Perovskite Supported Catalysts for the Selective Oxidation of Glycerol to Tartronic Acid

Exceptional selectivity of LaMnO3 perovskite supported Au catalysts for the oxidation of glycerol to the dicarboxylate tartronic acid is reported. Through using monometallic Au, Pt or bimetallic Au:Pt nanoparticles the tartronic acid yield could be altered significantly, with a maximum yield of 44% in 6 h with Au/LaMnO3 and 80% within 24 h. These LaMnO3 supported catalysts were compared with conventionally TiO2 supported catalysts, which at comparable reaction conditions produced lactic acid, via a dehydration pathway, in high yield and a maximum tartronic acid yield of only 9% was observed. The LaMnO3 catalysts produced minimal lactic acid regardless of the supported metal, showing that the support structure influences the prevalence of dehydration and oxidation pathways. The choice of metal nanoparticle influenced product selectivity along the oxidation pathway for both LaMnO3 and TiO2 supported catalysts. Au catalysts exhibited a higher selectivity to tartronic acid, whereas AuPt catalysts produced glyceric acid and Pt catalysts produced predominantly C–C scission products.


Introduction
The synthesis of carboxylic acids, from the aqueous oxidation of bio-derived polyols, such as glycerol, represents an important route to upgrading biomass to pharmaceuticals, polymers and fine chemicals [1,2]. Ideally, for environmental and economic reasons, these transformations should be performed at low temperature, pressure and with the use of molecular oxygen as the oxidant.
While most of these products can be attained in high yield, the selective formation of tartronic acid remains challenging. As a highly functional dicarboxylic acid, tartronic acid is considered a promising and versatile intermediate in polymer and pharmaceutical applications, and therefore is a highly desirable product from polyol oxidation [8]. Tartronic acid is formed by the further oxidation of glyceric acid, but frequently is observed as the minor product (selectivities < 20%) compared to glyceric acid [11].
Strategies for attaining high activity and good tartronic acid yields appear to be obvious; with greater catalytic activity and more aggressive oxidising conditions forcing the sequential oxidation of glyceric acid to tartronic acid. However, the limited number of papers reporting high tartronic acid yield suggests that this is in fact not easy to achieve. Firstly, highly active metals such as Pt or Pd, which can facilitate initial hydrogen abstraction from glycerol, unfortunately also enable scission reactions [4,5,11]. Supported Au catalysts have been shown to be highly selective for glycerol oxidation towards glyceric acid [4] and therefore, with a longer reaction, could perform the sequential oxidation to tartronic acid. Unfortunately, the activity of monometallic Au catalysts is low. Indeed, Cai et. al. demonstrated a high tartronic acid yield of 80% using a Au/HY zeolite catalyst but with a very low turn-over frequency (TOF) of 40 h −1 [16]. A method of improving the catalytic activity of Au has been to alloy with Pd or Pt but, while glycerol oxidation activity can be dramatically improved, selectivities towards tartronic acid remain low at the reaction conditions tested [6,7,17]. In an alternative strategy, Jin et. al. utilised a noble metal free Co/MgO-Al 2 O 3 catalyst that produced a tartrtonic acid yield of 63.5%, but again with very low TOF of 3 h −1 [18].
In addition to catalyst design, reaction conditions such as increasing the base:glycerol ratio and elevating temperature enhances further oxidation. Both the Au/HY zeolite and Co/MgO-Al 2 O 3 catalyst systems were run with high base:glycerol ratios of 4:1 and 7:1 respectively [16,18]. Interestingly, the logical step of increasing reaction temperature, from the conventional 60-70 °C, to increase reaction rate was not employed in either study. The probable reason is that at > 80 °C the dehydration/rearrangement pathway of glyceraldehyde to lactic acid becomes viable [12,13]. Therefore, at high temperature and base concentrations, tartronic acid formation is inhibited by these side reactions.
