Glycerol Oxidation over Supported Gold Catalysts: The Combined E ﬀ ect of Au Particle Size and Basicity of Support

: Gold nanoparticles supported on various oxides (CeO 2 , CeO 2 / TiO 2 , MgO, MgO / TiO 2 , La 2 O 3 , La 2 O 3 / TiO 2 ) (with 4 wt.% Au loading) were investigated in the liquid (aqueous) phase oxidation of glycerol by molecular oxygen under mild conditions, in the presence of alkaline earth (CaO, SrO and MgO) or alkaline (NaOH) bases. Full conversion and selectivity between 38 and 68% to sodium glycerate were observed on di ﬀ erent Au supported catalysts (Au / MgO / TiO 2 , Au / La 2 O 3 / TiO 2 , Au / CeO 2 and Au / CeO 2 / TiO 2 ). The combined e ﬀ ect of Au particle size and basicity of the support was suggested as the determining factor of the activity. Agglomeration of gold nanoparticles, found after the reaction, led to the deactivation of the catalysts, which prevents the further oxidation of sodium glycerate into sodium tartronate. Promising results were obtained with the use of alkaline earth bases (CaO, SrO, MgO), leading to the formation of free carboxylic acids instead of salts, which are formed in the presence of the more usual base, NaOH.


Introduction
Glycerol, a highly functionalized molecule, was identified as one of the top twelve most important bio-based chemicals in the world by the US Department of Energy [1]. Glycerol has a wide range of applications, from the production of food additives to pharmaceuticals, cosmetics, personal care products and detergents, and it is also transformed into non-toxic solvents [2][3][4][5].
Nowadays, glycerol is mostly generated as an undesired byproduct (around 10% of the total volume) during the production of biodiesel, one of the most important and valuable alternative liquid biofuels in the transportation sector [6][7][8]. In 2017, the production of biodiesel exceeded 21 million tons per year just in the EU [9]. Thus, despite the wide market for glycerol, the increasing growth  Table 1 briefly summarizes some of the catalytic results from research [27][28][29][30][31][32][33][34][35] on glycerol oxidation with molecular oxygen under pressure over gold supported catalysts in the presence of a base under different reaction conditions.
Carretin et al. [27] reported the selective oxidation of glycerol to glyceric acid (100% selectivity), probably via the initial formation of glyceraldehydes using gold supported on activated carbon or graphite under mild reaction conditions after 3 h. It was also noted that, with high concentrations of NaOH, exceptionally high selectivity to glyceric acid can be obtained. However, by decreasing the glycerol to gold ratio (R) from 540 to 214 and the oxygen concentration, an increase of glycerol conversion is obtained, with the formation of some tartronic acid via consecutive oxidation.
Porta and Prati [28] obtained a high selectivity towards glyceric acid (92%) with a 90% conversion of glycerol on a Au/C catalyst by optimizing the temperature and a NaOH/glycerol ratio. The higher the temperature, the higher the amount of tartronic acid formed (however, no data on selectivity values were provided). Also, larger gold particles (30 nm size) that were obtained with immobilization methods, maintained constant selectivity while particles with a smaller mean size (6 nm) obtained with incipient wetness or impregnation methods showed higher activity with a rapid change in selectivity. Besides that, the authors established the existence of two main pathways one leading to hydroxyacetone and glyceric aldehyde as the main oxidation products, and the other to glycolic and oxalic acids as the final compounds. In addition, the rapid oxidation of glyceraldehyde favors the formation of glyceric acid, rather than that of hydroxyacetone (Scheme 1).
Cai et al. [29] observed high catalytic performance (90-99% conversion after 9 h) in glycerol oxidation with different selectivity values on Au NPs deposited on various supports (CeO2, activated carbon and different Y type zeolites). However, it should be noted that a low ratio of glycerol to Au (R = 150) was used in all cases. Preferential formation of tartronic acid (from 44 to 82%) was observed when Y type zeolite was used as the support, while the primary products on Au/CeO2 and Au/C were oxalic acid (55%) and glyceric acid (89%), respectively. The authors proposed that the main reason for the formation of tartronic acid was the small size of the Au nanoclusters (1 nm) on the HY zeolite support.
