Facile conversion of glycerol to 1,3-dihydroxyacetone by using mesoporous CuO–SnO2 composite oxide supported Au catalysts

Selective catalytic oxidation of polyols, e.g., the selective catalytic oxidation of the secondary –OH bond in glycerol, remains a considerable challenge. In this study, a series of mesoporous CuO–SnO2 composite oxides were prepared by a hard-template method and used to support Au catalysts for the selective oxidation of glycerol to 1,3-dihydroxyacetone (DHA) under base-free conditions. Catalysts with different Cu:Sn molar ratios gave different catalytic performances. A high conversion of glycerol (100%) and selectivity for DHA (94.7%) were obtained in 2 h at 80 °C and PO2 = 1 MPa over the Au/CuO–SnO2-3:1 catalyst. Further investigation indicated that the high catalytic activity of Au/CuO–SnO2-3:1 is related to the small size and high dispersion of Au nanoparticles (NPs), the interactions between the Au NPs and the support, the synergistic effect between CuO and SnO2, and the amount of surface lattice oxygen species. Various reaction parameters, namely the glycerol:Au molar ratio, the reaction temperature, the initial O2 pressure, the reaction time, and the support calcination temperature were studied. Although the conversion rate by the catalyst decreased after four cycles, the selectivity remained above 86%. Density functional theory calculations showed that the synergy between CuO and SnO2 improves the catalytic activity in glycerol oxidation to DHA. The results show that mesoporous composite oxide supports have a wide range of potential applications in the selective oxidation of glycerol to other high-value-added products.


Introduction
Environmental concerns are attracting much attention because of the issues arising from the rapid development of society and the global economy, and increasing energy consumption. The use of renewable biomass to replace fossil feedstocks for producing fuels, commodity chemicals, and polymeric materials has attracted much interest. Renewable fuels are considered to be ideal energy sources. They can not only meet energy needs, but also reduce CO 2 emissions.
Biodiesel is considered to be a promising environmentally friendly renewable energy source because of its many advantages, e.g., it is readily available, clean, and degradable [1,2]. However, the boom in the biodiesel industry has led to a sharp increase in production of a major byproduct, i.e. glycerol. There is therefore a surplus of glycerol, and its production price has decreased.
DHA is a primary product of selective oxidation of the secondary O-H bond of glycerol. It is an important intermediate in the production of cosmetics and medicines, and in organic synthesis, and is expected to have further industrial applications. Controlling the selectivity for a target product 1 3 in the catalytic oxidation of glycerol is particularly difficult. Research on the catalytic oxidation of glycerol to DHA has mainly focused on supported noble-metal-based catalysts [9][10][11][12][13][14]. The activities of supported Au catalysts in the oxidation of diols are higher than those of Pt and Pd catalysts, and they have excellent antioxidant capabilities [15,16]. However, supported Au catalysts are only active in glycerol conversion under basic conditions [17][18][19][20]. Under such conditions, selective oxidation of the secondary O-H bond of glycerol is poor, and the selectivity for DHA is unsatisfactory. Demirel et al. reported that Au/activated carbon catalyzed the oxidation of glycerol to glyceric acid under basic conditions, with simultaneous production of DHA, but the selectivity for DHA was only 20% [21]. In 2007, they reported that a Au/C catalyst gave a glycerol conversion rate of 50% under basic conditions, and the DHA selectivity at this conversion level was 26% [22]. When they used Au/ CeO 2 to catalyze the oxidation of glycerol at pH 12, the glycerol conversion rate reached 30% and the DHA selectivity was 53% [23]. In 2014, Liu reported that the nature of the support significantly affects the Au active component [24]. Catalysts consisting of Au supported on different metal oxides, namely Al 2 O 3 , TiO 2 , NiO, ZrO 2 , and CuO, all showed high DHA selectivity in the catalytic oxidation of glycerol, but the glycerol conversion rates differed significantly. The glycerol conversion rate with the Au/CuO catalyst was significantly higher than those obtained with the other catalysts, and the selectivity for DHA was greater than 80%. These results show that metal-oxide-supported Au catalysts can give high selectivity for DHA in glycerol oxidation under base-free conditions. Use of CuO as a reductive oxide support can improve the dispersion of Au on the support and enhance the interactions between the active component and the metal support; this improves the catalytic activity. In recent years, the synthesis of DHA by selective oxidation of glycerol over Au catalysts supported on metal oxides such as ZnO [9,25,26], MgO-Al 2 O 3 , [27] Cu x O [10], Cu x Zr 1−x O y , [11,28] CuAlO [29], and ZnAlO [30] has attracted much attention.
