The challenges of sustainable development, such as increasing energy consumption, expanding commodity demands, and environmental concerns, are calling for the efficient utilization of renewable resources. As an easily available byproduct of biodiesel production, glycerol has drawn increased attention because of its abundant functional groups [1-3]. Various value-added derivatives from glycerol have been obtained through different reactions, including selective oxidation [4, 5], hydrogenolysis [6, 7], and dehydration [8, 9], increasing the feasibility of using glycerol as a bio-building block. Therefore, to explore new pathways of glycerol conversions and enhance the competitiveness of the biodiesel industry is attractive.

Pyruvic acid (PA) has been widely used as food additives, precursors for drugs and agrochemicals, as well as weight-loss supplements [10, 11]. Because of the many promising applications of PA, its commercial market is considerable. PA is produced by fermentation of glucose or dehydrative decarboxylation of tartaric acid [12]. However, the problems of these methods remain challenging because the product separation efficiencies in the biological methods are low, whereas the pyrolysis of tartaric acid consumes a large amount of KHSO₄ and energy [13]. Thus, developing renewable alternative pathways to produce PA under mild conditions is highly appealing.

The one-step transformation of glycerol to PA is a potential option with high atom efficiency, which could be an addition to the reported PA production methods, including oxidations of propylene glycol or lactic acid (LA) [14-16]. This one-step pathway is viable in a fermentation method [17], but there are limitations in scaling-up. As for chemical methods, recent progresses on PA production from LA (or lactate) [13, 18, 19] and LA production from glycerol [20-23] have shed light on this one-step strategy. In liquid-phase LA oxidation to PA, heterogeneous bimetallic catalysts have been frequently adopted [15, 18, 24]. These catalysts commonly consist of noble metals such as Pt or Pd as active sites, heavy metals such as Pb, Te, and Bi as additives, and carbon as a support [15, 18, 24]. For glycerol oxidation to LA, we previously reported that a 1.0 wt% Pt/AC catalyst performed well under mild conditions, with almost no PA detected [25]. Hayashi et al. reported that Pb was a key additive for Pd/C catalyst in PA production from LA [15]. Therefore, the introduction of Pb promoter to a Pt/AC catalyst is expected to enhance the product distribution and PA yield during glycerol oxidation.

We report the one-pot production of PA from selective oxidation of glycerol under mild conditions. The Pb loading is optimized for PA selectivity. The Pb-Pt/AC catalysts prepared via different methods are investigated, including impregnation, deposition precipitation, and further calcination under an Ar atmosphere. N₂ physical adsorption, X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), and transmission electron microscopy (TEM) analyses are adopted to uncover the structure-activity relationships of catalysts.

A series of Pb-Pt/AC catalysts were prepared by the following methods. The Pt loading was fixed at 5.0 wt% in all Pb-Pt/AC catalysts.

An amount of Pb (NO₃)₂ was dissolved in 4 mL of deionized water and impregnated onto 2.0 g of an activated carbon (AC) support. The sample was dried at 120 ℃ for 12 h and reduced with H₂ at 400 ℃ for 4 h. Sequentially, Pt was deposited on the as-prepared Pb/AC sample by a deposition-precipitation method [26]. By changing the loading of lead, a series of x Pb-5Pt/AC-Im-DP catalysts were obtained (x = 1, 3, 5, and 7 wt%, representing the Pb loading). If the 5Pb-5Pt/AC-Im-DP catalyst was further treated with Ar at 500 ℃ for 4 h, the obtained catalyst was denoted as 5Pb-5Pt/AC-Im-DP-500Ar. The 5Pt/AC - DP catalyst was prepared by a deposition-precipitation method.

AC support (2.0 g) was firstly dispersed in 200 mL of deionized water. Pb (NO₃)₂ and H₂PtCl₆ were introduced, reduced, and deposited simultaneously on the AC support by a deposition-precipitation method. The catalyst was denoted as 5Pb-5Pt/AC-Co-DP with the loadings of Pt and Pb both at 5.0 wt%. If the catalyst was further treated with Ar at 500 ℃ for 4 h, it was denoted as 5Pb-5Pt/AC-Co-DP-500Ar.

An amount of Pb (NO₃)₂ and H₂PtCl₆ were added into 4 mL of deionized water and co-impregnated onto 2.0 g of AC support. The sample was dried at 120 ℃ for 12 h and treated under H₂ or Ar atmosphere at 500 ℃ for 4 h. The metal loadings were the same as the catalysts in Section 2.1.2. The obtained catalysts were denoted as 5Pb-5Pt/AC-Co-Im-500H₂ and 5Pb-5Pt/AC-Co-Im-500Ar.