Villa et al. [30] described high activity (full conversion) of PVA protected Au nanoparticles (NPs) supported on NiO for glycerol oxidation, but with lower selectivity (55%) to glycerate after 4 h of reaction. Under the same reaction conditions, Au/TiO2 prepared the same way achieved 81% of glycerate selectivity but 64% for glycerol conversion. The improvement in activity by using NiO as the support was related to the stronger interaction between the support and the Au NPs. The XPS measurements revealed only Au 0 species on both support (NiO and TiO2) surfaces. Porta and Prati [28] obtained a high selectivity towards glyceric acid (92%) with a 90% conversion of glycerol on a Au/C catalyst by optimizing the temperature and a NaOH/glycerol ratio. The higher the temperature, the higher the amount of tartronic acid formed (however, no data on selectivity values were provided). Also, larger gold particles (30 nm size) that were obtained with immobilization methods, maintained constant selectivity while particles with a smaller mean size (6 nm) obtained with incipient wetness or impregnation methods showed higher activity with a rapid change in selectivity. Besides that, the authors established the existence of two main pathways one leading to hydroxyacetone and glyceric aldehyde as the main oxidation products, and the other to glycolic and oxalic acids as the final compounds. In addition, the rapid oxidation of glyceraldehyde favors the formation of glyceric acid, rather than that of hydroxyacetone (Scheme 1).
Cai et al. [29] observed high catalytic performance (90-99% conversion after 9 h) in glycerol oxidation with different selectivity values on Au NPs deposited on various supports (CeO 2 , activated carbon and different Y type zeolites). However, it should be noted that a low ratio of glycerol to Au (R = 150) was used in all cases. Preferential formation of tartronic acid (from 44 to 82%) was observed when Y type zeolite was used as the support, while the primary products on Au/CeO 2 and Au/C were oxalic acid (55%) and glyceric acid (89%), respectively. The authors proposed that the main reason for the formation of tartronic acid was the small size of the Au nanoclusters (1 nm) on the HY zeolite support.
Villa et al. [30] described high activity (full conversion) of PVA protected Au nanoparticles (NPs) supported on NiO for glycerol oxidation, but with lower selectivity (55%) to glycerate after 4 h of reaction. Under the same reaction conditions, Au/TiO 2 prepared the same way achieved 81% of glycerate selectivity but 64% for glycerol conversion. The improvement in activity by using NiO as the support was related to the stronger interaction between the support and the Au NPs. The XPS measurements revealed only Au 0 species on both support (NiO and TiO 2 ) surfaces.
Wang et al. [31], investigated Au catalysts supported on nanofibers with different degrees of graphitization under the same reaction conditions used in [30]. They found a similar size of Au NPs (3.5-3.8 nm) with a similar level of activity in glycerol oxidation (64-70%), but with different product distributions. The authors attributed this to the different shape and configuration of the Au particles. On Au/CNF-PS, 55% of glyceric acid (C3 product) and 45% of formic and glycolic acid (C1 + C2 products) were obtained, while 22% of glyceric acid and 77% of formic and glycolic acid were observed using Au/CNF-HT. The high selectivity of the latter catalyst to C1 and C2 products, which derives from the C-C bond cleavage, was supposedly associated with the exposure of low index Au planes surfaces caused by the direct contact between CNF-HT and the {111} surface of PVA Au NPs, while disordered carbon nanofiber surfaces (CNF-PS) led to random orientation of supported particles.
Sobczak et al. [28] obtained a Au/Nb 2 O 5 catalyst with catalytic properties comparable to Au/C, prepared by the gold-sol method and crystalline niobia. After 5 h, glycerol conversion was 67% with glyceric acid as the main product. However, Au on carbon showed a different behavior. Thus, it was concluded that oxidation of glycerol is a complex process and many parameters influence the activity and selectivity of the catalysts used. Au/Nb 2 O 5 was also tested in a second glycerol oxidation run by simple solvent decantation after the first run, without any drying and washing, and a small decrease in the activity (9%) with no change in selectivity was observed. Total selectivity (<100%) could not be reached for several reasons: polymerization of the reaction products, oxidation of the reaction products to CO 2 or adsorption of the products on the support surface. It was noted that glycolic acid formation proceeds via cracking of the C-C bond by the acidic centers on the catalyst surface. The authors suggested that the higher yield of tartronic acid was caused by the slow desorption of glyceric acid. The Au-Nb 2 O 5 catalyst was also investigated by Wolski [33], who found that with Au NPs smaller than 3 nm, the catalyst ability of C-C bond cleavage was enhanced and promoted formation of C2 and C1 products.
