Visible‐Light‐Controlled Oxidation of Glucose using Titania‐Supported Silver Photocatalysts

Abstract The visible‐light‐mediated photo‐catalytic selective valorisation of glucose using TiO2‐supported Ag nanoparticles is shown for the first time. The optimisation of the catalyst composition, substrate‐to‐catalyst ratio and reaction medium proved that a near total suppression of the mineralisation pathway could be achieved with a selectivity to partial oxidation products and small‐chain monosaccharides as high as 98 %. The primary products were determined to be gluconic acid, arabinose, erythrose, glyceraldehyde and formic acid. Under UVA light, the selectivity to organics decreases because of the production of CO2 from mineralisation. A reaction mechanism is proposed based on an α‐scission process combined with the Ruff degradation reaction, which explains the presence of the oxidation products, the smaller carbohydrates and formic acid. X‐ray photoelectron spectroscopy, UV/Vis spectroscopy and microscopy studies showed the presence of plasmonic 4 nm particles of silver that were oxidised to silver oxide over the course of the reaction, and recycling studies revealed that this was not detrimental to activity.


Introduction
To date, photo-catalysis has dealt typicallyw ith environmental remediation, sterilisation and decontamination of polluted water streamst hrough the total oxidation or "mineralisation" of organic contaminants. [1] Solar hydrogen production through water splitting and reformingi sa lso ap romising new technology for green fuel generation, although limitations in energy conversion efficiency,H 2 production volumes, catalystdeactivation and H 2 storages tillf etter this emerging technology. [2] Recent advances suggest that photo-catalytic routes can perform selectiveo xidation, [3] epoxidation,r eduction, carbonylation and cyclisation reactions [4] efficiently to offer ap otentially inexpensive, green and chemically benign method for the functionalisation and transformation of chemicals. The selective photo-catalytic oxidation of bio-derived C 1 -C 4 alcohols has been studied briefly,w hich indicates that high selectivities to the corresponding formates [5] and aldehydes [6] can be achieved if TiO 2 is used as ac atalyst.T he conversion andd ecarboxylation of someo rganic acids using metal-doped TiO 2 [7] and Pt/TiO 2 [8] catalysts have also been shown. The harvestingo fs unlight to drive chemical reactions for the productiono fh igh-value chemicals from waste by-products and renewable sourcess uch as biomass is the natural next step in the creationo fagreen and sustainable chemical industry,b ut there are still challenges to overcome. [9] Biomass comprises cellulose, hemicellulose and lignin, which are difficult substrates to work with because of their complex molecular structures and the large number of functional groups present on the carbon backbone. For this reason, researchs ince the early 1970s has focussed on the conversiono f smaller carbohydrates by photo-catalytic routes. Glucose represents the ideal substrate;i ti st he mostc ommon and cheapest carbohydrate availablei nn ature and it can be obtained from lignocellulosic waste biomass through the hydrolysiso fits constituent polysaccharides (cellulose and hemicellulose). Glucose can be valorised into platform chemicals, such as glucaric acid, arabitol, levulinic acid and hydroxymethylfurfural, under relatively mild conditions with good yields. [10] Early reviews surveyed the possible application of glucose oxidation products as chiral intermediates with potential applicationi nt he pharmaceutical industry and as precursors for vitamin Ca nd other high-value chemicals. [11] Commercially,g lucosei saprecursor of gluconica cid, which is used in the pharmaceutical, food, health and textile industries. [12] Glucose transformationsa re performed typically by fermentation and enzymatic routes but these often suffer from poor rates, low yields and the high cost of the enzyme used. [12] An alternative selective route using ah eterogeneous catalyst offers ap otentially more robust pathway.
