Gold catalyst recycling study in base-free glucose oxidation reaction

1 Gold catalyst recycling study in base-free glucose oxidation reaction C. Megías-Sayago*, L. Bobadilla, S. Ivanova, A. Penkova, M. A. Centeno, J. A. Odriozola Departamento de Química Inorgánica, Universidad de Sevilla e Instituto de Ciencia de Materiales de Sevilla, US-CSIC, Américo Vespucio 49, 41092, Sevilla, Spain corresponding autor: cristina.meg[email protected] Abstract This work is devoted to the study of viability of immobilized gold colloids on carbon as catalysts for the base-free glucose oxidation reaction with a special emphasis made on catalysts’ recycling, operational life and possible routes for deactivation/reactivation under batch conditions. The observed catalytic behavior is related to all possible manners of deactivation, like gold metal state changes (particle size agglomeration or leaching), support modifications or active sites blocking by intermediates. In an attempt to recover the initial catalytic activity, the samples are subjected to different treatments such as H2O and NaOH washings and calcination. The failure of the regeneration procedures to recover the initial activity and after detailed catalyst’ characterization allows us to find out the main cause of deactivation. Keywords: base-free glucose oxidation, gold, activated carbon, deactivation, recycling 2 Introduction In the last decades, an increasing necessity of readdressing the interest from fossil raw materials to renewable feedstocks arises due to the continuous depletion of the formers. A reorientation of the scientific interest towards carbohydrates as feedstock for fine chemistry is currently of actuality [1]. The “green catalysis” approach including orientation toward aqueous media low temperature catalytic transformations and carbohydrates highly functionalized type molecules increases by far its potential as fine chemical feedstock [2, 3]. It is then of crucial importance to explore this potential toward efficient, environmentally friendly and economically viable technologies for large-scale conversion of carbohydrates into industrially viable bulk or intermediate compounds for chemical, pharmaceutical or polymeric use [4]. Among the existing reactions for carbohydrates conversion, the selective oxidation of plant-biomass monomers (hexoses or pentoses) to corresponding carboxylic acids presents a possibility to easy integration in already existing technologies and therefore an important economic interest. The heterogeneously catalyzed transformation of glucose to gluconic acid received important attention [5-8]. Nevertheless, the proposed processes must show an important activity/selectivity balance and especially long term stability in order to be competitive to the existing biotechnological process for gluconic acid production. The slow to severe catalyst deactivation and side reactions products are considered as the main challenges to undertake for future industrialization of heterogeneous process [9]. Among the metal catalysts discussed for selective oxidation of glucose to gluconic acid, the noble metal based ones are the most frequently reported. Important productivity was reported for Pt and Pd based catalysts [10-12]. Especially improved activities and product selectivity of the Pt based catalysts were obtained after doping with Bi [12,13]. However, during the reaction, bismuth leaching was observed which converts this 3 catalytic system inadequate for pure chemicals production and may cause deactivation. In addition, alkaline conditions are required to increase the reaction rate and to avoid deactivation by reaction intermediates. Although beneficial for the reaction rate, the use of base conditions affects negatively the economics of the process as gluconate salt instead of pure gluconic acid is obtained at the end of reaction. Often, a decrease of selectivity due to the favored glucose to fructose isomerization reaction at high pH is observed [6, 14]. Considering that the catalysts operate continuously under oxygen rich atmospheres Pt and Pd based catalyst could suffer also deactivation by “over-oxidation” responsible for active sites lost by oxidation. The majority of these inconvenients could be resolved by using gold based catalysts instead of