Murthy and Selvam [34] reported 82-84% glycerol conversion after 5 h on mesoporous carbon supported nano-gold catalysts synthesized by the sol-immobilization method, with 70% selectivity to glyceric acid. The remarkable performance of both fresh and regenerated catalysts (second cycle), was attributed to the presence of pore channels in mesoporous carbons, which act as nanoreactors preventing the agglomeration of nanoparticles.
Zope et al. [35] studied the factors that influence the formation of diacids (tartronic and oxalic acids) in glycerol oxidation and of furandicarboxylic acid in 5-hydroxymethylfurfural (HMF) oxidation over gold supported catalysts (Au/TiO 2 and Au/C). Interestingly, hydrotalcite was used as an alternative to liquid bases for HMF reaction, which requires further neutralization. To effectively produce diacids, the use of a continuous reactor or high Au catalyst loadings in a semi-bath reactor was needed. Such conditions are necessary to prevent the formation of trace byproduct species from monoacid, since such species inhibit the rate of both glycerol oxidation and monoacid oxidation to diacid. Thus, the best results were obtained in the semi-bath reactor with R = 350: full conversion and 30% of diacid (25% of tartronic acid and 5% of oxalic acid) over Au/TiO 2 catalyst after 3 h at 60 • C under 11 bar of O 2 .
After analyzing these works, it can be concluded that it is essential to control the activity and selectivity in the process of liquid-phase oxidation of glycerol, and a comprehensive approach is necessary. This includes taking into account the particle size and shape of gold, the choice of the support and reaction conditions, that is, the amount of base and gold added relative to alcohol, and reaction temperature.
The aim of this work was to study the influence of the nature of the support and the base additives on the catalytic performance of gold catalysts in the liquid phase oxidation of glycerol by molecular oxygen.
Pretreatment in a hydrogen atmosphere for 1 h at a temperature of 300 • C was used for the decomposition of the products of the hydrolysis of the complex of gold (III) with urea on the support surface [39]. The catalysts were characterized by adsorption-desorption of N 2 at −196 • C on a Micromeritics 23x-Tristar 3000 Apparatus, Micromeritics Instrument Corporation (Norcross, GA, USA), X-ray diffraction (XRD) on a Philips XPert PRO diffractometer (Amsterdam, Netherlands), X-ray photoelectron spectroscopy (XPS), on a ESCALAB 200A, VG Scientific (Waltham, MA, USA), energy-dispersive X-ray spectroscopy (EDX), transmission electron microscopy (TEM), as well as scanning transmission electron microscopy-high angle annular dark field (STEM-HAADF) using one single microscope (JEOL JEM-2100F, JEOL Ltd., Tokyo, Japan) and the temperature-programmed carbon dioxide (CO 2 -TPD) method on a "Chemosorb" chemisorption analyzer (Neosib, Novosibirsk, Russia) as previously reported in [36][37][38][40][41][42], where the detailed description of the applied procedures can be found.
The size of the gold NPs used in the (i.e., after the reaction test) catalysts was investigated by high resolution transmission electron microscopy (HRTEM) using a JEM-2100F instrument. The samples were ground to a fine powder and sonicated in hexane at room temperature. Then, a part of the suspension was placed on a lacey carbon-coated Cu grid. For each sample, at least 300 particles were counted.
The determination of the basic sites is described in detail in previous papers by our group [37,38]. Briefly, the CO 2 -temperature-programmable desorption (TPD) method was applied using a "Chemosorb" chemisorption analyzer (Neosib, Novosibirsk, Russia), equipped with a thermal conductivity detector (TCD). After treatment in an inert atmosphere, samples were saturated with CO 2 for 1 h at 25 • C and the temperature was increased to 600 • C at a rate of 10 • C min −1 under an inert atmosphere. Total basicity was calculated from the area of the desorption peaks.
The catalytic tests were carried out in a semi-batch reactor operated under 3 atm of oxygen at 50 • C under stirring at 1100 rpm. Typically, the catalyst sample was added in a glycerol/gold ratio R = 1000 mol/mol to 10 mL of 0.3 M glycerol (87 wt.% solution, Fluka, Maurice, NJ, USA) and 1.2 M NaOH aqueous solution (distilled water) in a glass reactor equipped with heater, mechanical stirrer, gas supply system and thermometer. Small aliquots of the reacting mixture were taken after 15, 30, 60, 120 and 180 min to monitor the reaction progress. Catalytic experiments with alkaline earth bases (MgO, SrO, CaO) were performed with Au/La 2 O 3 /TiO 2 catalyst under the conditions described above (R = 1000, pO 2 = 3 atm) with different glycerol/base molar ratios (4 or 12) and reaction temperatures (50, 80 or 95 • C),over a 6 h period.