The visible-light-mediated photo-catalytic selective valorisation of glucose using TiO 2 -supported Ag nanoparticles is shown for the first time. The optimisation of the catalystc omposition, substrate-to-catalyst ratio and reaction mediump rovedt hat an ear total suppression of the mineralisation pathway could be achieved with as electivitytop artial oxidation products and small-chainm onosaccharides as high as 98 %. The primary products wered etermined to be gluconic acid, arabinose, erythrose,g lyceraldehyde and formic acid. Under UVAl ight, the selectivity to organics decreases because of the production of CO 2 from mineralisation. Ar eaction mechanism is proposed based on an a-scission process combinedw ith the Ruff degradation reaction, which explains the presenceo ft he oxidation products,t he smaller carbohydrates and formic acid. X-ray photoelectron spectroscopy, UV/Vis spectroscopy and microscopy studies showedt he presence of plasmonic 4nmp articles of silver that were oxidised to silver oxide over the course of the reaction, and recycling studies revealed that this was not detrimental to activity.
The selectivec atalytic oxidationo fg lucose continues to be of interest to researchers. [13] As early as the 1940s, an umber of scientists showed the efficiency of Pt and Pd-based catalysts for the production of gluconic acid, which resulted in several patents. [14] More recently,A u, Pd and Pt have been shown to be active and selective, whichh as led to ar esurgence and further effort to make the reaction am ore economically viable process. [13,15] However,t ypically, the reactions requiret he use of bubbling or pressurised O 2 along with the constant addition of base to maintain the catalyst activity. [13a, 15, 16] Along with the well-established chemical conversion of sugars to platform chemicals, in recent years, academic attention has shiftedt owards the application of photo-catalytic routes to obtain the same valuable chemicals using much less energy-intensive processes and milder reactionconditions.
The photo-conversion of carbohydrates to produce gluconic and glucaric acid using TiO 2 catalysts under UV light has been investigated by Colmenares et al. [17] More recently,d etailed studies have been published by Chong et al. [18] and Bellardita et al. [19] in which several reactionm echanismsh ave been suggested to explain the reactivity of carbohydrates and the possible interactions with the severalp hotoactive materials. [20] Most recently,w eh ave demonstrated that TiO 2 can successfully convert glucoset oh igher value products under visible light throught he formation of al igand to metal charget ransfer complex. [21] The TiO 2 -basedm aterials tested in the studies cited above were tested under pure UVAi rradiation or by using Xe lamps with no specific filters installed, which makest he comparison of the experimental results obtained difficult because of the variability of the reaction conditions.
Herein, we present ad etaileds tudy on the effect of both UVAa nd visible light upon the photo-conversiono fg lucose using TiO 2 -supported Ag nanoparticlesa nd demonstrate for the first time that glucose can be converted to gluconic acid and other monosaccharides under visible light (l > 420 nm). The systematic analysiso fs everal reaction parameters allowed the identification of an ew reactionm echanism that comprises three reaction pathways by using HPLC with quadrupole timeof-flight mass spectrometry (Q-TOF-MS) to enablet he determination of relevant and previously unaccounted for reaction products.

Photo-catalysed glucose transformation under visiblelight
The decoration of the catalysts urface with metal nanoparticles (typicallyA g, Au) extends the activity of the TiO 2 support to the visible part of the electromagnetic spectrum because of the plasmonic effect. This interaction is responsible for the excitation of metal nanoparticles under visible light and the energy transfer from these nano-antennas to the support and, subsequently,tot he substrate. [22] The 0.5-1.5 wt %A g/TiO 2 catalysts were prepared using aw et impregnationp rotocola nd tested both under UVAa nd visible light. Initialt ests performed using aqueous glucose sol-utions showed negligible catalytic activity,t herefore, the MeCN/H 2 O( 1:1v /v) system studied by Colmenares et al. [17d] was investigated. Blank reactions were performed in the dark by using aL uzchem photoreactor( Figure S1) at 30 8C. None of the catalysts was active for glucose conversion under these conditions.
The time-on-line (TOL) glucose conversion in which the bare TiO 2 support is compared with those that bear nanoparticles is showni nF igure 1. The addition of the metal nanoparticles enhances the activity of the material towards glucose oxidation under visible light. However,t his promotional effect is more pronounced if the lowest amounto fA gw as used (0.5 wt %), which resulted in at hreefold increased activity.