These side reactions can be mediated through choice of catalyst support [19][20][21]. Recently, it has been demonstrated that supporting AuPt onto high surface area lanthanum based perovskite supports (LaBO 3, where B was Cr, Mn, Fe, Co or Ni) allowed for control of the glycerol oxidation pathway [14,22]. It was found that at operating conditions of 100 °C, 3 bar oxygen and a ratio 4:1 base to glycerol the dehydration/rearrangement pathway to lactic acid or oxidation pathway could be switched on or off by choice of perovskite B site element. The LaMnO 3 support, with its high reported oxygen adsorption capacity and activity for the oxygen reduction reaction, was found to inhibit the dehydration/ rearrangement pathway and favour oxidation [23]. Here we report a systematic study to design a tartronic acid synthesis catalyst based on variation of the Au:Pt ratio while employing a LaMnO 3 support.

Preparation of Perovskite Supports
Lanthanum manganite (LaMnO 3 ,) was prepared by supercritical CO 2 anti-solvent precipitation (SAS). This SAS process was chosen as it produces high surface area materials with low impurity content (such as alkali metals) and exceptional mixing of components [14,24]. A summary of the preparation method is given below, with a more detailed experimental method described elsewhere [24]. A solution of lanthanum (III) acetylacetonate hydrate (4 mg ml −1 / Sigma Aldrich) and manganese (II) acetate tetrahydrate (2.7 mg ml −1 /Sigma Aldrich > 99.9% purity) in methanol (reagent grade, Fischer Scientific) was prepared. The metal salt solution was pumped at 4 ml min −1 (using an Agilent HPLC) through a coaxial nozzle, along with carbon dioxide (BOC) at 12 kg h −1 , into a vessel held at 130 bar and 40 °C. The resulting precipitate was recovered on a stainless steel frit; the CO 2 -solvent mixture flowed down stream where the pressure was decreased to separate the solvent and CO 2 . After 120 min the precipitation process was completed, the system was purged with CO 2 for 30 min at 130 bar and 40 °C. The system was then depressurised, and the dry powder collected. The SAS precipitate was then calcined at 750 °C in static air (ramp rate of 2 °C min −1 ) for 4 h to form the perovskite.

Addition of Au/ Pt Nanoparticles to Perovskite Supports
Au, Pt and AuPt nanoparticles were deposited onto the perovskite supports using the sol-immobilisation procedure [14]. For the preparation of sols, aqueous solutions of HAuCl 4 (Johnson Matthey), PdCl 2 (Johnson Matthey) and PtCl 2 (Johnson Matthey) were prepared at the desired concentrations. Polyvinyl alcohol (PVA, 1 wt% aqueous solution, Aldrich, MW = 10 kDa) was freshly prepared and used as the stabilizer. NaBH4 (Sigma Aldrich, 0.1 M aqueous solution) was also freshly prepared and used as the reducing agent. To an aqueous mixture of HAuCl 4 and PtCl 2 of the desired concentration (1: 1 metal weight ratio, 1 wt% total metal in final catalyst) the PVA solution was added [PVA/ (Au + Pt) (wt/wt) = 0.65] with vigorous stirring for 2 min. NaBH 4 was then added rapidly such that the NaBH 4 : total metal ratio (mol/mol) was 7.5. The support perovskite was then added to the solution and after 1 h of stirring the mixture was filtered, washed with distilled water and dried at 120 °C for 16 h.

Catalyst Testing
Reactions were performed using a 50 mL Radleys glass reactor. The aqueous glycerol solution [0.3 M, containing NaOH (NaOH/glycerol ratio = 4, mol/mol)] was added into the reactor. The reactor was then heated to 80 °C prior to being purged three times with oxygen. Then the desired amount of catalyst (glycerol/metal ratio = 1000, mol/mol) was suspended in the solution and the reactor heated to 100 °C. The system was then pressurised to 3 bar O 2 and the reaction mixture stirred at 900 rpm. After the stated reaction time, the reactor vessel was cooled to room temperature and the reaction mixture diluted by a factor of 10 before being analysed by HPLC (Agilent 1260 infinity HPLC). The HPLC was equipped with ultraviolet and refractive index detectors and a Metacarb 67H column (held at 50 °C) to separate products. The eluent was an aqueous solution of H 3 PO 4 (0.01 M), used at a flow rate of 0.8 ml min −1 . Quantification of reactants consumed and products generated was determined by an external calibration method. The reaction effluent was analysed for the following products; glyceric acid, tartronic acid, oxilic acid, glycolic acid and lactic acid.