The reactants and products were analyzed and quantified by high-performance liquid chromatography (HPLC, Agilent Technologies, 1220 Infinity, Santa Clara, CA, USA) using a column Alltech OA-10308 (L × i.d.: 300 mm × 7.8 mm, Fisher Scientific, Hampton, NH, USA) with ultraviolet (Varian 9050 UV, 210 nm) and refractive index (Waters RI) detectors, using 0.4 mL/min of 0.1% aqueous solution of H 3 PO 4 as the eluent. Attribution of peaks was made by comparison with chromatograms of standard samples. No activity of oxidation of glycerol was observed in the absence of support/catalyst.
The conversion of glycerol and selectivity to products were calculated in terms of moles of C atoms, according to Equations (1) and (2), respectively: Selectivity (% mol) = C moles of product formed C moles of all products × 100 Turnover frequency (TOF) was calculated with regard to the number of moles of gold during the first 15 min as follows: Conversion after 15 min × initial moles of glycerol moles of Au × 100 × 0.25 h The carbon balance in all the reported test data was within 100 ± 5%.

Catalytic Results
In our previous work [36,37,40], we found that for alcohol oxidation on gold supported catalysts, the optimal gold content was 4 wt.% and the optimal pretreatment was a reducing atmosphere. For this reason, glycerol oxidation was investigated on samples with 4 wt.% Au after pretreatment in H 2 ( Figure 1). the first 15 min as follows: Conversion after 15 min × initial moles of glycerol moles of Au × 100 × 0.25 h The carbon balance in all the reported test data was within 100 ± 5%.

Catalytic Results
In our previous work [36,37,40], we found that for alcohol oxidation on gold supported catalysts, the optimal gold content was 4 wt.% and the optimal pretreatment was a reducing atmosphere. For this reason, glycerol oxidation was investigated on samples with 4 wt.% Au after pretreatment in H2 ( Figure 1). It can be seen that samples of Au on modified titanium oxides and pure cerium oxide were extremely active. The full conversion of glycerol was observed after only 1 h on Au/MgO/TiO2 and Au/La2O3/TiO2 catalysts, and after 2 h on Au/CeO2/TiO2 and Au/CeO2 samples. Au/La2O3 and Au/TiO2 showed almost the same behavior with regard to the progress of the conversion with run time, and obtained a 51-56% conversion rate after 3 h. The least active sample was Au/MgO with a 27% conversion after 3 h.
In terms of TOF, after 15 min ( Table 2) the order of the catalyst's initial activity in glycerol oxidation was: Au/MgO/TiO2 ≥ Au/La2O3/TiO2 > Au/CeO2 > Au/CeO2/TiO2 > Au/La2O3 > Au/TiO2 > Au/MgO. It can be seen that samples of Au on modified titanium oxides and pure cerium oxide were extremely active. The full conversion of glycerol was observed after only 1 h on Au/MgO/TiO 2 and Au/La 2 O 3 /TiO 2 catalysts, and after 2 h on Au/CeO 2 /TiO 2 and Au/CeO 2 samples. Au/La 2 O 3 and Au/TiO 2 showed almost the same behavior with regard to the progress of the conversion with run time, and obtained a 51-56% conversion rate after 3 h. The least active sample was Au/MgO with a 27% conversion after 3 h.
In terms of TOF, after 15 min ( Table 2) the order of the catalyst's initial activity in glycerol oxidation was: In all cases, the main product was glyceric acid (38-77% selectivity) with some formation of glycolic, tartronic, formic, lactic and oxalic acids, after 3 h of the reaction (Figure 2). The exception in the selectivity trends was the Au/CeO 2 /TiO 2 sample, where a higher amount of tartronic acid (up to  Table 2, entry 7). It is also worth noting that with increasing time, for all of the active samples the selectivity changes towards the formation of tartronic acid as a product of the oxidation of glyceric acid (Figure 2b-d,f).
Processes 2020, x FOR PEER REVIEW 9 of 19 In all cases, the main product was glyceric acid (38-77% selectivity) with some formation of glycolic, tartronic, formic, lactic and oxalic acids, after 3 h of the reaction (Figure 2). The exception in the selectivity trends was the Au/CeO2/TiO2 sample, where a higher amount of tartronic acid (up to 32%) was formed ( Table 2, entry 7). It is also worth noting that with increasing time, for all of the active samples the selectivity changes towards the formation of tartronic acid as a product of the oxidation of glyceric acid (Figure 2b-d,f).  It should be mentioned that for all catalysts, only 2-5% of glycerol conversion was observed in the absence of bases (R = 500, T = 80 °C, 6 h).

