Competitive Adsorption of Substrate and Solvent in Sn‐Beta Zeolite During Sugar Isomerization

Abstract The isomerization of 1,3‐dihydroxyactone and d‐glucose over Sn‐Beta zeolite was investigated by in situ 13C NMR spectroscopy. The conversion rate at room temperature is higher when the zeolite is dehydrated before exposure to the aqueous sugar solution. Mass transfer limitations in the zeolite micropores were excluded by comparing Sn‐Beta samples with different crystal sizes. Periodic density functional theory (DFT) calculations show that sugar and water molecules compete for adsorption on the active framework Sn centers. Careful solvent selection may thus increase the rate of sugar isomerization. Consistent with this prediction, batch catalytic experiments show that the use of a co‐solvent, such as tetrahydrofuran, that strongly interacts with the Sn centers suppresses glucose isomerization. On the other hand, the use of ethanol as cosolvent results in significantly higher isomerization activity in comparison with pure water because of decreased competition with glucose adsorption on zeolitic Sn sites.

Glucose is the cheapest of all hexose sugars and can, in contrast to fructose and mannose, be obtained in relatively pure form from lignocellulosic biomass.I somerization of glucose into fructose is important fort wo reasons:1 )catalyzed by glucose isomerase, this reversible reaction is used at the industrial scale to produce high-fructose corn syrup;2 )ina ddition, the aldose-ketose isomerization is at the centero f the attention of the scientificc ommunity in the context of biomass valorization into fuels and chemicals, as fructose can be readily dehydrated in high yield to 5-hydroxymethylfurfural, ap rospective platform molecule. [1][2][3][4] For implementation of glucose isomerization in biorefineries, enzymatic isomerization needs to be replacedb ym ore active and robustc atalyst systems. Glucose isomerization by bases is known, but suffers from by-product formation and waste issues. [5] Recently,L ewis acids have been identified as the preferred chemocatalysts for glucose isomerization. [6][7][8][9] From these studies, Sn-Beta zeolite has emerged as particularly effective for catalyzing carbohydrate transformations. Sn-Beta can isomerize hexoses, pentoses, and triosesb yi ntramolecular hydride and carbon shift reactions in variouss olvents. [8,[10][11][12] Another platform molecule is lactica cid, which can be obtained from fructosevia retro-aldol condensation [13,14] and hydride shift reactions involving 1,3-dihydroxyacetone and glyceraldehyde as intermediates (Scheme1). Earlier,S n-Beta has shown its promise in catalyzing Meerwein-Ponndorf-Verley reactions [15] between alcohols and ketones as well as Baeyer-Villiger oxidation of cyclic ketones. [16] It is well known that the Lewis acidic Sn sites embedded in the microporous framework of Beta zeolite play ak ey role in these applicationsa nd their function resembles to some extentt he action of enzymes. Still, the activity of Sn-Beta in sugar isomerization is low compared with enzymes.A lthough it is wellk nown that slow diffusion in zeolitic micropores can limit the reactionr ate, [17][18][19][20][21][22][23][24] such effects have not been reported for Sn-Beta-catalyzed isomerization of sugars. [25] Scheme1.Sn-Beta-catalyzed isomerization of triose (1,3-dihydroxyacetone) and hexose (glucose) in the framework of biomass valorization.
Herein, we report that Sn-Beta is already effective at room temperature as ac atalystf or the isomerizationo f1 ,3-dihydroxyacetone( DHA)a nd d-glucose. We followed the isomerization of C3 and C6 sugars in Sn-Beta by in situ 13 CNMR spectroscopy and noted pronounceda ctivity differencesb etween dehydrated and hydrated zeolites exposed to an aqueous sugar solution. We confirm that the catalytic conversion of sugars in water is slow owing to sluggish displacemento fs olvent molecules coordinating to active Sn centers by sugar molecules. This insight allowed optimizing the solventsystem towards improvedi somerization activity.T he resultsa ttest to the importance of competitive adsorption of sugars and polar solvent molecules in sugar isomerizationb ySn-Betazeolite.