The use of ordered mesoporous materials as catalysts has been widely studied because of their highly ordered porous structures, adjustable pore sizes, and high surface areas and pore volumes. The high surface areas and porosities of mesoporous materials enable active components to be highly dispersed on the support material, and this enables effective diffusion of reactants to the active centers [31]. Liu et al. used Co-and Ru-modified Pt/MCM-41 bimetallic catalysts, and achieved glyceric acid selectivities of 82.5% and 80.1%, respectively [32,33]. In addition to the interactions between the support and the active components, the mesoporous structure of the MCM-41 support contributes to the high catalytic activity. Peng et al. prepared mesoporous Cu-Sn mixed-oxide nanorods [34]. A 1% Pd/SnO 2 catalyst showed excellent activity and stability in CO oxidation at low temperatures because of its high surface area and the presence of a large number of mobile oxygen species.
These findings indicate that Cu species are a key factor in the selective catalytic oxidation of glycerol to DHA over supported Au catalysts. In this work, we immobilized CuO on SnO 2 supports via a hard-template technique to produce ordered mesoporous composite oxides with a range of CuO contents. By changing the support preparation parameters, Cu-Sn mesoporous composite oxides with various compositions and characteristics were obtained. The results will enable the design of supports for catalyst for the selective oxidation of glycerol to high-value-added products.

Synthesis of SBA-15 hard template
Mesoporous silica SBA-15 with a bicontinuous porous structure was used as a hard template. SBA-15 was synthesized as follows. Pluronic®P-123 (400 g) was dissolved in deionized water (30.0 mL), and then HCl aqueous solution (120 mL, 2 mol L −1 ) was added. After stirring for 2 h, the solution was heated to 40 °C and TEOS (9.00 g) was added to the solution. The mixture was aged for 24 h at 40 °C. The resulting gel was transferred to a Teflon-lined autoclave and kept at 100 °C in an oven for 48 h. The solid product was separated by filtration, washed with deionized water, dried overnight at 65 °C, and calcined at 550 °C for 6 h (heating rate 1 °C min −1 ).

Synthesis of Cu-Sn mesoporous composite oxides
A hard-template method was used for the preparation of Cu-Sn mesoporous composite oxides. The synthesis of a sample with a Cu:Sn molar ratio of 1:1 is used as an example. Cu(NO 3 ) 2 ·3H 2 O (1.5 g) and SnC 2 O 4 (1.3 g) were premixed with SBA-15 silica (1.5 g) in n-hexane (15 mL). The resulting viscous mixture was ground with an agate pestle for approximately 30 min, and then dispersed in n-hexane (50 mL). The suspension was stirred for 12 h under reflux at 70 °C. The Cu-Sn precursor was separated by centrifugation, washed three times with n-hexane, dried at 70 °C for 12 h, and then calcined at 500 °C for 5 h. The Cu-Sn mesoporous composite oxides were obtained by treating the calcined precursor twice with NaOH aqueous solution (2 mol L −1 ) at 25 °C, for 2 h each time, and drying under vacuum at 60 °C for 12 h. Cu-Sn mesoporous composite oxides with other Cu:Sn molar ratios (3:1 and 1:5) were also prepared by changing the molar ratio of the metal salts. CuO and SnO 2 supports were synthesized by the same method. The Cu-Sn mesoporous composite oxides are denoted by CuO-SnO 2 -3:1, CuO-SnO 2 -1:1, and CuO-SnO 2 -1:5, respectively.