Study of the Influence of Alkaline Earth Base Additives
The above results show the efficiency of our catalysts in the presence of alkali in comparison with the other gold supported materials presented in Table 1. However, from an industrial point of view, there is a major technical advantage when base-free conditions are used, that is, the formation of free carboxylic acid instead of the corresponding salts, particularly when sodium hydroxide is applied. Since we observed that our catalysts had very low activity in the absence of NaOH, we replaced it with alkaline earth bases such as MgO, CaO and SrO in order to obtain free carboxylic acids (Table 3). Au/La₂O₃/TiO₂ (one of the most active catalysts in this study) was used for these experiments. As can be seen in Table 3, the order of activity, in terms of conversion, was MgO > CaO > SrO. In particular, the addition of alkaline earth bases had a positive effect on the activity, especially when MgO was used. In all cases, higher temperatures led to an increase in the conversion of glycerol (Table  3, 3rd column). It should be mentioned that for all catalysts, only 2-5% of glycerol conversion was observed in the absence of bases (R = 500, T = 80 • C, 6 h).

Study of the Influence of Alkaline Earth Base Additives
The above results show the efficiency of our catalysts in the presence of alkali in comparison with the other gold supported materials presented in Table 1. However, from an industrial point of view, there is a major technical advantage when base-free conditions are used, that is, the formation of free carboxylic acid instead of the corresponding salts, particularly when sodium hydroxide is applied. Since we observed that our catalysts had very low activity in the absence of NaOH, we replaced it with alkaline earth bases such as MgO, CaO and SrO in order to obtain free carboxylic acids (Table 3). Au/La 2 O 3 /TiO 2 (one of the most active catalysts in this study) was used for these experiments. As can be seen in Table 3, the order of activity, in terms of conversion, was MgO > CaO > SrO. In particular, the addition of alkaline earth bases had a positive effect on the activity, especially when MgO was used. In all cases, higher temperatures led to an increase in the conversion of glycerol (Table 3, 3rd column).
The product distribution showed that a larger quantity of glyceric acid was formed when glycolic acid was the second product. Moreover, on SrO and CaO, formic acid formed in approximately equal amounts as the glycolic acid (20-28%). Small quantities of tartronic and oxalic acid were determined as well. When MgO was added, the formic acid disappeared, resulting in higher conversion values (Table 3, entries 7 and 8).
Although full conversion of glycerol was not achieved, the use of alkaline earth bases is definitely advantageous compared to that of NaOH, since it promotes free carboxylic acids formation instead of corresponding salts. Also, we can assume that a higher performance could be obtained with proper development of the catalyst composition and reaction conditions.

Catalyst Characterization
The characterization of the samples used in this work is reported in our previous papers [36][37][38][40][41][42] and this is summarized in Table 4.  Figure S1) showed no alteration in the phase composition or the support structure after gold loading, except: (1) Au/La 2 O 3 ( Figure S1e), which showed peaks related to La(OH) 3 (hexagonal) and La 2 (CO 3 ) 2 (OH) 2 (orthorhombic) [38]; and (2) Au/MgO (Figure S1g), where transformation of the oxide to hydroxide Mg(OH) 2 occurred [42]. Additionally, no peaks related to Au were found, indicating either the small size of the Au particles and metal oxides or their amorphous structure. The metallic gold phase was found only in the case of the Au/MgO sample, with an average crystal size of 3.1 nm [42].
In all cases, the Au content detected by EDX analysis was close to the nominal one (Table 4). N 2 adsorption-desorption revealed no significant changes in the specific surface area after gold deposition on each support, except Au/MgO, where a 2.2-fold increase in S BET was found for MgO after gold addition, which is consistent with the changes in the support structure and composition revealed by XRD ( Figure S1g).
The mean particle size of gold as determined by HRTEM varied from 2.4 to 5.1 nm for modified titania samples, while gold on pure oxides was in the range of 2.6-6.1 nm ( Figure S2). In contrast to numerous studies on the oxidation of glycerol that show the direct dependence of the catalytic activity on the average size of the Au NPs (the smaller the NPs size, the higher the activity) [28,31,33,34], this trend was not observed in the present work ( Figure 3).