Preliminary experiments employing ahydrothermally synthesized Sn-Beta zeolite in HF medium( Sn-Beta-HF,S i/Sn = 108) and an anosized post-synthetically modified catalyst( Sn-Betaps, Si/Sn = 80) confirm that there is no effect of the crystallite size on the catalytic performance ( Figure 1). From this, we conclude that sugar isomerization under these relatively mild conditions is not limited by internal mass transport limitations, as was shown before for the isomerization of arabinose. [26] Having excluded internal mass transport limitations, we focused on the intrinsic reactivity of Sn-Beta-HF by studying the transformation of carbohydrate substrates using in situ NMR spectroscopy in detail. This was done by followingt he chemical nature of the carbohydrate species adsorbed in the Sn-Beta-HF zeolite catalyst in time. We first investigated room temperature sugar isomerization overS n-Beta-HF dehydrated at 170 8Ci nvacuo for 3h.T he dehydrated Sn-Beta-HF sample was then impregnatedw ith ac oncentrated solution of DHA in D 2 O( DHA/Sn = 2), followed by brief evacuation to remove bulk D 2 O. The role of the interaction of the sugarsw ith Lewis acid sites in Sn-Beta during sugar uptake has already been demonstrated. [27,28] Relevant 13 CNMR spectra ( Figure S1 in the Supporting Information) prove that DHA was convertedi nto lactic acid (LA), as evident from the decrease of the DHA dimer signal (d = 65 ppm) and the appearance of signals relatedt oL Aa t d = 20 ppm for ÀCH 3 ,a td = 67 ppm for CÀOH and at d = 179 ppm C=O. We confirmed that DHA isomerization did not occur overA l-Beta and dealuminated Beta under the given conditions. In an experiment where the amount of substrate was increasedt oD HA/Sn = 10, we found that the reaction went to completioni na bout 10 h( Figure 2). In this case, we observed an additional signal at d = 205 ppm owing to pyruvaldehyde, ak nowni ntermediate in the reaction from DHA to LA. [29] We also evaluated the performance of Sn-MCM-41 at room temperature ( Figure S2) and confirmed that its intrinsic reaction rate is significantly lower than that of Sn-Beta-HF in line with recently reported results. [30] The sharper lines with Sn-MCM-41 as the catalysta re ar esult of the high mobility of the substrate molecules inside the large MCM-41 mesopores. Owingt ot he severe overlap of signals from the substrate and products,o nly aq ualitative assessment of the DHA conversion can be given on the basis of these NMRr esults.
As isomerization of sugarsi st ypicallyc arriedo ut in water, we investigated the influence of pre-adsorbed water on the  isomerization reaction, For this purpose, we followed the conversion of DHA by 13 CNMR spectroscopy of the same Sn-Beta-HF sample after hydration in saturated water vapor (50 8C, 12.3 kPa). Thermogravimetric analysiss howedt hat this hydration procedure led to aw ater content of Sn-Beta-HF of 2.4 wt %. When this zeolite was then impregnatedw ith the same DHA/D 2 Os olution (DHA/Sn = 10) as used before for the dehydrated zeolite, we observed that the conversiono fD HA proceeded much slower.A fter approximately3 0h,t he conversion was not complete. This experiment shows that the presence of water molecules in the micropores of Sn-Beta-HF significantly slows the reactionofD HA under theseconditions.
Similarly,w ei nvestigatedt he conversion of d-glucose in Sn-Beta-HF.I nt his case, we used direct excitation 13 Cm agnetic angle spinning (MAS)NMRa llowing for quantitative monitoring of sugar conversion. Introducing a 13 Cl abel in the C1 position of d-glucose helped to quantify reactant and product. First, the reactivity of dehydrated Sn-Beta-HF was assessed by roomtemperature impregnation of a d-glucose/D 2 Om ixture (d-glucose/Sn = 5). d-fructose was one of the reaction products, as follows from the appearance of new signals in the 13 CNMR spectra at d = 62 and6 4ppm, which respectively relate to b-furanose and b-pyranosef orms of d-fructose. [31] At the same time, the signals at d = 96 and 92 ppm, related to the a-a nd b-anomers of d-glucose, (Figure 3a)d ecreased. Based on the 13 Ci ntensities of the reactant and products,w e estimate that about 50 %o ft he starting d-glucose was converted in 12 h, mostly into d-fructose (33 %). This approaches the glucose/fructose equilibrium ratio of 1.13 at room temperature. [8,32] The remainder of the 13 Cl abels were found in the carbonyl region (170-180ppm), which shouldb er elated to secondary reaction products such as 5-hydroxymethylfurfural. [33] The impregnation of hydrated Sn-Beta-HF with the same solution resulted in as ignificantly lower reaction( Figure3b), in keeping with the differenceo bserved for DHA isomerization.