The modest 2% conversion recorded with the bare TiO 2 after 120 min of irradiation increased to 6% with the addition of low concentrationsofAg. The suppressiono fthe mineralisation pathway to CO 2 was essentially complete in all cases, as shownb ythe > 99 %m ass balance values ( Figure 2). However,b yi ncreasing the Ag loading to values higher than 0.5 %t he glucose conversiond ecreased from 6t o3%f or the 1.5 wt %A g/TiO 2 catalyst, which indicated that the availability of the TiO 2 surface is acriticalparameter as reported previously for Ag-type systems by Grabowskae tal. [23] In ar ecent study, Fu et al. [24] showedt hat the H 2 generation from glucose reforming using Pt/TiO 2 catalysts was relatedt ot he dispersion of the nanoparticles on the catalyst surface. Therefore, it appears that the best metal loading for glucose oxidation under the experimental conditions used in this studyi s0 .5 wt %. The reason for the catalytic enhancement if Ag nanoparticles are supported on aT iO 2 semiconductor can be understood if we consider the Schottkyb arrier at the metal-support interface, which slows the electron-hole recombination, prolongs their lifetimea nd, subsequently,e nhances the activity of the TiO 2 ,a s reported elsewhere. [22a, 25]  There is ac orrelation between the number of Schottky barriers and the catalytic efficiency of the material. If the metal nanoparticles are too close to each other,t hey act as electronhole sinks or recombination centres, which thus reduces the availability of these speciestop articipate in redox processes.
Finally,d ifferent metal loadings have al imited impact on the product distribution after 120 min of irradiation, andt he selectivity towards the partial oxidation products is the same within experimentale rror.I na ll cases (which includes the bare TiO 2 ) the main reaction products are arabinose( > 35 %), formic acid (~30 %) and gluconic acid (15-18%)a long with erythrose and glyceraldehyde ( Figure 2). The product distribution values agree with previousresults. [17d, [18][19]

Photo-catalysed glucose transformation under UVAl ight
Glucose conversion under UVAl ight using bare TiO 2 andt he 0.5-1.5 wt %A g/TiO 2 catalysts is shown in Figure 3.
Undert hese conditions, the highest conversion (11.5 %) was achieved with the 1.5 wt %A g/TiO 2 catalystv ersus9%f or the bare TiO 2 .Under UVAirradiation, the presenceo fA gn anoparticles has no beneficial effect on the activity of the materials for the selective oxidation process but does promote the mineralisation reaction for the production of CO 2 ;t he mass balance values get lower with an increased Ag loading from 0.5 to 1.5 wt % ( Figure 4).
Unlike the previousc ase in which visible-light irradiation was used, under UVAi rradiation, both the support and the metal nanoparticles can be excited simultaneously,a nd the dominant reactionm echanism is the charge-separation step in the TiO 2 in which the photo-generated electrons are transferred from the support to the metal nanoparticles. [26] Also, the intra-bandt ransition of the electrons within the metal nanoparticle from the fully occupied db ands below the Fermi level to the half-filled sp bands has to be considered. The activity of the Ag nanoparticles under thesec onditions is associated with the promotion of 4d electrons to 5sp orbitals. [27] The holes left in the inner do rbital have ag reater tendency to capturee lectrons than the outermost sp orbitals and, therefore, act similarly to the electron-hole couple generated in as emiconductor in which the electron vacancy in the do rbital acts as ah ole. [28] The energy required to promote these transitions is much higher,h ence the necessity for UV irradiation. This mechanistic difference explainsw hy under UVAi rradiationt he best-performingm aterialw as that with the highest metal loading, which is different to our observations under visiblel ight.