Catalyst synthesis
Au nanoparticles (NPs) were deposited on the support surface by deposition-precipitation with urea. In a typical procedure, the Cu-Sn mesoporous composite oxide (1.00 g), HAuCl 4 aqueous solution (6.268 mL, 0.0243 mol L −1 ), and urea (3.66 g; urea/Au = 400:1 mol/mol) were dispersed in distilled water (50 mL). The suspension was aged for 16 h at room temperature after continuous stirring for 6 h at 80 °C. The Au/CuO-SnO 2 catalyst was obtained by separating the solid by filtration and washing with a large amount of distilled water to remove Cl − ions. The resultant solid was dried at 110 °C for 4 h and calcined at 200 °C in an air flow for 5 h. The same procedure was used to prepare the other catalysts. The catalysts are denoted by Au/CuO-SnO 2 -3:1, Au/CuO-SnO 2 -1:1, and Au/CuO-SnO 2 -1:5, respectively.

Catalyst characterization
The crystal phases of the catalysts and support were identified by X-ray crystallography (XRD; Rigaku Smart Lab X-ray diffractometer, CuKβ radiation source, 40 kV and 40 mA). High-angle diffraction patterns of the samples were recorded in the 2θ range 10°-80° at a scanning rate of 8° min −1 . Low-angle diffraction patterns of the samples were recorded in the 2θ range 0.6°-6° at a scanning rate of 0.3° min −1 .
The textural properties of the catalysts were investigated by recording N 2 adsorption-desorption isotherms at liquid N 2 temperature (77 K, Micromeritics ASAP 2020 HD88 instrument). Before the measurements, the samples were pretreated at 150 °C for 6 h under vacuum. The surface area (S BET ) was calculated by using the Brunauer-Emmett-Teller (BET) equation. The total pore volume (V p ) was estimated by determining the N 2 uptake at a relative pressure (p/p 0 ) of ca. 0.99. The average pore diameter (D p ) was estimated from the surface area and the total pore volume (D p = 4V p /S BET ).
The microstructures and crystalline sizes were determined, and energy-dispersive spectroscopy (EDS) element mapping images were obtained, by using an FEI Tecnai G2 F20 electron microscope at an accelerating voltage of 200 kV. The sizes and distributions of the Au NPs were obtained from transmission electron microscopy (TEM) images by averaging the values for a minimum of 200 particles.
X-ray photoelectron spectroscopy (XPS) was performed with a Thermo Scientific ESCALAB Xi + electron spectrometer with an Al Kα X-ray source. Binding energies were referenced to the C 1 s peak (set at 284.8 eV) of the sp 2 hybridized (C = C) carbon in the sample.
The elemental contents of the catalysts were determined by inductively coupled plasma optical emission spectroscopy (ICP-OES; Varian 710-ES instrument).
Hydrogen temperature-programmed reduction (H 2 -TPR) and CO 2 -temperature-programmed desorption (CO 2 -TPD) experiments were performed on the catalysts by using an Auto Chem II 2920 instrument with a thermal conductivity detector. Prior to the H 2 -TPR experiments, the sample (50 mg) was placed in a quartz reactor. the experiments were performed with 5% H 2 in Ar (30 mL min −1 ) from room temperature to 500 °C at a heating rate of 10 °C min −1 . For the CO 2 -TPD experiments, the sample (0.1 g) was pre-processed at 200 °C in He (50 mL min −1 ) for 2 h. After the temperature had dropped to room temperature, the quartz tube was purged with CO 2 until adsorption equilibrium was reached. The tube was then purged with He (50 mL min −1 ) to remove gas-phase CO 2 . The temperature of the reaction tube was increased to 600 °C at a heating rate of 10 °C min −1 .

Reactions and analysis
The selective catalytic oxidation of glycerol was performed in a stainless-steel autoclave reactor equipped with a mechanical stirrer and a temperature-control system. In a typical experiment, glycerol aqueous solution (24 mL, 0.1 mol L −1 ) and the catalyst (0.16 g, glycerol/Au = 100 mol/mol) were placed in the autoclave reactor. The autoclave was purged three times with O 2 at room temperature, and then filled with O 2 to a pressure of 1 MPa. The autoclave was heated and the reaction time began as soon as the desired temperature (80 °C) was reached; the reaction was performed under stirring at 500 rpm for 2 h. After each reaction, the autoclave was cooled with an ice-water bath, and the mixture was separated by filtration with an organic microfilter (0.22 μm).
