Biodiesel is a renewable, biodegradable fuel manufactured domestically from vegetable oils, animal fats, or recycled restaurant grease. It can be used in pure form or it may be blended with petroleum diesel at any concentration in most injection pump diesel engines. Glycerol is formed as a byproduct during the production of biodiesel, and the amount of glycerol produced has increased quickly with the rapid expansion of biodiesel applications [1, 2]. It was estimated that the production of glycerol in 2020 will be six times higher than demand [2].

In the past decade, much effort has been devoted to developing processes to convert glycerol into valuable products [3-7]. Among these processes, selective oxidation of glycerol with molecular oxygen over heterogenous catalysts to form dihydroxyacetone (DHA), glyceric acid (GLYA), glyceraldehyde (GLYHD), hydroxypyruvic acid (HPYA), and lactic acid has become a hot topic both in academic research and industrial applications [8-13]. Based on previously published works, it was generally concluded that the oxidation of glycerol with molecular oxygen is a structure-sensitive process, in which the reaction mechanism and product distribution vary significantly with different conditions (pH, temperature, and substrate to metal ratio) [11-22].

The use of carbon-supported Pt catalysts has been widely reported in the base-free aqueous liquid-phase oxidation of glycerol [23, 24]. But the deactivation of nano-sized Pt catalysts has often been reported during the selective oxidation of alcohols due to the weaker interactions between Pt and the carbon support, over-oxidation of surface Pt nanoparticles (NPs), and/or the strong adsorption of organic carboxylic acids[23-25].

Recently, it was recommended that adding and/or alloying Pt with another metal is an effective approach to preventing the deactivation of Pt NPs. For example, Au- [15, 26-28], Bi- [8, 9, 29], Ru- [30], Sn- [31, 32], Ni- [33], Co- [34] and Sb- [35, 36] promoted or alloyed Pt catalysts have been used for selective oxidation in the liquid phase. At the same time, it was also reported that N atoms in the framework of the carbon support can donate electrons to Pt, thereby enhancing the interactions between the support and the metal, and thus improving the dispersion, activity, and stability of Pt, and/or generating new active sites[24, 37-39].

Previous work in our laboratory showed that a PtSb alloy supported on multi-walled carbon nanotubes (MWCNTs) was highly active for the selective oxidation of glycerol in base-free aqueous solutions and that a PtSb alloy can hinder the cleavage of C–C bonds [35]. Recent works published by other groups have also confirmed that a Pt-Sb alloy was both more active and selective for the formation of DHA [36, 40]. But the fabrication of PtSb alloy NPs on MWCNTs is a tedious, time-consuming, and dangerous process [41], and commercial MWCNT supports are expensive.

In this work, we report a facile preparation method for the direct fabrication of Sb@PtSb₂ hybrids with a core-shell structure on N-doped carbon via simple pyrolysis of a mixture of glucose, melamine, and SbCl₃, according to the reaction scheme shown in Fig. 1. Furthermore, this Sb@PtSb₂/NC catalyst was characterized and tested for the selective oxidation of glycerol in base-free aqueous solutions.

Specific amounts of glucose, melamine, and SbCl₃ (with a mass ration of 6:6:1) were added and mixed manually in a ceramic mortar for 1 h, and then the mixture was dried in vacuum at 80 ℃ overnight. After that, the mixture was transferred to a quartz furnace and heated at 700 ℃ for 4 h (with a heating rate of 5 ℃/min) under a nitrogen flow of 20 mL/min. The obtained black powder was treated in 50 mL HCl (37 wt.%) under stirring for 6.0 h to remove the soluble species. The acid-treated solid sample was then filtered out, washed thoroughly with distilled water, and dried in vacuum at 60 ℃ overnight. The prepared sample is denoted as Sb/NC. Secondly, Sb/NC (1 g) was dispersed in a 7-mL aqueous solution of H₂PtCl₆ (0.01 g Pt/mL) under stirring, and further treated in an ultrasonic bath for 30 minutes. This suspension was then dried at 60 ℃, and the dried black powder was treated at 700 ℃ under hydrogen flow for 1 h. The final catalyst is denoted as Sb@PtSb₂/NC. The N-free Sb@Pt/C catalyst was prepared by following the same procedures as above without the addition of melamine in the feed.