Results
As previously reported in detail and summarised in Table 1, the SAS prepared LaMnO 3 comprised of 100% orthorhombic LaMnO 3 according to XRD, without evidence of single metal oxides, and with a surface area 32 m 2 g −1 . The high purity and reasonable surface area allowed for the successful deposition of Au, AuPt or Pt onto the support. Representative TEM analysis (Figs. 1 and 2) showed that metal particle sizes of 2.1 nm (± 0.9 nm) where achieved on the LaMnO 3 support. In comparison, TiO 2 with a surface area of 61 m 2 g −1 also allowed for the successful deposition of monometallic Au or Pt and bimetallic AuPt nanoparticles with a representative particle size of 1.7 nm (± 0.8 nm).
The sequential oxidation of glycerol to glyceric acid and then onto tartronic acid requires relatively strong oxidising conditions, such as a high base to substrate ratio and elevated temperatures; in this study a NaOH:glycerol ratio of 4:1 was used with a reaction temperature of 100 °C. As previously mentioned, while this will push sequential oxidation reactions it will also facilitate dehydration and rearrangement reactions [13].
Initially, the effect of metal composition with a conventional TiO 2 support, on product distribution was investigated. As observed from time-on line data in Fig. 3 and Table 2, irrespective of the use of Au, Pt or AuPt nanoparticles, TiO 2 supported catalysts all had a high selectivity, of ca. 70%, towards lactic acid (the product of glyceraldehyde dehydration and rearrangement). The activity of the Pt catalyst was higher than Au with a TOF of 350 h −1 vs 220 h −1 respectively, however a significantly higher C-C scission was seen with Pt as the reaction progressed. These observations, regarding oxidation pathway, are comparable to previously reported studies at lower temperatures of 60 °C [19]. The AuPt/TiO 2 catalyst retained the activity of the Pt/TiO 2 with a TOF of 320 h −1 but with significantly lower C-C scission selectivity and higher glyceric acid selectivity. Despite the ability of Au to mediate C-C scission, as a monometallic catalyst or as part of a AuPt alloy, the dominance of the dehydration pathway resulted in poor tartronic acid yield (below 10% after 6 h time online) regardless of metal combination. Another interesting observation was that the rate of glycerol conversion significant dropped after 2 h time online with the Au/TiO 2 catalyst, with conversion increasing from 36 to 42% from 2 to 6 h (this 6% representing a 22% growth in conversion). By comparison conversion increased from 55 to 81% (47% growth) over the same time period with AuPt/TiO 2 . This result suggest that the Au/TiO 2 catalyst deactivates over the reaction time period, relative to Pt containing catalysts. Lui and co-workers noted the same phenomenon, with Au only catalysts deactivating and Au-Pt ones being stable, when these catalysts were studied for the purposeful production of lactic acid and attributed this to a change in metal-support interaction [15]. Taking these observations into account, the effect of Au:Pt ratio on catalytic performance was investigated with a SAS prepared LaMnO 3 support. Figure 4 shows time online analysis of these catalysts for the glycerol oxidation reaction, with TOFs and product breakdown compared to the TiO 2 catalysts given in Table 2. Firstly, a general improvement in initial TOF for all metal compositions can be observed for LaMnO 3 supported catalysts, relative to TiO 2 supported catalysts. Of particular note is the significant improvement in the TOF of the Au only catalyst between the two supports, with Au/LaMnO 3 have a TOF of 370 h −1 vs that of 220 h −1 for Au/TiO 3 . As seen with Au/TiO 3 , the rate of glycerol conversion using Au/LaMnO 3 notably decreased with time online, however to a less significant extent to Au/TiO 2 . Strong metal-support reactions have previously been reported for Au/LaMnO 3 catalysts during gas phase oxidations, which potentially explains the difference in stability for this catalyst vs Au/TiO 2 [25] .