N₂ physical adsorption of the samples was performed on a Quantachrome Autosorb-1 instrument. The samples were degassed at 250 ℃ for 12 h before testing. The crystalline phases of the catalysts were tested by XRD with Cu K_α1 radiation on a PANalytical X'Pert PRO diffractometer at 40 kV and 40 mA. TEM and high-resolution TEM (HR-TEM) were performed on a JEM-2100 electron microscope operating at 200 kV. XPS was performed on a Thermo ESCALAB 250Xi X-ray spectroscopy spectrometer. The monochromatic Al K_α source (1486.6 eV) was operated at 15 kV. The C l s peak at 284.6 eV was adopted for correction of the charging effects.

The glycerol oxidation reaction was performed in a 100-mL three-neck flask in an oil bath, with 25.0 g of glycerol solution (10.0 wt%), 1.90 g of LiOH·H₂O (purity at 90%) and 0.25 g of catalyst [25]. After the reaction, the solution was filtrated, neutralized by H₂SO₄, and diluted with deionized water. An Agilent 1100 high performance liquid chromatography system equipped with a refractive index detector was applied for product analysis. In addition, an Alltech OA-1000 column was used at a separation temperature of 80 ℃. The mobile phase was a H₂SO₄ aqueous solution (0.005 mol/L) operated at a flowrate of 0.5 mL/min. All products were identified using the retention time of pure materials and quantified by an external standard method. The glycerol conversion and liquid product selectivity calculations were conducted using the following equations, with Con. representing the conversion of glycerol, S_a representing the selectivity of product a, R₀ and R_t representing the concentration of glycerol at the reaction time of 0 and t, and P_a representing the carbon molar concentration of product a at the reaction time of t. From our experiment observations and literature reports [15, 18, 24], the possible gas product was CO₂, which was in trace amount and would be converted to carbonate in any basic solutions present.

Con. = (R₀-R_t) × 100%/ R₀

S_a = P_a × 100%/∑ P_a

The N₂ physical adsorption results of the AC support and different catalysts are shown in Table 1 and Fig. 1. The specific surface areas, pore volumes, and average mesopore sizes of 5Pt/AC-DP and x Pb-5Pt/AC-Im-DP catalysts were considerably smaller than those of the AC support, whereas the micropore size remained unchanged. This result may be attributed to the partial blockage of mesopores by the deposited metal species [26]. When the Pb loading was increased to more than 3.0 wt%, the catalyst surface areas and pore volumes decreased dramatically, indicating that the blockage was considerable. The blockage was alleviated by treating the samples at 500 ℃ Ar (or H₂), as shown by the improved surface areas and pore volumes. This result is also shown from the isotherms in Fig. 1, where the introduction of Pb would result in the drop of isotherms, and the 500 ℃ gas treatment lead to the elevation of isotherms.

The XRD patterns of the x Pb-5Pt/AC-Im-DP catalysts are shown in Fig. 2(a). Characteristic diffraction peaks at 39.8°, 46.2°, 67.5°, and 81.3° are assigned to the Pt (111), (200), (220), and (311) reflections, respectively (JCPDS, 00-004-0802). Metallic Pt particles were present in all the samples. When Pb was supported on the catalysts at a loading of 3.0 wt% or higher, the characteristic reflection peaks of Pb₃(CO₃)₂(OH)₂ (JCPDS, 00-013-131) were observed. The major peaks at 34.2°, 27.1°, 24.6°, and 20.9° are indexed as (110), (015), (104), and (012) planes, respectively.

The XRD patterns of 5Pb-5Pt/AC catalysts prepared by different methods are shown in Fig. 2(b). They were denoted with their preparation method because of their same metal loadings. Characteristic Pt peaks were observed in all the 5Pb-5Pt/AC catalysts except Co-Im-500H₂ (sample (5)). This indicates that Pt particles were highly dispersed on the Co-Im-500H₂ catalyst. In contrast, the Co-Im-500Ar (sample (6)) catalyst shows more intense metallic Pt peaks. This result may be related with the availability of reducers during the reduction process, because H₂ (the reducer of Co-Im-500H₂) is more accessible than AC (the reducer of Co-Im-500Ar). From the XRD patterns, the particle sizes of Pt were calculated using the Scherrer equation (Table 2). The 500 ℃ Ar treatment of Im-DP and Co-DP catalysts also creates larger Pt particles (untreated at 8 nm vs treated at 14 nm).