Specific amounts of glucose and melamine (with a mass ratio of 1:1.5, in order to ensure the same N content in both Pt/NC and Sb@PtSb₂/NC) were added and mixed manually in a ceramic mortar for 1 h, and then the mixture was dried in vacuum at 80 ℃ overnight. After that, the mixture was transferred to a quartz furnace and heated at 700 ℃ for 4 h (with a heating rate of 5 ℃/min) under a nitrogen flow of 20 mL/min. The prepared sample is denoted as NC. Secondly, NC (1 g) was dispersed in a 7-mL aqueous solution of H₂PtCl₆ (0.01 g Pt/mL) under stirring. The suspension was then treated in an ultrasonic wave bath for 30 minutes. After that, the suspension was dried at 60 ℃, and the dried black powder was further treated at 700 ℃ under hydrogen flow for 1 h. The final catalyst is denoted as Pt/NC. N-free Pt/C catalyst was prepared by following the same procedures used to produce Pt/NC without the addition of melamine in the feed.

The real content of Sb and Pt in the prepared catalysts were measured using the following procedures. A sample (0.5 g) was treated in air from 25 to 900 ℃ (with a heating rate of 5 ℃/min), and then the residual solids were dissolved with 50 mL of aqua regia. Metal ions in the resulting solution were detected via inductively coupled plasma-atomic emission spectroscopy (ICP, Plasma-Spec-Ⅱ spectrometer), and these data are summarized in Table 1.

The texture and structure of all samples were characterized by N₂ adsorption measurements at –196 ℃ in a static volumetric apparatus (Micromeritics ASAP2020). Samples were first treated at 250 ℃ for 4 h under high vacuum. The specific surface area was calculated using the Brunner-Emmet-Teller (BET) method and pore size distributions were calculated using the Barrett-Joyner-Halenda (BJH) method. X-ray diffraction (XRD) was performed on RIGAKU D/MAX 2550/PC diffractometer using Cu K_α radiation at 40 kV and 100 mA. Diffraction data was recorded using continuous scanning at 0.02°/s and a step size of 0.02°. Scanning electron microscope (SEM) images were obtained with a Zeiss Sigma field-emission SEM (Model 8100). Transmission electron microscopy (TEM, JEOL-2010 F) was used to observe the morphologies and dimensions of catalysts at an accelerating voltage of 200 kV. X-ray photoelectron spectra (XPS) were recorded on a Perkin-Elmer PHI ESCA System, The X-ray source was a Mg standard anode 146 (1253.6 eV) operated at 12 kV and 300 W.

Oxygen adsorption and activation were measured over these catalysts via temperature programmed desorption of O₂ (O₂-TPD). The catalyst precursor was first pretreated at 700 ℃ in a purified H₂ flow of 30 mL/min for 1 h, and then the reactor was cooled to 50 ℃ in a H₂ atmosphere. Then, the synthesized catalyst was purged in a purified Ar flow of 30 mL/min at 50 ℃ for 1 h. After that, the catalyst was exposed to 10% O₂/Ar for 30 min, and then purged with Ar for 60 min at 50 ℃ in order to eliminate any physically adsorbed O₂. The desorption of adsorbed oxygen was conducted by ramping the catalyst temperature to 600 ℃ at a rate of 10 ℃/min, and O₂ (m/e = 32) was detected in the effluent and recorded as a function of temperature using a quadrupole mass spectrometer (Omni Star TM, GSD301, Switzerland).