For example, the two most active catalysts, Au/MgO/TiO 2 and Au/La 2 O 3 /TiO 2 have similar TOF values ( Table 2), but they have different average particle size of 5.1 and 2.6 nm, respectively. Another example is the comparison of Au/TiO 2 and Au/MgO/TiO 2 ( Figure 3). Thus, it can be assumed that the average size of the Au NPs is not the only determining factor.
The XPS results showed the presence of gold in four different electronic states: metallic (Au 0 ), monovalent (Au + ), trivalent (Au 3+ ) and electronegative (Au n δ− ) ( Table 4, Figure S3). No correlation was found between the surface concentration of the gold states and the catalytic data. It is known that the basic properties of catalysts play an important role in the oxidation of alcohols. To assess their impact, CO 2 temperature-programmed desorption experiments were carried out for both the catalysts and supports (Table 5) [37,38]. Processes 2020, x FOR PEER REVIEW 13 of 19 Figure 3. Turnover frequency (TOF) in glycerol oxidation for gold supported catalysts as a function of gold mean particle size.
For example, the two most active catalysts, Au/MgO/TiO2 and Au/La2O3/TiO2 have similar TOF values ( Table 2), but they have different average particle size of 5.1 and 2.6 nm, respectively. Another example is the comparison of Au/TiO2 and Au/MgO/TiO2 ( Figure 3). Thus, it can be assumed that the average size of the Au NPs is not the only determining factor.
The XPS results showed the presence of gold in four different electronic states: metallic (Au 0 ), monovalent (Au + ), trivalent (Au 3+ ) and electronegative (Aun δ− ) (Table 4, Figure S3). No correlation was found between the surface concentration of the gold states and the catalytic data.
It is known that the basic properties of catalysts play an important role in the oxidation of alcohols. To assess their impact, CO2 temperature-programmed desorption experiments were carried out for both the catalysts and supports (Table 5) [37,38]. Table 5. Basic properties of the catalysts and corresponding supports, TOF values and mean particle size (Dmean) of the catalysts. Adapted from [37,38], except the data for MgO-based catalysts (entries 6 and 7), which is presented for the first time in this paper.
After the deposition of gold, an increase in the general basicity was observed for all samples ( Table 5, 4th column), except Au/TiO2. This means that a redistribution of the base centers occurred due to the interaction of the gold precursor, urea and/or water with the support during the preparation of the catalyst [38,41]. For Au/MgO, and especially Au/La2O3, the increased basicity was associated with a complete change in the phase composition, which can be explained by the chemical properties of these supports, and exhibited in the interaction with water and urea hydrolysis products, during the preparation of the catalyst. Namely, for Au/MgO, the transition of the oxide to the hydroxide led to an increase of the Brønsted base centers (OH groups). For Au/La2O3, the presence . Turnover frequency (TOF) in glycerol oxidation for gold supported catalysts as a function of gold mean particle size. Table 5. Basic properties of the catalysts and corresponding supports, TOF values and mean particle size (D mean ) of the catalysts. Adapted from [37,38], except the data for MgO-based catalysts (entries 6 and 7), which is presented for the first time in this paper. TiO 2 , CeO 2 /TiO 2 , and La 2 O 3 supports possess weak basic properties, CeO 2 and La 2 O 3 /TiO 2 supports have an average number of basic centers, and supports with magnesium, MgO/TiO 2 and especially MgO, have a strong basicity, as expected (Table 5, 3rd column).