Although the combined excitation of the Ag nanoparticles and the TiO 2 enhances the production of the radical species responsible for the photo-activity of the materialf or example, OHC,r eactive oxygen species( ROS)a nd h + + ,t his results in unselective glucose conversion towards CO 2 .A dditionally,w eo bserved as hift in the product distribution values as arabinose is the primary reaction product with selectivity values above

Reaction mechanism
The susceptibility of the reaction products and intermediates to further radical attack along with the difficulties in the quantification of complexm ixtures of sugar isomers with oxidation and degradationp roducts makes the full quantification of reaction products veryc hallenging, that is, it is very difficult to achieve a1 00 %m ass balance.T herefore, it is extremelyd ifficult to ascertain an adequate and comprehensive reaction mechanism fora ll the observed reactivity.R eaction mechanisms for the photo-conversion of carbohydrates have been proposed before based on the observed reaction products and intermediates, [17d, 18,19] but the reactionp athways described are representative of the reaction set-up used in each study,t hat is, the light source,s ource power, the solvent used andt he photo-catalyst. In some cases, the presence of particularm olecules is neglected in the depiction of the reactions cheme. Stapley and BeMiller [29] reviewed the decarboxylation of sugars and sugar acids to produce smaller-chain carbohydrates. Specifically,t hey reviewed the so-called Ruff degradation that involves the decarboxylation of aldonic acids by Fe III and H 2 O 2 in aF enton-like system to produce smallerc arbohydrates. They also report that Ti IV behaves similarly under the same reaction conditions, which explains the reactionm echanism in the absence of Fe III species. [29,30] Based on this mechanism,w ed ecided to investigate the behaviour of gluconic acid stock solutions under UVAa nd visible-light irradiation to assess the resulting product distribution using the 0.5 wt %A g/TiO 2 catalyst (Figure 5a nd Figure S5). The use of a2 0mm gluconic acid solution in the MeCN/H 2 Om ixture produces as imilar product distribution to that of the glucose substrate. With the metalsupported catalyst, approximately 17 %g luconic acid could be detected after 30 min of UVAi rradiation, whereas 120 min was necessary to obtain as imilar selectivityi fb are TiO 2 was used ( Figure S5). However,i nt he case of gluconic acid, significant amountsofformic acid were found, and its presence could not be linked directly to the Ruff mechanism as it involves the productionofs olely CO 2 .
Therefore, am ore complex scheme that comprises multiple reaction pathways is neededt od escribe the system fully.
The overall reaction mechanism depicted in Scheme 1c ombines and explainst he observations by Colmenares et al. [17d] and Chong et al. [18] and, based on the findings of our experiments,a lso sees the inclusiono ft he Ruff degradation step. Our proposed mechanism agreesp artly with the mechanism suggested by Chong et al., [18] whereby a-scission generates the successive formationo fs horter-chain carbohydrates with the formation of equimolar hydrogen and formic acid. The H 2 produced was determined qualitativelyb yu sing headspace GC analysis over 24 hr eaction time ( Figure S13), but because of the nature of the reaction it was not possible to determine the H 2 /CO 2 ratio because of the mineralisation reactiont hat occurs in parallel with the a-scission pathway,t hat is, not only is the production so low as to be near the detection limits of the analytical method employed but any H 2 evolvedf rom water splitting cannotb es eparatedf rom the H 2 from the mineralisation of the glucose. However, we could obtain high amountso f gluconic acid both under visible and UVAl ight (Figures 2a nd  4), and an a-scission mechanism aloned oes not explain the formation of glucose oxidation products (Scheme 1). We believe that photo-catalytic oxidation reactions are responsible for the oxidation products observed andt hat the resulting acid products react further to decarboxylate as depicted in Scheme 1. From our data, it is apparent that the glucose is first oxidisedt og luconic acid before it undergoes the a-scission of the C 1 ÀC 2 bond to allow the formation of arabinose and formic acid. The arabinose then undergoes subsequent repeat CÀC cleavage to form erythrose and glyceraldehyde. Furthermore, the presence of CO 2 as ap roduct cannot be attributed solely to the mineralisation of formic acid as reported previously.