In these experiments, we observed that the concentration of the a-anomer of d-glucose was highert han that of the banomer.T his could be attributed to ac onfinement effect. The increased concentration of one of the anomers in the zeolite micropores should not affect the catalytic performance, as the barrierf or mutarotation on Sn sites ism uch lower than the barrierfor the H-shift reaction. [34,35] Figure 3c summarizes the sugar yields calculated from the NMR. One can see that for the hydrated Sn-Beta-HF,t he reaction proceeds very slowly after 4h and the conversion stays below 20 %e ven at prolonged reaction times. On contrary,d e-hydratedS n-Beta-HF afforded ag lucosec onversion of about 50 %a fter 12 h. The data show that cyclic d-fructosew as formed much faster than previously reported in as imilarN MR experiment at room temperature. [33] Apart from signals correspondingt od-fructose, the 13 CNMR spectra also contained signals in the carbonyl region (d = 160-200 ppm). One of the signals (d = 178.6 ppm) is related to 5-hydroxymethylfurfural. [33] The other signals could not be identified. They are not resulting from the acyclic aldohexoses, because thesec ompoundsw ould give rise to as ignal at d = 205 ppm. [36] After 22 ho fr oom temperature reaction, the rotor andi ts contentsw ere placed in an oven at 100 8Cf or 2h. 13 CNMR spectra recorded after cooling to room temperature evidenced that glucose and fructose were completely converted as followed by the absence of signals in the 60-100 ppm range (Figure S3). This is surprising, as in batch experiments the isomerization reactiong oes to equilibrium (glucose/fructose % 0.75 at 100 8C) [8,32] with comparatively little formation of byproducts. [8] Instead, the spectrum of the heated sample contained signals owing to carbonyl groups (d % 180 ppm) and CH 3 groups (d % 20 ppm). These findings show that the sugarsw ere also converted to lactic acid. We postulate ar eactionm echanism (Scheme S1) that helps to explain the occurrenceo f 13 Cl abels at ÀCH 3 and carbonyl carbon atoms following reactions of 13 C1-d-glucose. This mechanism is consistentw ith recenti nsights from Lewis acid-catalyzed retro-aldolization of hexoses. [13,37] In the light of the presented data, we hypothesize that the sugar reactant and water solvent molecules competef or coordination to Sn sites. In this way,d ifferencesi nt he coverage of the active sites with sugar would affect the catalytic activity. Previously as imilarh ypothesis has been put forward to rationalize the crucial role of hydrophobic reactione nvironment on glucosei somerization reaction. [38] This taken together with our findings inspired us to furthere xplore whether competitive adsorptione ffects play ar ole during sugar isomerization by comparing catalytic performance of Sn-Beta-HF in water,i ne thanol/water = 9/1 v/v and in THF/water = 4/1 v/v mixtures (THF = tetrahydrofuran). The solvent mixtures were selected to ensure sufficient variation in the adsorptions trength with Sn sites as comparedw ith water,y et to still be able to dissolve the sugar reactant. Figure 4s ummarizest he resultso ft hese activity measurements at 90 8C. In the THF/water mixture the isomerization activity is much lower than the reference case in water. On the other hand, the reaction rate was much higher when the reaction was carriedo ut in the mixed ethanol/water solvent.