Glycerol oxidation was carried out in a 25-mL custom-designed stainless-steel autoclave with a glass inner layer, which contained 5 mL of a 0.2 g/mL glycerol solution and 0.100 g of catalyst. The reactor was sealed, filled with 0.6 MPa of O₂, placed in water bath preheated to the required temperature, and maintained at that temperature for a given time under vigorous stirring with a magnetic stirrer. After the reaction, the catalyst was removed by centrifugation and the aqueous solution was analyzed using a high-performance liquid chromatograph (HPLC, Agilent 1100 series) equipped with a refractive index detector (RID) and a Zorbax SAX column (4.6 mm × 250 mm, Agilent). The analysis procedures are described in detail in our previous work [42]. The selectivity of each product was calculated as: (mmol of product in reaction mixture) × (the number of carbon atoms in product) / ((initial mmol of glycerol − mmol of glycerol left) × 3) × 100%.

Fig. 2 shows the N₂ adsorption isotherms and pore size distributions of Pt/C, Pt/NC, Sb@Pt/C, and Sb@PtSb₂/NC. All catalysts were found to exhibit type Ⅰ pattern isotherms, according to the International Union of Pure and Applied Chemistry (IUPAC) classification, which suggested that they were typical microporous materials. On the basis of adsorption data, the calculated BET surface areas of Pt/C, Pt/NC, Sb@Pt/C, and Sb@PtSb₂/NC were 115, 144, 264, and 275 m²/g, respectively. These results indicate that introducing N and/or Sb into the precursor (by adding melamine and/or SbCl₃) can increase the surface area and pore volume of the prepared catalyst. In addition, the effect of introducing Sb to the carbon support can be observed clearly in their typical SEM images. The outlines of Pt/C and Pt/NC look like closely packed solid laminates (see Fig. 3a–3d), while Sb@Pt/C and Sb@PtSb₂/NC exhibited porous structures (see Fig. 3e–3h).

Fig. 4 compares the XRD patterns of Pt/C, Pt/NC, Sb@Pt/C and Sb@PtSb₂/NC. It can be seen that the main diffraction peaks of Pt were detected in Pt/C and Pt/NC, while the calculated Pt crystallite sizes based on the full-width at half-maximum of the Pt(111) peak were 36.2 and 32.5 nm, respectively. These results indicate that introducing N into the catalyst can improve the dispersion of Pt, which might be attributed to the fact that N can increase the surface area of the carbon support and induce defects in the carbon support, strengthening the interactions between Pt and the carbon support [42]. At the same time, it is interesting to note that a PtSb₂ alloy (JCPDS 14-0141) was detected in the Sb@PtSb₂/NC catalyst along with metallic Sb (JCPDS 35-0732), which means that Pt can enter into the lattice of Sb. Peaks around 23.8° and 28.6° confirmed the existence of a separate Sb phase. On the other hand, separate Sb and Pt phases and trace amounts of Pt₅Sb alloy were detected in the XRD pattern of Sb@Pt/C.

Typical HRTEM images indicated that a layer of PtSb₂ alloy formed on the surfaces of well-crystallized Sb particles in the Sb@PtSb₂/NC catalyst (see Fig. 5), as the observed d-spacing values of 0.325 nm and 0.350 nm correspond to the (200) plane of PtSb₂ alloy and the (101) plane of Sb (see Fig. 5b), respectively. Line-scan EDS results further confirmed that Pt was mainly located in the shell of the Sb@PtSb₂ hybrid (see Fig. 5c–5f).

Raman analysis further confirmed that introducing SbCl₃ into the mixture of glucose and melamine can induce defects in the final catalyst (see Fig. 6), as the calculated intensity ratio of the D band to G band (I_D/I_G) increased from 0.863 (in Pt/C) to 0.915 (in Sb@Pt/C), 0.934 (in Pt/NC), and 0.953 (in Sb@PtSb₂/NC). In addition, the introduction of N can also increase the amount of carbon defects, as the ratio of the D band to the G band (I_D/I_G) of Pt/NC increased compared to that of Pt/C. And these defects can bring additional active sites, which encourages the oxidation reaction [29, 43].