The liquid-phase composition was determined with a Sykam S-501 high-performance liquid chromatography system with refractive index (S3590, held at 40 °C) and ultraviolet-visible (S3250, wavelength 210 nm) detectors. H 2 SO 4 aqueous solution (0.005 mol L −1 , flow rate 0.5 mL min −1 ) was used as the mobile phase in conjunction with a Bio-Rad Aminex HPX-87H ion-exclusion column held at 45 °C. The selectivities for various products and the glycerol conversion were calculated by using the following equations:

Density functional theory (DFT) calculations
DFT calculations were performed by using CASTEP software (Materials Studio) [35]. The CuO (111) and SnO 2 (110) surfaces were cut from the optimized CuO and SnO 2 bulk crystals, respectively [36,37]. A periodic slab model was used to describe the surface. The CuO (111) and SnO 2 (110) surfaces were modeled as three-layered slabs, in each of which the bottom layer was fixed and the upper two layers and adsorbates were allowed to relax. We used (6 × 3 × 1) unit cells for CuO (111) with the adjacent slabs separated by a 10 Å vacuum, and (2 × 2 × 1) unit cells for SnO 2 (110) with the adjacent slabs separated by a 10 Å vacuum. We selected an energy cutoff of 381.0 eV for a plane-wave basis set and the 1 × 1 × 1 Monkhorst-Pack k-point for the first Brillouin zone. The Perdew-Burke-Ernzerhof (PBE) approximation, which is based on the generalized gradient approximation (GGA), was used in the calculations. We used the following equation to calculate the adsorption energy (E ads ): and E M represent the calculated energies of the adsorption system, adsorbate, and substrate, respectively.

Catalyst characterization
The mesoporous structures of the Cu-Sn composite oxides were investigated by using high-resolution TEM (HRTEM), XRD, and N 2 physisorption. The TEM image of the CuO-SnO 2 -3:1 sample in Fig. 1(a) indicates the development of mesoporosity during the thermal treatment. Figures 1(b) and S1 show low-angle XRD patterns of the oxide supports and supported Au catalysts, respectively. The patterns of all the samples, except that of SnO 2 , show one Glycerol conversion(% ) = Moles of converted Glycerol Moles of initial Glycerol × 100% Selectivity(% ) = Moles of product formed Total moles of C 1 + C 2 + C 3 products formed × 100% diffraction peak at 1.5°-2.1°. This indicates the presence of mesoporous CuO and a Cu-Sn solid solution. The Au loading did not affect the crystal morphology of the support. Figure 1(c) and (d) show the N 2 adsorption-desorption isotherms and corresponding pore size distribution plots for the supported Au catalysts. All the isotherms are typical type-IV curves with an H3-type hysteresis loop, which indicates a mesoporous structure with slit-like pores [38]. The BET surface areas (S BET ), pore volumes (V p ), and pore diameters (D p ) of the supported Au catalysts were calculated; the results are listed in Table 1 Figure 2 shows wide-angle XRD patterns of the supports and catalysts. The patterns show the presence of both the tetragonal rutile SnO 2 [39] phase (JCPDS no. 72-1147) and the monoclinic CuO [40] phase (JCPDS no. 80-1268) in the Cu-Sn mesoporous composite oxides and corresponding supported Au catalysts. The intensity of the CuO diffraction peak distinctly decreased with decreasing Cu content, whereas that of the SnO 2 peak increased. The CuO and SnO 2 diffraction peaks for the Cu-Sn mesoporous composite oxides shifted to higher 2θ angles (Fig. 2a); this is attributed to interactions between Cu and Sn. The results show that Cu-Sn mesoporous composite oxides with high degrees of crystallinity were successfully prepared. Figure 2(b) shows that the Au loading had no effect on the support crystal morphology and diffraction peak intensities. For the Au/SnO 2 catalyst, the characteristic Au diffraction peak at approximately 38.2° was not observed. This is attributed to the low loading or fine particles (less than 3 nm, as shown in the TEM image) of Au, and complete overlapping of the Au peak with the CuO diffraction peak at 38.2°.