The photo-catalytic conversion of gluconic acidp roduces am uch-simplified reaction profile (Scheme 1), whereby the only detected products correspond to consecutive a-scission products in the order 1) arabinose,2 )erythrose, 3) glyceraldehyde and4 )formic acid. However,w ew ere only able to detect formic acid after 90 min ( Figure 5). It is clear that the carbohydrates are indeed formed from gluconic acida nd that the reaction proceeds through the partial oxidation of glucoset og luconic acid and sequential decarboxylation.N otably,a nalysiso f the products from the gluconic acid experiment by using MS revealed no glucaric acid present,w hich indicates that under these conditions the oxidation of gluconic acid to glucaric acid does not take place. Furthermore, no arabitol was observed in any of the reactions as had been reported previously. [17d] Althought heser eactions were conducted under different conditions, the presence of arabitol would be possible throught he reduction of arabinose, but this would be unexpected in our case. Furthermore, the Q-TOF-MS analysis ( Figures S6 and S7) shows the absence of arabitol for the two control reactions performed under UVAand visible light. The role of the photo-catalyst is to generate h + + during the photo-catalytic process that act as anodest oo xidise the organic molecules absorbed on the surfaceo ft he catalyst: initially glucose to gluconic acid. As ar esult of the high oxidising potential, molecules adsorbed on the surface such as gluconic acid can undergo decarboxylation through am echanism similar to aR uff degradation (seen typically with Fe III and H 2 O 2 ), but performedhere by Ti IV and the photo-generated radical species. [30b] This photo-oxidativep athway does not result in the production of formica cid in each step, so it appears to be the dominantpath during the early stages of the photo-oxidation process ( Figure 5). The appearance of formic acid at longer reactiont imes under visible light suggests that both the a-scission and aR uff-type degradation mechanism take place simultaneously.However,under UVAirradiation, the presence of gluconic acid can be detected after 30 min of irradiation as highlighted previously ( Figure S5). The production of the two acids (formic and gluconic) can only be explained in this case if the two reaction mechanisms (a-scission and Ruff degradation) occur simultaneously.I nt his respect, it is evident how the metal nanoparticlesp romote the formation of gluconic acid as an intermediate in the glucose oxidative decarboxylation. The kinetic production of the acid intermediate is clearly faster than the a-scission process as the 17 %s electivityo bserved at the beginning remains constant throughout the reaction. Further work on reactionc onditions:[ O 2 ]c ontrola nd the nature of the catalyst, reaction mediuma nd irradiation source will potentially offer control with regard to whichr eaction pathways can be promoted or demoteda nd be the key to obtain selectivity and further insights into the mechanism. The effect of the substrate was analysed by using three glucose stock solutions with concentrations in the range of 2.8-20 mm. If the most dilute glucoses olutionw as used (2.8 mm)t he mineralisation pathway played as ignificant role in the glucose conversion with am ass balance value < 90 %u nder visible light and as low as 82 %u nder UVAi rradiation. If the substrate concentrationw as increased, the mass balance was significantly better with values greater than 95 %f or the 20 mm stock solution.The increased substrate concentration did not affect the product distribution,a nd the relative ratio of the partial oxidation products remains within the experimental error (Figure S8). Therefore, it is clear how the surfacec overage of the catalystp lays ap ivotal role to determine the activity of the system andw hich reactionpathway will be more dominant.

Catalysts characterisation and recycling studies
Solid-state UV/Vis spectroscopy was used to assess the presence of the metal nanoparticles on the catalysts urface. We used aT auc plot [31] to evaluatet he band gap of solid samples, and if the Kubelka-Munk functioni su sed with the reflectance plotteda gainst the wavelength energy [32] it is feasible to isolate the presence of the metal nanoparticles from the support as shown in Figure 6.
Scheme1.Proposed reaction mechanism for the photo-catalytic conversion of glucose. This global pathway includes the a-scission of sugars suggested by Chong et al. [18] along with the oxidative decarboxylation mechanism typical of the Ruff degradation. [29] This mechanism was found to be applicabletod escribet he reaction products obtained under visible and UVAlight.T he products showni nb rackets could not be isolated from the reactionm ixture. The TEM images and particle size distribution determined from 300 particles fort he 0.5-1.5 wt %A g/TiO 2 catalysts are shown in Figure 7. All catalysts have particles in the size range of 1-8 nm with am ean particles ize of approximately 4nm. Althought he 1% Ag/TiO 2 had am ean average of 3.4 nm, recycling experiments indicatet hat an increase in particle size as ar esult of photo-induced agglomeration has am inimal effect on the activity of the catalyst (Figures 8and 9).P revious results show that plasmonic effects become negligible in nanoparticles smaller than 2nm, and size differences between particles of < 5nmh ave an egligible effect on their absorbance wavelength. [33] Elemental analysis showedt he actual loading to be lower than the theoretical value in all cases, which was tolerable (Table 1).