Uptake measurements of the different solvents on Sn-Beta-HF ( Figure S6) show that the adsorption of THF and ethanol on the zeolite is much stronger than that of water at room temperature and also at 90 8C. This indicates that the interaction differences with zeoliteB eta itself cannot explain the activity differences, but should rather be relatedt ot he differences in specific interactions with the active Sn-sites, that is, competitive adsorption. Further support for the hypothesis that competitive adsorption betweent he solvent and sugarsa ffects catalytic performancew as obtained from periodic density functional theory (DFT) calculations ( Figure 5). For this purpose, we used ac omplete zeolite Beta unit cell containing an open lattice Sn site (Sn/Si = 63). This choice was based on recent studies emphasizing the role of such partially hydrolyzed open lattice sites as the active centers for sugar isomerization in Sn-Beta zeolite catalysts. [31,33,34] The adsorption configurations were constructed following the results of ap revious theoretical study on the mechanism of glucose isomerization by Sn-Beta. [39] We determined the specific interaction energies of adsorbates with the Sn sites (Table S3,F igure S5) and neglected van der Waals (vdW) interactions with the zeolite pore wall, as the dispersion-corrected DFT (DFT-D) method tends to overestimate the vdW contribu-tion. [40] An open Sn site can favorably coordinate two water molecules, two THF and two ethanol molecules or one acyclic glucose molecule. Ethanol adsorbs much weaker on the open Sn site (À43 kJ mol À1 )t han THF,w ater,a nd acyclic glucose (À100, À58, and À48 kJ mol À1 ,r espectively;T able S3). Here we chose acyclic glucose as the reference state over glucopyranose dominating the solution because the interaction of the former with the catalytic Sn sites is much stronger andi ti s also required to enable the rate-determining H-shift reaction within the adsorption complex (Table S3). The current findings on the preferred adsorption of acyclic glucoseo ver glucopyranose on open Sn site are in an apparent disagreement with the previous mechanistic studies, [33,35,41] in which am etastable semi-open configuration of the acyclicg lucose has been discussed as ar eactioni ntermediate within the catalytic path from glucose to fructose. In this work, we selected the most stable conformation of the acyclicg lucose inside beta micropores characterized by af ully stretched and openc arbohydrate backbone. The very strong adsorption of THF to the active Sn centers hinders the adsorption of glucose. In other words, the replacement of THF by glucose is endothermic, which will lead to low coveragew ith the sugar reactant during isomerization catalysis. On the other hand, ethanol can be easily displaced by glucose and this will increase the coverage and activity.T he influence of temperature and entropyo nt he adsorption process over the open site is further evaluated by calculation of the Gibbs free energy of adsorption at at ypical temperature of 298 Ka nd 1atm. The data in Figure 5s how that the entropyl osses owing to adsorption in the pores substantially decrease the favorability of the confinement of all substances inside the zeolitea sc ompared to the energetics predicted on the basis of the electronic DFT energies. However, the qualitative trends in the interaction strengthsa re not affected by correcting for the entropic effects. The differences in isomerization activity can thus be understood in terms of  The stronger the interaction with the solvent molecules, the lower the coveragea nd the isomerizationa ctivity are. The increaseda ctivity seen for the dehydrated zeolite in the in situ NMR experiment can be explained by the lower coverage of Sn centers with water as compared with the hydrated zeolite.
Qualitativelys imilar resultsh ave been obtained with an alternative semi-openS n(m-OH) 2 Si configurationi dentified as intrinsically more stable site in the current periodic Sn-Beta model (see the Supporting Information). Because of the 5-fold coordination of the Sn center in the semi-open configuration, it can only accommodate one solventm olecule in the first coordination shell of Sn ( Figure S5). DFT calculations predict very similar adsorption energies for water,e thanol, and acyclic glucose on semi-open SnOH site in zeolite Beta (À49, À56, and À47 kJ mol À1 ,r espectively). Similart ot he true openS nOH, this site forms very strong complexes with THF (À75 kJ mol À1 )a nd binds glucopyranose very weakly (À1kJmol À1 ).
In summary,wehave shown that competitive adsorption between sugar molecules and polar solvent molecules for the Lewis acid Sn sites in Sn-Beta affect the overall isomerization activity.R oom-temperature NMR experiments demonstrate that under solvent-leanc onditions trioses( DHA) andh exoses (glucose) are more rapidly isomerized than in the presence of water.D FT calculations providef urthers upport for the hypothesis that water competes with theses ugars for adsorption on open Sn sites embedded in the zeolite framework. As olvent, such as ethanol, that interactsw eakly with the active Sn centers allows for higheri somerization activity;i somerization will be suppressed by as trongly interacting solvent, such as THF. The present resultss uggest the importance of competitive adsorptionb etween the reactant and polar solvent molecules during sugar conversion reactions in Sn-Betai somerization catalysts. These insights help to identify improved reaction conditions fort he efficient isomerization of glucose, whichi sa ne ssentialstep in the valorization of lignocellulosic biomass.