Temperature-programmed desorption of O₂ (O₂-TPD, see Fig. 7) over Pt/NC and Sb@PtSb₂/NC catalysts indicated that mainly atomic oxygen was adsorbed on the surface of both catalysts [44, 45], but the amount of desorbed O₂ over Sb@PtSb₂/NC was much higher than that over Pt/NC. The above XRD analysis indicated that the crystalline structure of the PtSb₂ alloy consists of a Pa-3 unit cell structure with a dimension of 0.644 nm (see Fig. 4), while the crystalline structure of elemental Pt consists of an Fm-3m unit cell structure with a dimension of 0.392 nm. The enhanced activation of oxygen over the Sb@PtSb₂ catalyst might be attributed to the formation of a PtSb₂ alloy shell and the enlarged distance between Pt atoms, which are more favorable for the dissociative adsorption of oxygen (see Fig. 7). In other words, the formation of a PtSb₂ alloy shell in the Sb@PtSb₂ hybrid increased the space between metal atoms, allowing the adsorption and surface diffusion of more oxygen atoms in the catalyst.

XPS analysis provided evidence of electron transfer from N to Pt, as the detected binding energy of the Pt 4f peaks shifted from 71.4 and 74.7 eV (in Pt/C) to 71.0 and 74.3 eV (in Pt/NC), and from 71.3 and 74.6 eV (in Sb@Pt/C) to 70.8 and 74.1 eV (in Sb@PtSb₂/NC) (see Fig. 8A and 8B). At the same time, XPS also confirmed the formation of a PtSb₂ alloy in Sb@PtSb₂/NC, as the detected binding energy of Pt 4f in Sb@PtSb₂/NC falls well within the range of values reported for PtSb₂ in the literature [35]. Deconvolution of the N 1s spectra (see Fig. 8C) suggested the formation of pyridinic nitrogen (N1, 398.5 eV), pyrrolic nitrogen (N2, 400.1 eV), quaternary nitrogen (N3, 401.1 eV), and pyridine N-oxide nitrogen (N4, 403.3 eV) in Pt/NC and Sb@PtSb₂/NC [42]. It was reported that pyridinic N can reduce the energy barrier for the adsorption of reactants on adjacent carbon atoms and accelerate the rate-limiting first-electron transfer process, resulting in a significant enhancement of catalytic activity of the carbon surface [46, 47]. Moreover, quaternary N in a graphene structure can bring about a non-uniform electron distribution and a significant enhancement of the catalytic activity of the carbon surface [48]. Introducing Sb and N dopants into the catalyst can enhance the interactions between Pt NPs and the carbon support, leading to an easier transfer of electrons from N atoms to Pt NPs, and these electron-enriched Pt particles are more active in the oxidation of glycerol.

The above characterizations confirmed that N-doped carbon-supported Sb@PtSb₂ hybrids with a core-shell structure could be fabricated in a facile manner via the direct pyrolysis of a mixture of glucose, melamine, and SbCl₃, followed by impregnation with Pt.