The chemical environment and valence state were determined by using X-ray photoelectron spectroscopy (XPS), and the Ag 3d 3/2 and 3d 5/2 peaks are shown in Figure 10. The binding energies (BEs) for Ag, Ag 2 Oa nd AgO are very close within 367.3-368.4 eV,b ut there is ac lear shift in binding energy from 368.1 to 367.7 eV with increased Ag loading indicative of the formation of Ag + +1 species.T he ease of oxidation of Ag nanoparticles under air is not without precedent. [34] The presence of the oxidised Ag speciesc annot be linked directly to    the nature of the reaction considered and did not hindert he photo-activity of the material, and Ag 2 Os pecies on TiO 2 supports have been shown previously to be active under visible irradiation. [35] The presence of both metallic Ag and the oxides in all the catalysts cannotb ed iscounted as Ag 0 wase vident from the TEM analysis of 0.5 wt %A g/TiO 2 ( Figure S9) in which the interplanar distance of 0.24 nm for the particles can be attributed to the preferentiale xposure of the 111 plane (ICDD 01-071-3672).
Representative solid UV/Vis spectrao f1wt %A g/TiO 2 (Figure S10) display ar edshift in the plasmonic-resonance peak after multiple re-uses both under visiblea nd UVAl ight. This slight shift from l = 383 to 420 nm is caused by oxide formation and changes in the nanoparticles morphology after multiple re-uses. [35d] XPS of the catalyst after multiple re-uses showed that the bindinge nergy shifts to 367.3 eV indicative of the presence of Ag 2+ + species,inthis case, AgO;this has been reported previously for similar systems. [35d, 36] The signal attenuation corresponds to an apparent decreasei nt he surface concentration from 0.08 to 0.03 at %( Ta ble 1). However,t he energy-dispersive X-ray spectroscopy (EDX) microanalysis shows that the bulk concentration remains constanta ta pproximately 0.1 at %( av alue in good agreement with the 0.08 at %o btained by using XPS), whichd emonstratest hat there is no leaching of the metal and that the lower surfaceconcentration is caused by particle sintering.
Further investigation by using EDX mapping showed that the lower surface concentration determined by using XPS is ar esult of light-induced particle agglomeration as large particles > 200 nm can be seen throughout the titania matrix ( Figure 8).
Interestingly,t he same effect wasn ot observed for the catalysts recycled under UVAl ight (data not shown). Ag nanoparticles are known to show photo-chromicb ehaviour if they are exposed to different light sources, which involves morphological changes of the supported metal nanoparticles because of the interaction of the incident light as reported elsewhere. [37] Upon illumination of the supported nanoparticles, it is possible to influencet heir shape to obtain smaller satellite metallic structures from biggerp articles. [38] Additionally,t he accumulation of the well-dispersed nanoparticles into largera gglomerates was also observed. This effect can be explained by the super-heating of the metal nanoparticles upon irradiation. Huang et al. [39] showed recently that Ag nanoparticles can be super-heatedw ith femto-laser pulses and that the strong electric fields cause the agglomeration of the particles into larger structures. Even in our case after exposure to the variousl ight sources, the powders turned to ab rownish grey colour as reported by Naoi et al. [40] This colour change is caused by changes in the morphologya nd size of the nanoparticles as well as in the refractive index of the support. The glucosea dsorptionm echanism and the colour change associated with the formation of the glucose-TiO 2 complexr eported by Kim et al. [41] show how the refractive index of the TiO 2 support is affected by the support-substrate interaction as shown by the FTIR spectroscopica nalysis performed on the 0.5 wt %A g/TiO 2 catalysta fter multiple re-uses under visible and UVAl ight (Figure S11).
However,despite the morphological changes of the catalysts after multiple re-uses, the activity and product distribution valuesr emained unaffected. The data presented in Figure 9 show that the glucose conversion under both UVAa nd visible light for 1wt% Ag/TiO 2 remains at approximately 10 and 4%, respectively,a nd retains the same mass balance andp roduct distribution values after three runs ( Figure S12).

Conclusions
Here, for the first time, we have shown how visible light can be used to transform the renewable feedstock glucose to higher-value organicss uch as gluconic acid, arabinosea nd formic acid using Ag/TiO 2 catalysts. Thep romotion of TiO 2 using plasmonicA gn anoparticles resulted in enhanced conversion and high selectivity (> 98 %) to the desired products with an ear total suppression of the mineralisation pathway.