platinum and palladium based ones. High selectivity and activity were reported for supported gold catalysts and convert them in the most promising candidates for future industrial implantation [8, 15-19]. Gold catalysts successfully operate under oxygen. The inert nature of the metal avoids loss of activity by oxidation and present an important activity for aldehyde group oxidation and inertness towards secondary alcoholic group oxidation [6]. Probably the most important findings concerning gold based systems is that they could operate under a wide range of pH with the same activity, which allows avoiding the glucose to fructose isomerization under base conditions. Recently, the great potential of gold catalysts for direct glucose oxidation in base free conditions was demonstrated [20-23]. It was confirmed that catalyst productivity is barely influenced by the pH of the reaction and that the proper choice of support is very important for catalyst performance. Mineral oxides, as pure ceria or mixed cerium oxides, are not suitable as supports, since the catalysts suffer slow deactivation by metal leaching [21, 23]. On the other hand, basic supports are reported to be most suitable in acidic conditions [20], although their stability towards hydroxylation has to be improved. 4 Pointing industrial application, the catalysts should withstand continuous operation and repeated recycling for a long time, which implies low deactivation rate and/or easy activity recovery. Carbon based catalysts present excellent stability in acidic media and low active metal leaching which converts them in potential candidates to study their operation under base free glucose oxidation reaction. In this context, the aim of this work is to study the catalytic performance of a series of Au/C catalysts prepared by gold colloids immobilization and their recyclability under repeated operation. All possible deactivation causes as gold leaching, sintering and chemical poisoning (active sites blocking by reaction intermediates) are contemplated and related to the variation of the initial gold particle size. The viability of some catalyst’ treatments between the operation cycles aiming to recover the initial deactivation was also investigated. Experimental Catalyst preparation Gold was deposited from pre-formed colloids prepared by reducing of the PVA stabilized gold precursor with NaBH4. 5.10-4 M aqueous solution of HAuCl4 (2wt% nominal value) was mixed with the corresponding quantity of PVA (1 wt. % aqueous solution) and stirred for 20 min. After, appropriate amount of 0.1 M freshly prepared NaBH4 solution was quickly added for other 20 min and the solution was put in contact with commercially available activated carbon powder DARCO® (Sigma Aldrich, 100 mesh particle size). After aging, the final mixture was centrifuged at 15000 rpm for 20 minutes and resulted solids separated by filtration. Samples were finally calcined in static air at 300 ºC for 2 hours. 5 Following this procedure, four samples were prepared keeping constant PVA:Au weight ratio of 0.85 and varying the NaBH4 :Au molar ratio from 3 to 10. The later was chosen in order to get different final gold particle size, as reported in previous studies [24]. Table 1 summarizes the synthesis parameters and labelling of the catalysts. Table 1. Synthesis parameters of the prepared samples. Sample NaBH4 : Au molar ratio AuC_I 3 AuC_II 5 AuC_III 10 Characterization The gold contents were estimated through ICP analysis by using Horiba Jobin Yvon spectrometer. Transmission electron microscopy (TEM) study on particle size and dispersion of the catalysts was performed on PHILIPS CM-200. The average gold particle size was estimated based on surface distribution calculations as shown in equation 1 D [3,2]= ∑D𝑖 3𝑣𝑖 n 1 ∑D𝑖 2𝑣𝑖 n 1 eq. 1 where Di is the geometric diameter of the ith particle, and vi the number of particles with this diameter. For particle size distribution, the total number of measured particles overcomes 200 for every sample. The DRIFTS spectra were recorded at room temperature without sample dilution using a Thermo Nicolet Nexus FT-IR spectrometer equipped with a liquid nitrogen cooled MCT 6 detector at 4 cm-1 resolution and average of 128 scans. The whole optical path was purged with CO2- and