The Au loading, and Cu and Sn contents of all the catalysts are listed in Table 1. The data show that the actual Au loading was between 2.59 wt% and 2.95 wt%, which is lower than the theoretical loading of 3 wt%. The actual Au loading increased with increasing Sn content. This may be related to the specific properties of the support, because Au can affect the surface chemical states of SnO 2 . [41] The actual Cu:Sn molar ratios are similar to those expected. Figure 3 shows TEM images of the Cu-Sn mesoporous composite oxide Au catalysts and the corresponding Au particle size distributions. The figure shows that the supported Au catalysts consist of approximately spherical Au particles and the average Au particle diameters are similar. The mean Au particle size is approximately 1.76-2.12 nm, which is lower than those for the Au/CuO (3.03 ± 0.52 nm) and Au/ SnO 2 (4.48 ± 0.95 nm) catalysts (Fig. S2). This indicates that interactions between the Au particles and Cu-Sn mixed oxide support increased, which increased the mean size and dispersion of the Au NPs. The mean Au particle size is lowest for the Au/CuO-SnO 2 -3:1 catalyst (1.76 ± 0.30 nm). Generally, the catalytic activity of a supported Au catalyst is correlated with the mean particle size. Small Au particles give a larger exposed metal surface area and therefore a higher activity [42].   The distribution of Au on the catalyst surface was investigated by EDS elemental mapping of the Au/CuO-SnO 2 -3:1 catalyst; the results are shown in Fig. 3(d) and Fig. S3. The Au, Cu, Sn, and O elements have good chemical homogeneity in this catalyst. This confirms that Au species are uniformly dispersed on the mesoporous framework. Fig. 4(a) shows H 2 -TPR profiles for the Cu-Sn mesoporous composite oxides, and the CuO and SnO 2 supports. The CuO profile shows a reduction peak at approximately 175-330 °C, which is ascribed to overlapping of the wide bands for reduction of Cu 2+ to Cu + and Cu + to Cu 0 . [43] In the H 2 -TPR profile of SnO 2 , a wide reduction peak is observed at 600-900 °C. This peak is attributed to the reduction of both surface Sn 2+ and bulk Sn 4+ to Sn 0 . [44] The profiles for the Cu-Sn mesoporous composite oxides show two reduction peaks within the investigated temperature range (100-900 °C). The peaks at 200-240 °C are ascribed to reduction of Cu + to Cu 0 and those at 520-560 °C are ascribed to reduction of Sn 4+ to Sn 0 . The reduction temperatures of the Cu-Sn mesoporous composite oxides are clearly lower than those of CuO and SnO 2 . The CuO reduction peak shifts to lower temperature, and the SnO 2 reduction peak shifts to higher temperature, with decreasing Cu content. This provides evidence of strong synergistic effects between CuO and SnO 2 , [45] and indicates that the Cu-Sn mesoporous composite oxides have good oxygen delivery capacities [46,47].
The reduction behaviors of the supported Au catalysts were similar to those of the corresponding supports, but the reduction peaks shifted to lower temperatures, as shown in Fig. 4(b). In particular, the Au/CuO-SnO 2 -3:1 catalyst showed two reduction peaks at about 210 and 279 °C. The low temperature reduction peak is associated to bulk of copper species with weak interaction with Au, and the high temperature reduction peak is associated to bulk of copper species [48]. These results suggest that the reducibility is enhanced, which is beneficial to the catalytic oxidation of glycerol. Figure 5 shows the Au 4f, Cu 2p, Sn 3d, and O 1 s XP spectra of the supported Au catalysts. Figure 5(a) shows that the Au 4f 7/2 peaks of all the supported Au catalysts were located at binding energies between 83.5 to 84.3 eV. This suggests the presence of metallic Au [49,50]. The Au 4f peak intensity decreased with decreasing Cu content, which indicates electron transfer from Cu and/or Sn to Au. For the Au/CuO catalyst, in addition to surface Au 0 species (84.3 eV), there was a small quantity of Au δ+ species (85.5 eV). These results suggest that the electronic interactions between Au and Sn favor the formation of Au 0 .