The catalyst was re-usable and showed no loss in activity or changes in the product distribution. We used TEM analysist o reveal how the nanoparticles are unstable under reactionc onditions but this was not detrimentalt oa ctivity.I na ddition to Ag, we anticipate that other plasmonic nanoparticles, such as Au, Cu and their alloys, could promote this reaction similarly to offer an ew avenue to control the selective photo-catalytic upgradingo fb io-derived polyols and saccharides using visible light.

Experimental Section
Ag/TiO 2 synthesis

Recycling of Ag/TiO 2
The recycling of the Ag catalyst under visible and UV radiation was performed following ap yramidal scheme:t hree reactions were run for 2h under the same experimental conditions using 14 mg of catalyst. The recovered catalyst was centrifuged, and the supernatant was removed. The catalysts were then washed with H 2 Oa nd ethanol three times to remove any organics adsorbed on the catalyst. The washed powders were dried overnight at 50 8Ca nd then ground. The recovered catalyst was used in the second run for two reactions and, after the reaction, it was treated following the procedure described above. Finally,f or the third run, only one reaction was analysed. The values provided for the conversions and the product selectivity for the first and the second run are the average of the results obtained for each of the reactions.

Catalyst testing
The

Product analysis
The standard solution and the reaction products were analysed by using a1 200 HPLC Agilent (Agilent, USA) system equipped with an inline degasser,aquaternary pump, an autosampler and ac olumn switch. The selected detectors were ap hotodiode array detector (DAD) and ar efractive index detector (RID). The analytical column was an Aminex HPX-87H (300 mm 7.8 mm), 9 mmp article size (Bio-Rad CA, USA) column kept at 65 8Cw ith 0.025 m H 2 SO 4 as eluent with af low rate of 0.65 mL min À1 .B efore analysis, the samples were centrifuged at 13 400 rpm for 1min to remove any suspended particles. The glucose and the reaction products were determined using commercially available standards. The accurate mass of the oxidised products obtained from glucose was analysed by using an Agilent 6510 LC-Q-TOF-MS system and interpreted by using Agilent MassHunter Workstation Software (Version B.06.00). The column used for the MS analysis was aV arian MetaCarb 67H (300 mm 6.5 mm;A gilent, USA) kept at 65 8Cu sing a0 .1 %w /w formic acid aqueous solution at af low rate of 0.8 mL min À1 .T he Q-TOF-MS was operated in positive ESI mode.

Catalystc haracterisation
TEM was performed by using aJ EOL 2100 instrument (Jeol Ltd, JPN) operated at 200 kV.S amples were prepared by dispersion in methanol with sonication and deposited on a3 00 mesh holey carbon film.
XPS analysis was performed by using aT hermo K-Alpha (Thermo Scientific, East Grinstead, UK) with am icro-focused mono-chromatic AlK a source (1486.6 eV,1 2kV, 3mA, 36 W) with as pot size of 400 800 mm. The data acquired were obtained from the analysis of three positions per sample with ag eneral 30 scan survey and a1 0scan survey for the high-resolution regions. The raw data were corrected by using the C1sb inding energy at 284.7 eV.T he recorded spectra were fitted with least squares to produce Gaussian-Lorentzian functions after the subtraction of background noise.
Solid-state UV/Vis spectroscopy was performed by using aU V-2550 Shimadzu spectrophotometer equipped with an ISR-2200 integrating sphere (Shimadzu Corp, JP) in the range of 200-800 nm with a0 .5 nm sampling interval and a5nm slit using BaSO 4 as reference. The reflectance data were used to calculate the Kubelka-Munk function using the absolute reflectance (R 1 )t od etermine the plasmon resonance of the supported Ag nanoparticles.
The EDX and SEM analysis of the samples was performed by using aH itachi S-4800 field-emission microscope equipped with an Oxford Instruments Inca Energy EDX detector.T he electron accelerating voltage was 30 kV with ap robe current of 20 mA. Samples were analysed uncoated and the EDX measurements represent the average of am inimum of three points over the material.