H2O-free nitrogen. About 50 mg of fresh or used catalyst finely grounded was loaded in the Praying MantisTM cell for each measurement. XPS measurements were carried out on Leybold-Hereus LHS-1020 instrument coupled with EA200 detector and using non chromatic Mg K (220W, 11kV, 1253,6 eV). Prior to use the sample were pressed into a thin disk. The XPS spectra of all sample were recorded at room temperature and the binding energy was calibrated on C1s at 284,6 eV with an uncertainty ±0.2 eV. The spectra were recorded with constant pass energy of 44 eV and 0.1 eV resolution for the studied zones. Base free aerobic oxidation of glucose The oxidation of glucose was performed in a glass batch reactor (50 mL) at constant temperature and stirring rate at saturated oxygen atmosphere (approximate PO2 of 0.1 MPa). In a typical experiment, the reactor was charged with 0.2M glucose aqueous solution, catalyst in Glucose/Au molar ratio of 100 and oxygen saturated by bubbling 20 mL/min of pure oxygen flow for few minutes. The reactor was then closed and reaction mixture stirred at 600 rpm, at 40ºC during 18 hours in base-free conditions. The recycling study was carried out in a similar manner separating the catalyst from the products by filtration between the runs and re-using it maintaining Glucose/Au molar ratio of 100 at every run. In some cases reactivation treatments were performed. The samples after filtration were treated either with distilled water, 0.1 M NaOH solution or thermally treated at 300ºC for 2h before every run. 7 The reaction products were identified and quantified by HPLC equipped with refractive index detector (Varian 360-LC) and Hi-Plex H column (300 × 7,7 mm) set to 40ºC using MilliQ water as mobile phase. Glucose conversion and product selectivity to gluconic acid (GA) were calculated as described by equation (1) and (2). 𝐶𝑜𝑛𝑣𝑒𝑟𝑠𝑖𝑜𝑛 (%)= [𝐺𝑙𝑢𝑐𝑜𝑠𝑒]𝑖𝑛−[𝐺𝑙𝑢𝑐𝑜𝑠𝑒]𝑜𝑢𝑡 [𝐺𝑙𝑢𝑐𝑜𝑠𝑒]𝑖𝑛 ∗100 eq (1) 𝑆𝑒𝑙𝑒𝑐𝑡𝑖𝑣𝑖𝑡𝑦 𝐺𝐴 (%)= 𝑚𝑜𝑙𝑒 𝐺𝐴 𝑡𝑜𝑡𝑎𝑙 𝑚𝑜𝑙𝑒𝑠 𝑜𝑓 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑠*100 eq (2) Results and discussion The real gold loading measured by ICP and the mean particle size deduced from TEM is presented in Table 2. Table 2. Gold metal loadings and particle size for fresh and spent catalysts. Sample Au wt.% (fresh), Au, wt.% (spent 4th cycle) Au particle size, nm (fresh) Au particle size, nm (spent 1st cycle) Au particle size, nm (spent 4th cycle) AuC_I 2.2 1.9 16.1 n.m. 20.4 AuC_II 2.4 1.9 8 17 16.6 AuC_III 2.3 1.9 4.8 8.9 7.7 n.m. not measured All fresh samples present similar gold loadings close to the targeted value. The average gold particle size varies within the series attributed to the modification of the synthesis parameters, chosen in a way to have a real disparity of the initial particle size. Higher the NaBH4:Au ratio at a constant PVA:Au ratio, lower the average gold particles size. The 8 synthesis parameters influence strongly the average particle size but not the actual metal loadings. The representative micrographs of AuC_II sample and its particle size distribution are presented in Figure 1. Figure 1. Representative TEM micrographs and gold size distributions for fresh and spent (4th cycle) AuC_II samples. Catalytic activity and recycling study The catalyst activity of all samples and in every cycle, expressed as glucose conversion, is shown in Figure 2a. The calculated carbon balance ranges from 86 to 95 % (first vs. last cycle) indicating possible adsorption of reactive or intermediates after the first cycle but not after all successive ones. The gluconic acid selectivity on the base of liquid nonadsorbed products was found 100% and will not be discussed from now on. No matter the gold particles size similar glucose conversions are obtained in the first cycles neighboring 80%. Systematically the second cycle gives always less conversion than the first one being the difference in the range of 20-25 %. The performed blank test with activated carbon shows that at this temperature only 8% of glucose could be adsorbed on the carbon surface and false the initially calculated glucose conversion. Anyway the deactivation of the catalyst between the 1st and the 2nd cycle occurs 9 unequivocally. It is also observed very similar 2nd and 3rd cycles and slightly different 4th cycle. The sample AuC_III