-support and/or metal-support-reactant interactions to achieve high dispersion in the reaction mixture [29, 35, 36]. It has been confirmed that graphene and reduced graphene oxide are good supports for metal NPs [34, 37], and that graphene-based catalysts can be easily dispersed in the reaction mixture [38].

In this study, MWCNTs-pillared nitrogen-doped graphene (NG) was prepared via direct pyrolysis of melamine on MWCNTs. NG grew smoothly on the MWCNT surfaces during melamine pyrolysis. This composite, i.e., NG-MWCNTs, provides an excellent support for ultrafine dispersed Pt NPs, and a Pt/NG-MWCNTs catalyst was highly active and selective in the oxidation of glycerol to GLYA in a base-free aqueous solution.

MWCNTs (purity > 97%, diameter 40±10 nm, and length less than 2 μm) were purchased from the Shenzhen Nanotech Port Co., Ltd. (China). The MWCNTs were pretreated in a mixture of nitric acid (65-68 wt%) and sulfuric acid (98 wt%) (1:1, V/V) at 80 ℃ for 5 h. The pretreated MWCNTs were separated by filtration, washed with distilled water until the pH of the effluent was 6-7, and dried under vacuum overnight; this product is denoted by H-MWCNTs.

H-MWCNTs (2 g) were dispersed in deionized water with poly (vinyl pyrrolidone) (PVP, PVP/H-MWCNTs mass ratio=1:1) under stirring. Melamine (melamine/H-MWCNTs mass ratio=10:1) was added to the suspension and the mixture was stirred for 1 h at room temperature. The solvent was removed from the mixture, using a rotary evaporator, at 10 kPa and 60 ℃ for 1 h. The resulting solid was transferred to a quartz tube and pyrolyzed at 700 ℃ in a N₂ flow for 2 h; the final product is denoted by NG-MWCNTs. As a reference, bare melamine was also pyrolyzed using the same procedure but without the addition of H-MWCNTs; the product is denoted by CN_x.

NG-MWCNTs (1 g) were immersed in ethylene glycol (EG, 99.5%, 50 mL, Sinopharm Chemical Reagent) under stirring, and pretreated ultrasonically for 1 h to give a homogeneous suspension. An aqueous solution of H₂PtCl₆ (0.01 g-Pt/mL, 4.0 mL) was added, and the pH of the mixture was adjusted to 11-12 using NaOH/EG solution. The mixture was then heated in a microwave oven (Sineo, MAS-Ⅱ, 600 W, 2.45 GHz) at 145 ℃ for 7.5 min. The solid product was isolated, washed with distilled water until free of Cl^-, and dried under vacuum overnight at 40 ℃. As references, MWCNT-and CN_x-supported Pt catalysts were prepared using the same method as for Pt/NG-MWCNTs. The Pt contents of the prepared catalysts were determined using inductively coupled plasma-atomic emission spectroscopy (ICP-AES; Plasma-Spec-Ⅱ spectrometer).

X-ray diffraction (XRD) was performed using a RIGAKUD/MAX2550/PC diffractometer with Cu K_α radiation at 40 kV and 100 mA. The diffraction data were recorded by continuous scanning at a rate of 0.02°/s with a step of 0.02°. N₂ adsorption was performed at-196 ℃ using an auto-adsorption analyzer (Micromeritics, TriStar Ⅱ). The Brunauer-Emmett-Teller (BET) surface areas and Barrett-Joyner-Halenda (BJH) pore size distributions were calculated from the desorption isotherms. Scanning electron microscopy (SEM) was performed using a JEOL (JSM6700F) scanning electron microscope. The morphologies and dimensions of the catalysts were determined using transmission electron microscopy (TEM; JEOL-2010F) at an accelerating voltage of 200 kV. X-ray photoelectron spectroscopy (XPS) was performed using a Perkin-Elmer PHI ESCA system, with a Mg standard anode 146 (1253.6 eV) as the electron source, at 12 kV and 300 W. The binding energy of the C 1s peak, i.e., 284.5 eV, was used as the standard for calibration.

Glycerol oxidation was performed in a 25 mL custom-designed stainless-steel autoclave with a glass inner layer, using glycerol solution (0.1 g/mL, 10 mL) and a catalyst (0.1 g). The reactor was sealed, filled with 0.5 MPa of O₂, and placed in an oil bath preheated to the required temperature; the temperature was maintained for a given time under vigorous stirring with a magnetic stirrer (MAG-NEO, RV-06M, Japan). After the reaction, the catalyst was removed by filtration and the aqueous solution was analyzed using an Agilent 1100 series high-performance liquid chromatography system equipped with a refractive index detector and a Zorbax SAX column (4.6 mm × 250 mm, Agilent).