The types of oxygen species on the catalyst surface were identified by O 1 s XPS; the results are shown in Fig. 5(b). The peak for the Au/CuO catalyst was decomposed into two peaks with binding energies of approximately 529.7 and 531.6 eV, which correspond to surface lattice oxygen species (O latt ) and surface chemisorbed oxygen species (O ads ), [51,52]    495.0 eV, which correspond to Sn 3d 5/2 and Sn 3d 3/2 , respectively, and can be ascribed to the Sn-O [53]. The binding energy tended to shift to higher values with increasing Cu content. This indicates that the chemical environment of Sn had changed, possibly as a result of electron transfer from Sn to Cu. The Cu 2p XP spectra (Fig. 5d) were used to investigate the Cu species on the catalyst surfaces. Two main regions, namely 950-957 eV (Cu 2p 1/2 ) and 930-938 eV (Cu 2p 3/2 ), were fitted for all the Cu-bearing catalysts; two satellite peaks were present at 939-946 and 960-964 eV, respectively. For the Au/CuO-SnO 2 -1:1 and Au/CuO-SnO 2 -1:5 catalysts, Cu 2p 3/2 can be fitted to Cu + 2p 3/2 (932.9 eV) and Cu 2+ 2p 3/2 (934.5 eV) [54,55]. However, for the Au/CuO and Au/CuO-SnO 2 -3:1 catalysts, there were three peaks in the range 930-940 eV. The binding energies at 933.9 to 934.5 eV can be assigned to Cu 2+ species [56][57][58]. Various studies have shown that the positions of these peaks depend on the chemical composition of the sample, particularly on the environment near the Cu 2+ cation [59,60]. The above results prove that Cu-Sn composite oxides were successfully prepared. In addition, the changes in the Cu and Sn binding energies in the Cu-Sn composite oxides relative to those in CuO and SnO 2 indicate that interactions occur between CuO

Reactions and analysis
The catalytic performances of the supported Au catalysts were evaluated by performing reactions in a batch reactor; the results are listed in Table 2. In addition to DHA, byproducts such as glyceric acid, lactic acid, and oxalic acid were detected in small amounts. The glycerol conversions over the Au/SnO 2 and Au/CuO catalysts were 30.9% and 72.2%, respectively, with corresponding DHA selectivities of 86.3% and 96.4%. The catalytic activities of the Au catalyst supported on Cu-Sn mesoporous composite oxides were considerably affected by the Cu:Sn molar ratio. The glycerol conversion increased with increasing CuO content from 58.1% for the Au/CuO-SnO 2 -1:5 catalyst to 100% for the Au/CuO-SnO 2 -3:1 catalyst. This indicates that incorporation of CuO into the Cu-Sn mesoporous composite oxides significantly increases the catalytic activity. The Au/CuO-SnO 2 -3:1 catalyst gave high selectivity for DHA, namely 94.7%, which is higher than those achieved with the other Au catalysts supported on Cu-Sn mesoporous composite oxides and the Au/SnO 2 catalyst, and nearly the same as that obtained with the Au/CuO catalyst. These results suggest that construction of a composite oxide catalyst can greatly improve the catalytic activity under base-free conditions, and the interactions between Cu and Sn promote the oxidation of glycerol and selectivity for DHA. For the Au/ SnO 2 and Au/CuO catalysts, the larger Au particle sizes may account for the poorer catalytic performances. A combination of the XPS, BET, and H 2 -TPR results shows that the Au/CuO-SnO 2 -3:1 catalyst has the highest catalytic activity because it had a higher Au 0 content, stronger low-temperature reduction ability, and smallest Au particle size.
The Au/CuO-SnO 2 -3:1 catalyst was used as the primary catalyst for investigation of the effects of the reaction conditions on the catalytic oxidation of glycerol and the selectivity for DHA. First, the effect of the glycerol/Au molar ratio on the conversion of glycerol to DHA was studied; the results are shown in Fig. 6. The results show that different glycerol/ Au molar ratios led to different catalytic results. The glycerol conversion decreased with increasing glycerol/Au molar ratio from 100 to 1000. The catalytic activity decreases because a lower Au content leads to a decreased number of active sites on the supported catalyst. However, only slight changes in the DHA selectivity were observed for different glycerol/Au molar ratios.