Entry
After the deposition of gold, an increase in the general basicity was observed for all samples ( Table 5, 4th column), except Au/TiO 2 . This means that a redistribution of the base centers occurred due to the interaction of the gold precursor, urea and/or water with the support during the preparation of the catalyst [38,41]. For Au/MgO, and especially Au/La 2 O 3 , the increased basicity was associated with a complete change in the phase composition, which can be explained by the chemical properties of these supports, and exhibited in the interaction with water and urea hydrolysis products, during the preparation of the catalyst. Namely, for Au/MgO, the transition of the oxide to the hydroxide led to an increase of the Brønsted base centers (OH groups). For Au/La 2 O 3 , the presence of lanthanum hydroxycarbonate along with lanthanum hydroxide complicated an evaluation of the base sites for this sample due to the contribution of residual carbonates to the total release of CO 2 in addition to CO 2 desorbed from the base sites [38]. The highest basicity values (230-236 µmol/g) were observed for Au/CeO 2 and Au/MgO/TiO 2 , which showed an increase in the total basicity of 1.9 and 1.2 times, respectively, compared to the concentration of base sites for CeO 2 and MgO/TiO 2 . After the deposition of gold on CeO 2 /TiO 2 and La 2 O 3 /TiO 2 , the number of base centers increased 2.2 and 1.5 times, respectively, and the corresponding catalysts had approximately the same basicity level (153-161 µmol/g). A comparison of the obtained basicity results with TOF (Tables 2 and 4) did not reveal a direct correlation between the concentration of basic centers and the activity of the studied catalysts in glycerol oxidation (Figure 4). deposition of gold on CeO2/TiO2 and La2O3/TiO2, the number of base centers increased 2.2 and 1.5 times, respectively, and the corresponding catalysts had approximately the same basicity level (153-161 µmol/g). A comparison of the obtained basicity results with TOF (Tables 2 and 4) did not reveal a direct correlation between the concentration of basic centers and the activity of the studied catalysts in glycerol oxidation (Figure 4). However, if the TOF results are compared with the average particle size and total basicity, it is clear that samples with Au nanoparticles smaller than 2 nm and a high amount of base sites have the highest TOF values. Although Au/MgO/TiO2 is characterized by a small number of small particles and has one of the highest average sizes (5.1 nm), it had the largest TOF among all catalysts after 15 min of the reaction due to its high basicity (236 µmol/g). It is likely that the role of the basic groups of the support, as in the case of added NaOH, is to enable faster deprotonation of glycerol, thereby facilitating the primary oxidation step. Prior to the deprotonation, adsorption of glycerol and oxygen and the further oxidation of the alcohol after activation of the hydroxyl group by the base occur on gold [43,44], depending on its distribution and size on the surface of the support. So, although Au/CeO2 has half the average gold size of Au/MgO/TiO2, it showed an initial TOF value smaller than that of Au/MgO/TiO2 because Au/CeO2 has fewer basic centers compared to Au/MgO/TiO2. This also explains why Au/TiO2, with a size of 3 nm showed medium activity, as this catalyst has the lowest basicity. Unfortunately, it was not possible to directly assess the basicity of Au/La2O3 because of the presence of carbonates and higher average particle size (3.6 nm); therefore, we cannot draw definite conclusions for this sample. It should also be taken into account that the specific surface of this sample was the smallest (Table 4), which can also affect its catalytic performance. Finally, Au/MgO has the lowest TOF value, which can be explained by the fact that it has the largest Au NPs size.
Thus, we can conclude that a combination of small particle size and high concentration of basic groups is necessary for the effective oxidation of glycerol over gold supported catalysts.
However, there still remains an unanswered question as to why Au/CeO2/TiO2, which has the lowest TOF among the active catalysts, shows the highest yield of tartronic acid, a product of the oxidation of glyceric acid, as observed in the reaction products.
Perhaps, as many authors have noted [29,[45][46][47], deeper oxidation of glycerol occurs with the formation of tartronic acid, and products associated with the cleavage of the C-C bond of glycerol are formed (formic, glycolic and oxalic acids) when small particles of gold (smaller than 2 nm) are present. In order to better study the effect of the Au NPs size and assess whether the gold distribution However, if the TOF results are compared with the average particle size and total basicity, it is clear that samples with Au nanoparticles smaller than 2 nm and a high amount of base sites have the highest TOF values. Although Au/MgO/TiO 2 is characterized by a small number of small particles and has one of the highest average sizes (5.1 nm), it had the largest TOF among all catalysts after 15 min of the reaction due to its high basicity (236 µmol/g). It is likely that the role of the basic groups of the support, as in the case of added NaOH, is to enable faster deprotonation of glycerol, thereby facilitating the primary oxidation step. Prior to the deprotonation, adsorption of glycerol and oxygen and the further oxidation of the alcohol after activation of the hydroxyl group by the base occur on gold [43,44], depending on its distribution and size on the surface of the support. So, although Au/CeO 2 has half the average gold size of Au/MgO/TiO 2 , it showed an initial TOF value smaller than that of Au/MgO/TiO 2 because Au/CeO 2 has fewer basic centers compared to Au/MgO/TiO 2 . This also explains why Au/TiO 2 , with a size of 3 nm showed medium activity, as this catalyst has the lowest basicity. Unfortunately, it was not possible to directly assess the basicity of Au/La 2 O 3 because of the presence of carbonates and higher average particle size (3.6 nm); therefore, we cannot draw definite conclusions for this sample. It should also be taken into account that the specific surface of this sample was the smallest (Table 4), which can also affect its catalytic performance. Finally, Au/MgO has the lowest TOF value, which can be explained by the fact that it has the largest Au NPs size.