Similar to the Im-DP catalyst, characteristic peaks of Pb₃(CO₃)₂(OH)₂ were also observed in Co-DP catalyst. After 500 ℃ Ar treatment, these signals disappeared because of the decomposition of Pb₃(CO₃)₂(OH)₂ [28]. Platinum lead alloys emerged, i.e., PtPb and Pt _x Pb (JCPDS, 00-006-0374 and 00-006-0574). For Co-Im-500H₂ and Co-Im-500Ar catalysts, neither Pb₃(CO₃)₂(OH)₂ nor platinum lead alloys were detected.

The XPS spectra of the x Pb-5Pt/AC-Im-DP catalysts were conducted (Fig. 3) to study the oxidation states of the surface metals on the catalysts. The spectra data was processed by a software called XPS Peak 4.1. The backgrounds of curves were subtracted by the Shirley method. A Lorentzian function was convoluted with an experimental Gaussian curve (G = 0.2) for data analysis. The position gap between deconvoluted Pt 4 f_7/2 and Pt 4 f_5/2 peaks was fixed at 3.3 eV, whereas that for Pb 4 f peaks was fixed at 5.0 eV. The full width at half maximum of the curve of each group of deconvoluted Pt 4 f_7/2 and Pt 4 f_5/2 peaks was kept the same, with the peak area of Pt 4 f_5/2 being 0.75 times of the area of Pt 4 f_7/2. A similar method was applied for the Pb 4 f spectra.

The Pt 4 f spectra of monometallic 5Pt/AC-DP catalyst and bimetallic x Pb-5Pt/AC-Im-DP catalysts are shown in Fig. 3(a), with two doublet peaks originated from the spin-orbital splitting of the Pt 4 f_7/2 and Pt 4 f_5/2 states. The deconvoluted peaks of monometallic 5Pt/AC-DP catalyst at 71.5 and 74.8 eV were ascribed to the metallic form of Pt, meaning that its surface Pt species generally remained unoxidized [26]. However, for Pb-promoted catalysts, the deconvoluted peaks at 72.6-72.8 and 75.9-76.1 eV indicate the presence of oxidized Pt species, similar to the results in our previous report [16].

For Pb 4 f spectra in Fig. 3(b), the binding energy (BE) peaks were deconvoluted into two or three groups. The BE peaks at 138.5, 137.0-137.4, and 135.5-135.6 eV were ascribed to the Pb⁴⁺, Pb²⁺, and Pb⁰ species, respectively [29-31]. Unlike samples (3), (4), and (5), almost no Pb⁰ species was detected on the surface of the 1Pb-5Pt/AC-Im-DP catalyst. Therefore, the overall oxidation state of Pb species on 1Pb-5Pt/AC-Im-DP is higher than other x Pb-5Pt/AC-Im-DP catalysts, influencing the oxidation state of the surface Pt species, which showed a higher BE peak at 72.8 eV than the other three catalysts at 72.6 eV (Fig. 3(a)).

To observe the influence of the 500 ℃ Ar treatment on the metal oxidation state, two representative 5Pb-5Pt/AC catalysts (i.e., the Co-DP and the Co-DP-500Ar catalysts) were adopted for XPS analysis (Fig. 4). The Pt 4 f peaks (slightly) and the Pb 4 f peaks (obviously) shifted to a lower BE position after Ar treatment at 500 ℃. Considering that carbon is frequently used as a reducing agent, the treatment of Co-DP catalyst may favor the reduction of metals by the AC support. In addition, the Pt 4 f peaks of 5Pb-5Pt/AC-Co-DP (Fig. 4(a), sample (1)) and 5Pb-5Pt/AC-Im-DP catalysts (Fig. 3(a), sample (4)) were of similar shape and position, whereas their Pb 4 f peaks were obviously different (Fig. 4(b) vs Fig. 3(b)). In a previous discussion, these two catalysts were of similar XRD patterns (Fig. 2(b)), but the XPS analysis demonstrated that their surface Pb species were different, which would result in different catalytic performance.

The TEM images of two groups of representative catalysts (Im-DP vs Im-DP-500Ar, Co-DP vs Co-DP-500Ar) are presented in Fig. 5. The Ar treatment at 500 ℃ resulted in the metal particle aggregation (Fig. 5(b) vs Fig. 5(d)), which agrees with the XRD results. The HR-TEM images demonstrated the respective characteristic lattice fringes of the metallic particles. In images of Im-DP and Co-DP, only characteristic patterns of Pt (111) planes (d = 2.28 Å) (JCPDS, 00-004-0802) were observed. However, PtPb alloys (JCPDS, 00-006-0374) were also detected in Im-DP-500Ar and Co-DP-500Ar images. In Im-DP-500Ar, PtPb (101) planes (d = 3.04 Å) were observed, whereas in Co-DP-500Ar, PtPb (100) and (102) planes with d = 3.68 and d = 2.20 Å, respectively, were observed. This also agrees with the XRD analysis that Ar thermal treatment favored the Pt-Pb alloys formation.