Recycling experiments were performed in a 50 mL custom-designed stainless-steel autoclave with a glass inner layer, using glycerol solution (0.1 g/mL, 20 mL) and a catalyst (0.2 g). After reaction for 6 h, the catalyst was isolated, washed with distilled water, and dried overnight in a vacuum oven at 40 ℃. The glycerol/Pt ratio in the feed was controlled to 581 in each cycle.

Fig. 1 shows the N₂ adsorption-desorption isotherms of CN_x (i.e., the pyrolysis product of melamine without MWCNTs), MWCNTs, and NG-MWCNTs. The pristine MWCNTs gave a type Ⅳ isotherm with an H1 hysteresis loop at higher pressure because of condensation of N₂ in voids among the aggregated MWCNTs. CN_x and the NG-MWCNTs gave type Ⅳ isotherms and H3 hysteresis loops at lower pressure. The amount of N₂ adsorbed by the NG-MWCNTs in the low p/p₀ was greater than that adsorbed by the MWCNTs, indicating the presence of mesopores and a large proportion of wide slit-like pores derived from the gaps between the MWCNT-pillared graphene layers. The sorption isotherms were used to calculate the BET surface areas (A_BET), pore volumes, pore diameters, and size distributions; the results are summarized in Table 1. The NG-MWCNTs surface area (173 m²/g) and pore volume (0.50 cm³/g) were higher than those of the MWCNTs and CN_x. Unlike the MWCNTs and CN_x, the NG-MWCNTs had a bimodal pore size distribution. The large pores, i.e., around 42.6 nm, can be attributed to the gaps between MWCNT-pillared NG layers, and the other pore channels (around 4 nm) could be derived from the accumulation of separated NG layers. This improvement in the hierarchical porous structure (including mesopores and macrospores) and the larger surface area increase contact between the active sites and the substrate.

SEM images show that the pristine MWCNTs consisted of tubes of diameter 20-50 nm (Fig. 2(a)), and CN_x had an irregular morphology with loose macropore channels (Fig. 2(b)). Film-like graphene sheets can be observed clearly on the surfaces of the MWCNTs in the NG-MWCNTs (Fig. 2(c)). The XRD patterns of the Pt/MWCNTs, Pt/CN_x, and Pt/NG-MWCNTs are shown in Fig. 3. The strong (002) diffraction peak at 2θ≈26°, corresponding to a graphite structure, can be seen in all the samples, but the angles are higher for Pt/CN_x and the Pt/NG-MWCNTs than for the Pt/MWCNTs (see inset in Fig. 3). These results indicate the generation of skeletal NG during thermal pyrolysis of melamine. This is accompanied by a decrease in the average interlayer distance to smaller d values because the nitrogen atoms, which are smaller than carbon atoms, compact the carbon sheets in CN_x. Also, the higher electronegativity of nitrogen may result in enhanced interactions between CN_x layers. The diffraction peaks of Pt in Pt/CN_x were broad, and these peaks became weaker and obscured in the Pt/MWCNT and Pt/NG-MWCNT patterns, indicating high Pt dispersions on these supports. The average crystalline sizes of Pt, calculated from the half-width of Pt (111), in the Pt/MWCNTs, Pt/CN_x, and Pt/NG-MWCNTs were 2.2, 3.0, and 1.2 nm, respectively.

XPS spectra of the Pt/MWCNTs, Pt/CN_x, and Pt/NG-MWCNTs are shown in Fig. 4. The spectra show that that the nitrogen contents of Pt/CN_x and Pt/NG-MWCNTs were 15.6 and 5.3 at%, respectively (Table 2).