Thus, we can conclude that a combination of small particle size and high concentration of basic groups is necessary for the effective oxidation of glycerol over gold supported catalysts.
However, there still remains an unanswered question as to why Au/CeO 2 /TiO 2 , which has the lowest TOF among the active catalysts, shows the highest yield of tartronic acid, a product of the oxidation of glyceric acid, as observed in the reaction products.
Perhaps, as many authors have noted [29,[45][46][47], deeper oxidation of glycerol occurs with the formation of tartronic acid, and products associated with the cleavage of the C-C bond of glycerol are formed (formic, glycolic and oxalic acids) when small particles of gold (smaller than 2 nm) are present. In order to better study the effect of the Au NPs size and assess whether the gold distribution changed after the reaction, we performed HRTEM of the samples in which high TOF values were achieved (Au/CeO 2 /TiO 2 , Au/CeO 2 , Au/La 2 O 3 /TiO 2 and Au/MgO/TiO 2 ) ( Figure 5). An EDX analysis was also performed in parallel, and no leaching of gold was detected after the reaction. However, for all catalysts, except Au/CeO 2 /TiO 2 , there was a noticeable increase in the average particle size after the reaction, as compared with the corresponding values before the reaction (Table 6). Such particle growth may indicate some deactivation of the catalysts, which prevents the further oxidation of glyceric acid to tartronic acid. Indeed, some tendencies were observed including the smaller the particle growth observed, the higher the amount of tartronic acid formed. all catalysts, except Au/CeO2/TiO2, there was a noticeable increase in the average particle size after the reaction, as compared with the corresponding values before the reaction (Table 6). Such particle growth may indicate some deactivation of the catalysts, which prevents the further oxidation of glyceric acid to tartronic acid. Indeed, some tendencies were observed including the smaller the particle growth observed, the higher the amount of tartronic acid formed.

Conclusions
The efficiency of gold catalysts supported on pure oxides (TiO 2 , CeO 2 , La 2 O 3 and MgO) or modified titania (CeO 2 /TiO 2 , La 2 O 3 /TiO 2 and MgO/TiO 2 ) was investigated in the aerobic oxidation of glycerol under mild (T = 50 • C, 3 atm of O 2 and added base) conditions. Gold catalysts were highly effective in the oxidation of glycerol in the presence of an alkaline base (NaOH). The order of the catalyst's initial activity (TOF in 15 min) was found to be as follows: Au/MgO/TiO 2 ≥ Au/La 2 O 3 /TiO 2 > Au/CeO 2 >Au/CeO 2 /TiO 2 > Au/La 2 O 3 > Au/TiO 2 > Au/MgO. A combination of small particle size and high concentration of basic groups was proposed as requisite for the effective oxidation of glycerol.
The main product was sodium glycerate (38-72%). However, on the most active catalysts (Au/La 2 O 3 /TiO 2 , Au/CeO 2 , Au/MgO/TiO 2 and Au/CeO 2 /TiO 2 ) that exhibited full conversion, an increase in the formation sodium tartronate was observed with run time, and the selectivity to tartronate on these materials after 3 h (glycerol/Au = 1000) was as follows: 15, 19, 25 and 32% respectively. The tendency of these catalysts is as follows: the higher the growth of Au NPs after the reaction, the less formation of tartronate, which may indicate some deactivation of the catalysts, which prevents the further oxidation of glyceric acid to tartronic acid.
Although replacing traditional alkaline base (NaOH) with alkaline earth bases (CaO, SrO and MgO) did not provide higher activity, it led to the formation of free carboxylic acids in contrast to the salts formed in the case of NaOH. The highest conversion of 20% was achieved when MgO and Au/La 2 O 3 /TiO 2 catalyst were used, giving a 52% selectivity to glyceric acid in 6 h at T = 95 • C, glycerol/MgO = 4 and glycerol/Au = 1000.
Author Contributions: E.P. carried out the preparation and pretreatment of the catalysts, performed the catalytic tests with NaOH, and wrote the first draft of the paper; E.K. was responsible for the methodology of the catalytic tests with alkaline earth base additives, supervised those experiments and participated in the writing; M.S., A.V. and L.P. dealt with the methodology of catalytic tests with NaOH; D.G. was responsible for the catalytic results with alkaline earth base additives; M.S., A.V., L.P., S.A.C.C., N.B., V.C.C. and A.P. provided the means for the realization of this work and contributed to the supervision and paper revision. All authors read and approved the final manuscript.