shows continuous decrease of glucose conversion, while the AuC_I and AuC_II stabilized after the initial loss of activity. Nevertheless, from statistical point of view, within the series of samples only AuC_II behave differently (Figure 2a). Figure 2. a) Glucose conversion on fresh and spent catalysts under 4 reaction cycles b) TOF (bars) and particle size (points) comparison in 1st and 4th cycles on fresh and spent catalysts. As reported above the reasons for deactivation could be assembled in two groups, those altering directly the active sites, as sintering or metal loss, and those responsible for catalyst’ blocking or support modifications. The leaching of gold was checked by ICP analysis of the remaining gold on catalyst after accomplished recycling study (Table 2) and in the filtration water before every cycle. No matter the starting catalyst and/or particle size, the same gold loss is detected averaging 15% of the initial gold loading and always the gold loss between the 1st and the 2nd cycle represents almost 80% of the total leaching. 16 indicating that nor the adsorbed intermediates nor the modification of the support play important role in catalyst deactivation. The similar 2nd, 3rd and 4th cycles demonstrate also that the presence of adsorbed intermediates or modifications of the support does not inhibit the reaction. It seems that gold metal state, i.e. gold particle size variation and metal leaching, is the most important parameter inducing loss of activity. On one hand the particles agglomeration could result in a beneficial optimal particle size and increase in the specific activity (AuC_II). And on the other, in our previous study [23] it was reported that the leached gold could participate in the reaction with around 10% conversion increase between the 4th and the 18th hour of reaction which together with the 8% of glucose adsorption on activated carbon accounts for the 20% of glucose conversion decrease between the 1st and the 2nd cycle. All this leads to the conclusion that the most important problem to solve is gold metal leaching. This statement leads to the assumption that the carbon support is not a key parameter in this reaction and that the glucose oxidation is occurring only on the low defect surface gold particles with preferable average size of 15-20 nm. Ishida et al. [8] also suggested that the glucose oxidation reaction occurs preferably on gold particles surface and not on the gold/support interface or support. Our observations also support this idea, as the most important deactivation is produced by the gold state change. The mechanism of glucose oxidation in alkaline conditions is already known and includes preferable glucose adsorption on hydroxylated Au surface on which the glucose CHO group is firstly activated. After water is released by formyl group H removal and gluconic acid is formed by OH transfer from the Au surface [35]. However, in base free conditions the glucose oxidation should proceed through carbonyl conjugated radical [36] and probably will include oxidation through hydrogen peroxide formation, as the activation of molecular oxygen is improbable on gold. Comotti et al. [37] proposed as a key reaction 17 step the presence of electron-rich gold species, formed by hydrated glucose anions on gold surface atoms, which activate the molecular oxygen by nucleophilic attack and lead to the formation of hydroxyperoxides, represented schematically in Figure 6. Figure 6. Representation of possible mechanism of glucose oxidation in base free conditions. This step could be considered as one of the of Baeyer –Villiger reaction steps, including addition of peroxide to formyl group with hydroxyperoxide formation and its rearranging with cleavage of the C-C bond and products formation [38]. The proposed reaction steps are only a tentative for predicting the reaction mechanism in base free conditions. However a detailed study is needed to confirm the viability of the proposed reaction pathway. Conclusions Several sample with varying gold particle size are prepared and tested in base free oxidation of glucose. The samples show similar behavior around 80% of glucose conversion toward 100% selective gluconic acid production in the first cycle of reaction. The recycling of the samples reveals that