The stability of LaMnO 3 catalysts to leaching, under these reactions has been previously reported [14,22]. Consultation of the relevant Pourbaix diagram (Fig. 4) shows that in the absence of the NaOH, LaMnO 3 is highly prone to dissolution from acid products [22]. However, the addition of NaOH maintains a pH of > 12 (even in the event of 100% conversion to formic acid) under which the support is stable, as evidenced by only 60 ppm La and 20 ppm Mn being observed over a 2 h catalyst cycle [14]. Elemental analysis has shown there to be no observable Au or Pt leaching during reaction.
In contrast to the TiO 2 supported catalysts, all LaMnO 3 catalysts (regardless of Au/Pt composition) had low yield (below 10%) towards lactic acid. This finding agreed with our previous observations of low lactic acid selectivity for lanthanum perovskite supports with high oxygen absorption capacity (LaMnO 3 and LaCoO 3 ) and high lactic acid selectivity with perovskites exhibiting low oxygen absorption capacity (LaCrO 3 , LaFeO 3 and LaNiO 3 ) [14]. The employment of only AuPt nanoparticles in the previous study raised the question as to whether the effect of the support was to modify the AuPt morphology, which then effected selectivity towards oxidation or dehydration products. Clearly, the Fig. 1 Representative Transmission electron microscopy of AuPt/LaTiO 2 (a) and associated particle size distribution (b) Fig. 2 Representative Transmission electron microscopy of AuPt/LaMnO 3 (a) and associated particle size distribution (b) low lactic acid selectivity for both the Au/LaMnO 3 and Pt/ LaMnO 3 catalysts, shown in Fig. 5, demonstrated that the support itself controls/inhibits dehydration and the lactic acid reaction pathway and not the nanoparticle composition.
The inhibition of the dehydration pathway is quite remarkable, given that it has been demonstrated that initial glycerol oxidation products of glyceraldehyde and 1,3,dihydroxyacetone have been shown to readily undergo dehydration and rearrangement to lactic acid in the presence of base alone (i.e. with an absence of a catalyst) [13]. Specifically, lactic acid selectivity is dictated by a reaction step that occurs in solution. If a carboxylic acid has been produced from the oxidation of these shared intermediates (with dehydration pathway), lactic acid cannot form (see scheme 1). Therefore, inhibition of the dehydration reaction by the LaMnO 3 support can be attributable to a significantly  reduced desorption of glycerol oxidation intermediates during their sequential oxidation. While LaMnO 3 supresses lactic acid, product selectivity associated with the oxidation reaction pathway is affected by the metal composition. As observed with TiO 2 supported catalysts, Au prevents C-C scission, with both Au/LaMnO 3 and AuPt/LaMnO 3 catalysts exhibiting less than 10% selectivity towards scission products; while the reaction with the Pt/LaMnO 3 catalyst had high C-C scission product selectivity that exceeded 50% over the reaction. The significantly higher C-C scission selectivity for the Pt/LaMnO 3 compared to Pt/TiO 2 (C-C scission after 6 h of 58% selectivity and 25% selectivity respectively) can be attributed to the prevalence of the dehydration reaction in the latter catalyst.