Table 3 shows the results of glycerol oxidation reactions over 5Pt/AC-DP and x Pb-5Pt/AC-Im-DP catalysts with Pb loading ranging from 1.0 to 7.0 wt%. From our previous optimization studies of alkaline salts and their ratios, we adopted LiOH in this work [25]. Although the monometallic Pb/AC catalyst showed no activity in converting glycerol (data not shown), the addition of 1.0 wt% Pb to 5Pt/AC catalyst improved catalyst turnover frequency (TOF) compared with the monometallic 5Pt/AC-DP catalyst. For other x Pb-5Pt/AC-Im-DP catalysts, the increase of Pb loading resulted in a decrease in catalyst TOF (as well as glycerol conversion), and the changing trend is consistent with that of their N₂ physical adsorption data. This result indicates that the decrease in surface area and pore size can cause the decrease in catalyst activity (Table 1).

As shown in the XRD analysis (Fig. 2(a)), the increase in Pb loading ( > 1.0 wt%) leads to Pb₃(CO₃)₂(OH)₂ formation. This might be one of the reasons for the decreased surface area and pore volume. For the 1Pb-5Pt/AC-Im-DP catalyst, although its surface area was smaller, its catalytic performance is higher than that of 5Pt/AC-DP, and no characteristic peaks for Pb₃(CO₃)₂(OH)₂ were detected. Therefore, Pb₃(CO₃)₂(OH)₂ might be the main reason for the catalyst activity decrease at high Pb loading ( > 1.0 wt%).

In addition, the analysis of Pb 4 f spectra (Fig. 3(b)) of x Pb-5Pt/AC-Im-DP catalysts suggested that surface Pb⁰ species are unfavorable for the glycerol conversion. Because the Pb⁰ species were observed in all x Pb-5Pt/AC-Im-DP catalysts except the 1Pb-5Pt/AC-Im-DP catalyst. The role of Pt species in the enhanced glycerol conversion of the 1Pb-5Pt/AC-Im-DP catalyst could not be determined, because Pt⁰ and Pt²⁺ species were detected in all the bimetallic catalysts (Fig. 3(a)).

In terms of the target product PA, the introduction of Pb showed enhanced PA selectivity, which may be because of the presence of Pb and the generation of oxidized Pt species in all the four x Pb-1Pt/AC-Im-DP catalysts, which was caused by the Pb promoter as shown in the XPS analysis (Fig. 3(a)). From the product distributions, the introduction of Pb favored PA production and contributed to glyceric acid selectivity. However, the productions of LA and oxalic acid were decreased. This result may be because the presence of Pb favored LA consecutive oxidation towards PA, but not glyceric acid consecutive oxidation and C-C cleavage reaction to oxalic acid [32]. The 5Pb-5Pt/AC-Im-DP catalyst showed the highest performance with a PA selectivity over 14%.

To further explore the previous speculation on the roles of surface Pb⁰ species and Pb₃(CO₃)₂(OH)₂, we prepared a series of 5Pb-5Pt/AC catalysts by different methods. Fig. 6(a)shows the catalyst TOF, and Fig. 6(b) shows the glycerol conversion over these 5Pb-5Pt/AC catalysts with time on stream. As discussed in Section 3.2.1, the decreases in glycerol conversion were related with the Pb⁰ species. The comparison of catalyst TOF in Fig. 6(a) and glycerol conversion in Fig. 6(b) between Im-DP and Co-DP could further investigate the role of Pb⁰ species. The XRD patterns of Im-DP and Co-DP were similar (Fig. 2), with Pb₃(CO₃)₂(OH)₂ observed in both of them. Their Pt 4f spectra were also of similar shape and position. However, their Pb 4f spectra (Fig. 3(b) sample (4) vs Fig. 4(b) sample (1)) were different, with the Pb⁰ species present only in the Im-DP catalyst. Thus, the lower TOF and glycerol conversion of the Im-DP catalyst might be related with the Pb⁰ species.