deactivation occurs after the 1st cycle independently on the initial gold particle size. Gold leaching is confirmed to be the 18 primary reasons for catalyst deactivation. Also a gold agglomeration occurs and results beneficial for the specific activity improvement when optimal gold average size is achieved. On the other hand, the support nature, its chemical state or adsorbed species do not seem to be very important. All reactivation treatments result in the same activity trend indicating that the catalyst operation is not affected by intermediates adsorption, removal or support modification. Therefore, every loss of activity is due to gold metal leaching after the 1st cycle and when the leaching stops the activity remains constant (2nd, 3rd, 4th cycle). Acknowledgements Financial support for this work has been obtained from the Spanish Ministerio de Economía y Competitividad (MINECO) (ENE2013-47880-C3-2-R) co-financed by FEDER funds from the European Union. 19 References [1]. C. Chatterjee, F. Pong, A. Sen, Green Chem., 2015, 17, 40-71. [2]. J. C. Serrano-Ruiz, R. Luque, A. Sepúlveda-Escribano, Chem. Soc. Rev., 2011, 40, 5266-5281 [3] J. Chedda, G. W. Huber, J. A. Dumesic, Angew. Chem. Int. Ed., 2007, 46, 7164- 7183. [4] Frieder W. Lichtenthaler, Siegfried Peters, C. R. Chimie 7 (2004) 65–90. [5] M. Besson, P. Callezot, Catalysis Today, 57, 2000, 127-144 [6] S. Biella, L. Prati, M. Rossi, J. Catal, 206, 2002, 242-247. [7] P. Beltrame, M. Comotti, C. Della Pina, M. Rossi, Appl. Catal. A Gen. 297 (2006) 1–7. [8] T. Ishida, N. Kinoshita, H. Okatsu, T. Akita, T. Takei, M. Haruta, Angew. Chemie Int. Ed. 47 (2008) 9265–9268. [9] M. J. Climent, A. Corma, S. Iborra, Green Chem., 13, 2011, 520-540. [10] D. J. Verraest, J. A. Peters, H. van Bekkum, Carbohydr. Res. 306 (1998) 197-203 [11] J.M.H. Dirkx, H.S. van Der Baan, J. Catal. 67 (1981) 1–13. [12] M. Besson, F. Lahmer, P. Gallezot, P. Fuertes, G. Fléche, J. Catal. 152 (1995) 116– 121. [13] P. Gallezot, Catal. Today. 37 (1997) 405–418. 20 [14] C. Kooyman, K. Vellenga, H.G.J.D.E. Wilt, Carbohydrate Research, 54 (1977) 33- 44. [15] C. Baatz, U. Prüβe, J. Catal. 249 (2007) 34-40. [16] N. Thielecke, K. D. Vorlop, U. Prüβe, Catal. Today 122 (2007) 266-269. [17] U. Prüβe, M. Herrmann, C. Baatz, N. Decker, Appl. Catal. A, 406 (2011) 89-93. [18] A. Mirescu, H. Berndt, A. Martin, U. Prüβe, Appl. Catal. A, 317 (2007) 204-209. [19] C. Baatz, N. Thielecke, U. Prüβe, Appl. Catal. B, 70 (2007) 653-660. [20] P. J. Miedziak, H. Alshammari, S. A. Kondrat, T. J. Clarke, T. E. Davies, M. Morad, D. J. Morgan, D. J. Willock, D. W. Knight, S. H. Taylora, G. J. Hutchings, Green Chem., 2014, 16, 3132-3141. [21] Y. Wang, S. Van de Vyver, K. K. Sharma, Y. Román-Leshkov, Green Chem., 2014, 16, 719-726. [22] P. Qi, S. Chen, J. Chen, J. Zheng, X. Zheng, Y. Yuan, ACS Catal. 2015, 5, 2659−2670. [23] C. Megías-Sayago, S. Ivanova, C. López-Cartes, M.A. Centeno, J.A. Odriozola, Catalysis Today 279 (2017) 148–154. [24] J. Luoa, W. Chu, S. Sall, C. Petit, Col. Surf. A: Physicochem. Eng. Aspects 425, 2013, 83–91. [25] H. Okatsu, N. Kinoshita, T. Akita, T. Ishida, M. Haruta, Appl. Catal. A: Gen. 369 (2009) 8–14. 21 [26] L. Prati, A.Villa, A. Lupini, G. Veith, Phys. Chem. Chem. Phys.2012, 14, 2969- 2978. [27] B. Zope, R. Davis, Green Chem. 2011, 13, 3484-3491. [28] B.J. Meldrum, C.H. Rochester, Fuel 70 (1991) 57-63. [29] Brian J. Meldrum and Colin H. Rochester, J Chem Soc Faraday Trans 86 (1990) 861-865. [30] Sócrates, Infrarred and Raman characteristic group frequencies, 3rd edition, Wiley [31]. J.L Figueiredo, M.F.R Pereira, M.M.A Freitas, J.J.M Órfão, Carbon 37 (1999) 1379-1389. [32] Brian J. Meldrum and Colin H. Rochester, J Chem Soc Faraday Trans 86 (1990) 1881-1884. [33] Abdel-Nasser A. El-Hendawy, J Anal Appl Pyrolisis 75 (2006) 159-166. [34]. Douglas B. Mawhinney, John T. Yates Jr., Carbon 39 (2001) 1167-1173. [35] T. Ishimoto, Y. Hamatake, H. Kazuno, T. Kishida, M. Koyama, Appl. Surface Sci. 324, 2015, 76-81 [36] S: J. H. F. Arts, E. J. M. Mombang, H. van Bekkum, R. Sheldon, Synthesis 6 (1997) 597-613 [37] M.Comotti, C. Della Pina, E. Falletta, M. Rossi, Adv. Synth. Catal. 2006, 348, 313 – 316. [38] Hassall, C. H. 2011. The Baeyer-Villiger Oxidation of Aldehydes and Ketones. Organic Reactions. 9:3:73–106.