Next, the reaction temperature, initial pressure of O 2 , reaction time, and support calcination temperature were optimized. The most active catalyst, i.e., Au/CuO-SnO 2 -3:1, was selected as the catalyst and the glycerol/Au ratio was 500 mol/mol. The results are listed in Table 3. The glycerol conversion increased with increasing reaction temperature (Table 3, entries 1-4). The selectivity for DHA did not change significantly with increasing reaction temperature from 40 to 80 °C, but it decreased rapidly to 71.5% at 100 °C. This result is attributed to oxidation of DHA to other byproducts such as glyceric acid and lactic acid at high temperatures. The initial pressure of O 2 slightly affected the   (Table 3, entries 3 and 5-7). When the initial O 2 pressure was increased from 5 to 10 bar, the glycerol conversion increased from 41.9% to 50.7%, and then gradually decreased with further increases in the initial pressure. This is because the number of active sites on the catalyst surface occupied by O 2 increases with increasing P O2 . The highest glycerol conversion was obtained when P O2 was 10 bar. The selectivity for DHA decreased slightly with increasing P O2 . In terms of glycerol conversion and DHA selectivity, the optimal reaction pressure is 10 bar. As expected, the glycerol conversion increased gradually with reaction time, and the selectivity for DHA decreased gradually (Table 3, entries 3 and 8-11). When the reaction time was extended to 3 h, the selectivity for DHA rapidly decreased to 80.9%. The results show that the optimal reaction time for the conversion of glycerol to DHA is 2 h.

Catalyst stability
The stabilities of the Au catalysts supported on Cu-Sn mesoporous composite oxides were evaluated by investigating recycling of the Au/CuO-SnO 2 -3:1 catalyst. After each cycle, the catalyst was separated by filtration, washed with deionized water, and dried overnight at 80 °C. The recycled catalyst was then used for the next cycle under the same reaction conditions. The results of a four-cycle recycling test are shown in Fig. 7. The fresh catalyst gave 50.7% glycerol conversion and 92.3% DHA selectivity. In the second cycle, the glycerol conversion decreased to 28.2%, and in the fourth cycle, the conversion declined considerably to 9%. The selectivity for DHA changed only slightly.
To explore the reason for the catalyst deactivation, the used Au/CuO-SnO 2 -3:1 catalyst was characterized by XPS. The results of the XPS analysis confirmed that the binding energy of Au 0 4f 7/2 decreased after four runs (from 84.02 eV to 83.68 eV), which indicates that the gold-support interaction changed after the reaction. After four runs, the binding energy at 529.7 and 530.7 eV were disappeared and a new peak at 530.20 was formed, which indicates that the Cu-O species and Sn-O species changed after the reaction. The binding energy of the Sn 3d 5/2 slightly decreased after four runs (from 486.64 eV to 486.48 eV). The binding energy of the Cu 2p 3/2 peaks have remarkable changes after four runs. These results indicated that the chemical state of Cu changed after recycling. In addition, the content of Au, Sn and Cu on the surface clearly decreased, which may be caused by dissolving of elements during the reaction process. The abovementioned changes were responsible for the decrease of the catalytic activity (Fig. 8).

Discussion
The results described in Sect. 3 indicate that the Au/ CuO-SnO 2 -3:1 catalyst gave an excellent catalytic performance, with 100% glycerol conversion and 94.7% selectivity for DHA. The high activity of the Au/CuO-SnO 2 -3:1 catalyst can be explained by the catalyst properties. The intensities of the SnO 2 and CuO peaks in the XRD patterns of the Cu-Sn mesoporous composite oxides were higher than those in the patterns of other composite oxides. The positions of the CuO and SnO 2 diffraction peaks for the Cu-Sn mesoporous composite oxides shifted to higher 2θ angles, which confirms that Cu-Sn interactions occurred. TEM images showed that the Au particles in the Au/CuO-SnO 2 -3:1 catalyst had a smaller mean size and higher dispersion. The excellent catalytic performance is therefore mainly ascribed to higher exposure of metal on the surface because of the smaller Au particles.