The role of Pb₃(CO₃)₂(OH)₂ could be illustrated by the following comparison. For the Im-DP and Co-DP catalysts, their TOF and glycerol conversions were generally enhanced after treatment under Ar at 500 ℃. On the one hand, this suggests that the removal of Pb₃(CO₃)₂(OH)₂ might be beneficial to glycerol conversion. On the other hand, such an improvement might be related with the formation of PtPb and Pt _x Pb alloys. The BE peak shift of Pb to a lower state might be associated with the catalyst performance. The decomposition of Pb₃(CO₃)₂(OH)₂ and the formation of Pt _x Pb alloys can cause a BE peak shift. Thus, further analysis is still necessary to clarify the mechanism. For simplicity, the following discussion focuses on the Pb₃(CO₃)₂(OH)₂ and Pt _x Pb alloy species. For Co-Im-500H₂ and Co-Im-500Ar catalysts, Ar or H₂ treatment at 500 ℃ prevented Pb₃(CO₃)₂(OH)₂ formation; moreover, neither PtPb nor Pt _x Pb alloys were observed. They still showed higher TOF and glycerol conversion than the Im-DP and Co-DP catalysts. Therefore, the removal of Pb₃(CO₃)₂(OH)₂ is essential for enhancing catalyst activity.

Similar to the enhancement effect on catalyst activity, the PA selectivities of Im-DP and Co-DP catalysts were improved after their treatment at 500 ℃ in Ar (Fig. 6(c)). The PA selectivity of Co-DP-500Ar was almost 4.6 times higher than that of the untreated Co-DP catalyst, the PA yield reached 18.4% under the present reaction conditions. Thus, the PA selectivity might also be related with the removal of Pb₃(CO₃)₂(OH)₂ and the formation of platinum lead alloys.

The role of PtPb and Pt _x Pb alloys could be partly demonstrated by the comparison between the Co-DP-500Ar and the Co-Im-500Ar catalysts. In both cases, the Pt and Pb metals were reduced simultaneously on AC, and their calculated Pt particle sizes were similar (Table 2). The Co-DP-500Ar catalyst contains two more species than the Co-Im-500Ar catalyst, which were the PtPb and Pt _x Pb alloys. The former catalyst showed slightly better glycerol conversion and much improved PA selectivity than those of the latter one (Fig. 6). Therefore, the platinum lead alloys might favor the glycerol activation slightly and its transformation towards PA greatly. For fresh catalysts with PtPb and Pt _x Pb alloys present (i.e., Im-DP-500Ar and Co-DP-500Ar catalysts), their PA selectivity maintained the highest among all the catalysts tested.

We characterized three representative spent catalysts (i.e., Co-DP, Co-DP-500Ar and Co-Im-500Ar catalysts, with three different phases), to find more information about the role of platinum lead alloys. The alloys were observed only on fresh Co-DP-500Ar catalyst, yet they were observed on all the three spent catalysts, as confirmed by both HR-TEM and XRD analyses (Figs. 7 and 8). In addition, the XPS analysis of spent Co-DP and Co-DP-500Ar showed that the Pb 4 f peak of Co-DP shifted from 139.4 to 138.6 eV (Fig. 4(b) vs Fig. 9(b)) after the reaction. The surface Pb species on Co-DP-500Ar catalyst remain the same before and after the reaction, whereas the Pb species on spent Co-DP catalysts resemble that of the spent Co-DP-500Ar catalyst, which agrees with the HR-TEM and XRD results. The spent Co-DP and Co-DP-500Ar catalysts were evaluated under the same reaction conditions, and they exhibited slightly better PA selectivity than the fresh ones (Table 4). This suggested that the PtPb and Pt _x Pb alloys might be the active sites for PA production. However, their formation and catalytic mechanisms still require further investigation.

We propose a possible reaction pathway for glycerol oxidation to PA and other products (Scheme 1). Glycerol is first dehydrogenated to form intermediates, followed by their dehydration and intramolecular Cannizzaro rearrangement to LA [33, 21]. The introduction of Pb was beneficial for glycerol conversion when Pb₃(CO₃)₂(OH)₂ and surface Pb⁰ species were absent, and the presence of platinum lead alloys favored the transformation of LA to PA.

A one-step production of PA by selective oxidation of glycerol was investigated. Both the Pb loading and the preparation methods greatly affect the catalyst performance, because of the alteration of Pt and Pb species. The Pt-Pb alloys are favorable for glycerol transformation to PA, whereas the Pb₃(CO₃)₂(OH)₂ and the surface Pb⁰ species adversely affect glycerol conversion. By treating the 5Pb-5Pt/AC-Im-DP (or 5Pb-5Pt/AC-Co-DP) catalyst under a 500 ℃ Ar atmosphere, Pb₃(CO₃)₂(OH)₂ can be prevented, and platinum lead alloys can form. The highest PA yield was 18.4% under the present reaction conditions. Further optimization and better understanding of the reaction are required.