Another notable effect regarding metal composition is the extent of the sequential oxidation of glycerol to tartronic acid (via glyceric acid). Yields of tartroic acid with reaction time on-line are shown in Fig. 6, and final product yield after 6 h are given in Table 2. Due to the high C-C scission rate, very little tartronic acid was formed using the Pt/LaMnO 3 catalyst during the reaction. The AuPt/LaMnO 3 catalyst resulted in high glyceric acid selectivity over the 6 h time on line, with a slight increase in tartronic acid formation towards the end of the reaction to give a final yield of 13.5%. The secondary oxidation of glyceric acid to tartronic acid appears to be suppressed while significant glycerol substrate remains present in the reaction. However, low C-C scission and lactic acid formation over the 6 h reaction shows that tartronic acid formation could be achieved by extending the reaction time, as demonstrated by the complete conversion of all glycerol and glyceric acid to give 87% tartronic acid yield over 24 h reaction time. Interestingly, the Au/LaMnO 3 catalyst has high activity for the conversion of glyceric acid into tartronic acid, with a yield of 44% at 6 h reaction time, while glycerol was still present in the reaction mixture. After 24 h this catalyst gave an exceptional 92% selectivity towards tartronic acid, yet even after this time-period, glycerol conversion was only 87%, giving a final tartronic acid yield of 80%.
As noted, Au only catalysts showed a marked decrease in glycerol conversion with time (vide supra), yet it seems that further oxidation of glyceric acid occurs after glycerol oxidation has stopped or markedly slowed. Conformation of this can seen from simple plots of reactant and product concentrations with time on-line (Fig. 7). A reaction with AuPt/ LaMnO 3 shows a clear increase in glycerol acid concentration as that of glycerol drops which, as previously reported, reaches a maximum concentration once all glycerol has been consumed, before itself being consumed to produce tartronic acid. In the case of reactions using Pt/LaMnO 3 , glyceric acid concentration remains at steady state from 60 to 360 min, while glycerol continues to be consumed to produce C-C scission products and small quantities of tartronic acid. Most interestingly, glyceric acid concentration peaks within the 1st h of reaction when using the Au/LaMnO 3 catalyst and then steadily decreases, even while the rate of glycerol consumption slows dramatically. Tartronic acid is formed at a steady rate throughout the 6 h reaction period, suggesting that oxidation of glyceric acid to tartronic acid is preferred over glycerol oxidation. As glycerol conversion almost stops at 6 h but tartronic acid formation and glyceric acid consumption continue, there must be different reaction pathways and/or different catalytically active sites for these two reactions on the Au catalysed surface.
Freakley, Hutchings and co-workers showed that selectivity towards glyceric or tartronic acid could be influenced by the amount of polymeric stabiliser used in Au nano particle synthesis. High tartronic acid selectivity being observed when little or no stabiliser was used, while high glyceric acid yields were reported when higher concentrations of stabiliser were used [26]. A potential cause of the switch between a glycerol oxidation pathway and a glyceric acid oxidation pathway seen in Au/LaMnO 3 could be that the surface coverage of the PVA stabilising polymer decreased during the reaction, which is expected under aqueous reaction media at elevated temperatures. Alternatively, specific active sites for glycerol oxidation were blocked by product inhibition during the reaction.

Conclusion
Using reaction conditions to favour sequential oxidation of glycerol to produce tartronic acid, i.e. temperatures of 100 °C and a 4:1 NaOH: glycerol ratio, has limited effect when using conventional TiO 2 supported Au, AuPt or Pt nanoparticle catalysts. The reason for this being the preference for the dehydration and structural rearrangement of initial aldehyde products to form lactic acid. This lactic acid reaction pathway is switched off when a LaMnO 3 support is used, irrespective of the choice of metal nanoparticle, i.e. monometallic Au or Pt, or bimetallic catalysts produce none or minimal lactic acid over 6 h reaction. It is therefore clear that the support structure is responsible for shutting down the dehydration pathway. Choice of nanoparticle composition did however influence reaction selectivity, with Pt alone facilitating C-C scission and alloying with Au supressing these reactions to favour glyceric acid production. Au only catalysts demonstrated interesting behaviour under the reaction conditions used; firstly, Au only catalysts had comparable turn over frequencies to AuPt alloyed catalysts. Secondly, while the catalysis of glycerol oxidation underwent rapid  deactivation, the sequential oxidation of glyceric acid to tartronic acid was relatively unaffected. Such an observation was only possible due to the lactic acid pathway being suppressed by the LaMnO 3 support.