The main C 1s peak for the nitrogen-free MWCNTs, located at approximately 284.5 eV, with a tail peak at a 288.7 eV, is assigned to electron-deficient carbons bound to oxygen (C-O/C=O). The main C 1s peak shifted to higher energy (284.9 eV) and the tail peak signal was strengthened in the NG-MWCNTs spectrum (Fig. 5). The enhanced tail peak at 289.0 eV can be attributed to incorporation of nitrogen species into the carbon skeleton [14, 35]. The N 1s spectra of Pt/CN_x and the Pt/NG-MWCNTs were broad and asymmetric (Fig. 6), and these peaks can be deconvoluted to four well-fitted peaks at 398.5, 400.1, 401.1, and 403.3 eV; these are attributed to pyridinic nitrogen (N1), pyrrolic nitrogen (N2), quaternary nitrogen (N3), and pyridine N-oxide nitrogen (N4), respectively[29, 35, 39]. The most active nitrogen species in the oxygen reduction reaction (N1, pyridinic nitrogen) [35] was predominant in both the NG-MWCNTs (84.3%) and CN_x (87.1%) (see Table 3).

High-resolution Pt 4f spectra (Fig. 7) show that the proportions of metallic Pt on the CN_x (78.1%) and NG-MWCNT (68.8%) surfaces were higher than that on the pristine MWCNTs (64.2%), and the Pt signals shifted to lower binding energies with increasing nitrogen content. This was attributed to the strong electron-donating effects of nitrogen dopants in carbon skeletons [29].

The morphology and microstructure of the Pt/NG-MWCNTs were unambiguously identified using TEM. Bare CN_x consisted of a thick plate with bulk carbon on the surface. MWCNT-pillared transparent NG with crumpling and agglomeration was clearly observed in the Pt/NG-MWCNTs, indicating that the MWCNTs act as a scaffold in exfoliation of the bulk CN_x sheet to give graphene during melamine pyrolysis. Pt NPs were well dispersed on these supports, and the calculated mean particle sizes of Pt based on 500 particles in the Pt/MWCNTs, Pt/CN_x, and Pt/NG-MWCNTs were 2.4±0.8, 2.9±0.9, and 1.4±0.4 nm, respectively. Scanning TEM (STEM) and linescans of the Pt/NG-MWCNTs (Fig. 8(e) and (f)) show good spatial correspondence between the carbon and nitrogen elemental maps and the intensities of their K lines, indicating that nitrogen is incorporated throughout the entire hybrid. These MWCNTs in NG-MWCNTs can act as nanoscale spacers to generate mesopores (as shown in Fig. 8(c) and (d)) and prevent restacking of NG.

The activities and product distributions in selective oxidation of glycerol over these catalysts in base-free aqueous solution at 60 ℃ and 3 h are summarized in Table 4. The data show that GLYA was the main product in all the experiments. The glycerol conversions over the Pt/MWCNTs and Pt/CN_x were 29.0% and 31.6%, respectively. When the Pt NPs were supported on NG-MWCNTs, the glycerol conversion increased to 64.4%. Time-on-stream data shows that the glycerol conversion over the Pt/NG-MWCNTs increased continuously, but the GLYA selectivity (~80%) was constant (see Fig. 9). The specific activities of the surface metallic Pt atoms were calculated; the results are summarized in Table 4. The calculated turnover frequency (TOF) at 3 h over the Pt/NG-MWCNTs was 267.2 h^-1, i.e., higher than those over Pt/CN_x (211.4 h^-1) and the Pt/MWCNTs (221.7 h^-1). The higher activity of the Pt/NG-MWCNTs is attributed to its mesoporous structure and highly dispersed Pt NPs. High dispersion of Pt NPs provides more active sites for catalytic oxidation, and the highly mesoporous structure provides easy access of reactants to active sites. The nitrogen-rich surface nitrogen results in good hydrophilicity, which improves catalyst dispersion in water and wettability of the substrate on the support, resulting in acceleration of glycerol oxidation.

The Pt/NG-MWCNT catalyst can be easily recovered and reused five times with only a slight decrease in the glycerol conversion, i.e., from 64.4% to 61.2%, in the fifth cycle (Fig. 10). The GLYA selectivity was stable. Deactivation was probably caused by leaching of Pt and nitrogen species from the catalyst.

MWCNT-pillared NG was prepared by direct pyrolysis of melamine on MWCNTs. The MWCNTs acted as a scaffold during graphene formation from melamine. This hybrid had a high surface area and bimodal pore channels. NG-MWCNTs provided a superior support for Pt NPs. The resulting Pt/NG-MWCNTs catalyst was highly active in selective glycerol oxidation, and was recyclable. This high performance can be attributed to the increased surface area, and highly dispersed and electron-sufficient Pt NPs; all these effects were the result of doping with nitrogen. This synergistic strategy provides a simple, efficient, and versatile blueprint for the low-cost fabrication of nitrogen-doped carbon materials and will extend their use in catalysis.