Catalytic reaction results confirmed that this Sb@PtSb₂/NC catalyst was highly active for the selective oxidation of glycerol in base-free aqueous solutions. As is summarized in Table 2, Sb/NC was inactive for the oxidation of glycerol, which indicates that Pt and/or PtSb₂ provide the active sites for this reaction. The detected conversion of glycerol over the monometallic Pt/C catalyst was low (13.5%), and mainly GLYA (75.2%), DHA (12.8%), and HPYA (5.6%) were detected. The conversion of glycerol over Pt/NC improved significantly to 21.5% with 77.3% selectivity for GLYA. The higher activity of Pt/NC may be attributed to its higher surface area, increased amount of carbon defects, and higher dispersion of Pt, while electron donation from N to Pt can also accelerate the activation of molecular oxygen and improve the reaction rate. When Sb was added to the catalyst (Sb@Pt/C), the conversion of glycerol further increased to 26.0% with a 73.3% selectivity for GLYA. It is interesting to note that the detected conversion of glycerol jumped to 65.3% over Sb@PtSb₂/NC under the same reaction conditions, and the selectivity for DHA increased to 39.2%. The calculated TOF value on the basis of total Pt metal in Sb@PtSb₂/NC at a conversion of 10% reached 130.1 h^–1, which is higher than those of Pt/C (34.3 h^–1), Pt/NC (67.0 h^–1), and Sb@Pt/NC (76.3 h^–1) [49, 50]. The enhanced activity of Sb@PtSb₂/NC could be attributed to its higher surface area (see Fig. 2 and Table 1) and increased amount of surface defects, which promoted the adsorption and activation of glycerol and oxygen. At the same time, the formation of a PtSb₂ alloy in Sb@PtSb₂/NC was also favorable for the adsorption and transfer of oxygen (see the O₂-TPD in Fig. 7). It was also found that the Sb@PtSb₂ hybrid was more selective for the formation of DHA, which might be attributed to the fact that the PtSb₂ alloy can suppress the further oxidation of DHA (see Fig. 9) [40].

Fig. 9 shows the time course of glycerol oxidation over Sb@PtSb₂/NC at 60 ℃ in 0.2 g/mL glycerol aqueous solution. It can be seen that the conversion of glycerol increased quickly to 70.2% within the first four hours, and then increased steadily in the following time. The reaction rate slowed down with increasing conversion, which might be caused by the strong adsorption of the formed products. This phenomenon was discussed by Davis et al. in 2014 [23], and several works from Zhou et al. also confirmed that strongly adsorbed intermediates can decrease the reaction rate over Pt/AC(PO) catalysts [51, 52]. When these adsorbed intermediates were washed off, the Sb@PtSb₂/NC catalyst could be recovered. At the same time, it was found that the selectivity to DHA reached 56.9% at the beginning of the reaction, and it decreased continuously with time on stream. These results indicate that DHA is an important intermediate of the reaction. The selectivity to GLYA increased from 37.5% (at the beginning of the reaction) to 70.2% (at 4 h), and then changed only slightly. On the other hand, the selectivity of HPYA and products derived from C–C bond cleavage (such as GLYCA) increased continuously. In this case, we think that the primary product DHA and/or GLYA would further oxidize to HPYA and GLYCA during the reaction. The proposed reaction network is illustrated in Fig. 10.

That is, GLY was first oxidized to DHA or GLYHD, then GLYHD was further oxidized to GLYA, and then GLYA could be oxidized to GLYCA, OXA, or even to CO₂. At the same time, DHA was further oxidized to HPYA, and then HPYA could be oxidized to GOX, OXA, or even to CO₂ (see Fig. 10).

Fig. 11 shows the performance of the recycled Sb@PtSb₂/NC catalyst for the selective oxidation of glycerol under base-free conditions. The conversion of glycerol decreased from 65.3% to 43.1% after the catalyst was used five times, while the selectivity of GLYA and DHA remained almost the same. These results show that Sb@PtSb₂/NC still maintained high activity for the oxidation of glycerol after five cycles. The deactivation might be attributed to the mass loss of the catalyst during the reaction and recycling process because of the vigorous stirring involved.

Sb@PtSb₂ hybrids with a core-shell structure could be fabricated on the surface of N-doped carbon (NC) via a facile synthesis route. It was confirmed that SbCl₃ can increase the surface area of a binary glucose + melamine pyrolyzed support, and a PtSb₂ alloy shell could be formed on the surface of Sb particles during the hydrogen reduction. The synthesized Sb@PtSb₂ hybrid was more active for the absorption and activation of O₂, and also exhibited prominent activity for the selective oxidation of glycerol with molecular oxygen. The best specific activity based on the total amount of Pt reached 130.1 h^–1.