For the Au/CuO-SnO 2 -3:1 catalyst, the CuO reduction peak shifted to a higher temperature and the SnO 2 reduction peak shifted to a lower temperature than those for the other Cu-Sn mesoporous composite oxide supported Au catalysts. This provides evidence of strong synergistic effects between the CuO and SnO 2 components and indicates that the Cu-Sn mesoporous composite oxides have good oxygen delivery capacities. This leads to enhanced reducibility, which is beneficial to the catalytic oxidation of glycerol.
The XPS results indicate that the interactions between Au particles and the support, and between CuO and SnO 2 , and the amount of surface lattice oxygen species affect the catalytic activity. These effects are favorable for the selective catalytic oxidation of the secondary O-H bond of glycerol.
The BET analysis results (Table 1) show that the S BET increased with decreasing Cu content in the Cu-Sn mesoporous composite oxide supported Au catalysts, and the values were higher than those for the Au/SnO 2 and Au/CuO catalysts. The average pore diameter was much lower than those for the Au/CuO and Au/SnO 2 catalysts. This shows that the differences among the catalytic performances are independent of the surface area and pore diameter. The characterization results show that the synergistic effect of CuO and SnO 2 improves the catalytic activity of the mesoporous CuO-SnO 2 composite oxide supported Au catalysts. The experimental results were clarified by performing DFT calculations for the Au/CuO-SnO 2 -3:1 catalyst. First, in accordance with the XRD results, a CuO (111) surface was cut on the basis of an optimized CuO cell; this has been proved to be the most stable surface by many researchers [36,[61][62][63]. Figure 9 and S4 show the structural properties of the stable configurations. The three O-Cu bond lengths were 1.887, 1.906, and 1.933 Å, respectively, and the O atom was exposed on the CuO (111) surface, which is beneficial for Au atom adsorption. When a Au atom was adsorbed on the CuO-SnO 2 composite oxide, a Au-O bond of length 2.074 Å was formed simultaneously. The CuO was changed by supporting Au, and the lengths of the three new O-Cu bonds increased to 1.932, 1.983, and 2.028 Å, respectively. The Au adsorption energy on the CuO (111) surface was − 4.55 eV, which indicates that Au atoms can be easily adsorbed on the CuO (111) surface. Subsequently, a SnO 2 (110) surface was cut on the basis of the optimized SnO 2 cell. This surface has been proved to be the most stable surface and the most popular one [64,65]. As shown in Fig. 9(c) and (d), the Sn atom was exposed on the SnO 2 (110) surface, and the O atom occupied a vacancy site, which can be beneficial for adsorbing the -OH of glycerol on the top Sn atom. The optimized Sn-O bond lengths were 2.149 and 2.148 Å before the adsorption calculations. When a glycerol molecule was adsorbed on the SnO 2 (110) surface, two new Sn-O bonds were formed of length 2.879 Å (secondary hydroxyl) and 3.274 Å (terminal hydroxyl). The lengths of the Sn-O bonds that were present before adsorption also changed, from 2.194 to 2.095 Å, and from 2.106 to 2.208 Å. The glycerol adsorption energy on the SnO 2 (110) surface was − 2.06 eV, which indicates that glycerol can be easily adsorbed on the SnO 2 (110) surface. These results indicate that synergy between CuO and SnO 2 can improve the catalytic activity in glycerol oxidation to DHA.

Conclusions
In this work, a series of mesoporous CuO-SnO 2 composite oxide supported Au catalysts were prepared by a hardtemplate method and evaluated in the selective oxidation of glycerol to DHA under base-free conditions. The effects of the Cu:Sn molar ratio and reaction conditions on the catalytic performance were investigated.
The results show that Au/CuO-SnO 2 -3:1 had the best catalytic activity; the glycerol conversion rate reached 100% and the selectivity for DHA reached 94.7% under the optimal conditions. The characterization results show that the small size of the Au NPs, the properties of the support, the synergistic effect between Au NPs and the support, and the basicity of the catalyst all significantly affect the catalytic performance.