Selective Catalytic Oxidation of Cyclohexene with Molecular Oxygen: Radical Versus Nonradical Pathways

Abstract We study the allylic oxidation of cyclohexene with O2 under mild conditions in the presence of transition‐metal catalysts. The catalysts comprise nanometric metal oxide particles supported on porous N‐doped carbons (M/N:C, M=V, Cr, Fe, Co, Ni, Cu, Nb, Mo, W). Most of these metal oxides give only moderate conversions, and the majority of the products are over‐oxidation products. Co/N:C and Cu/N:C, however, give 70–80 % conversion and 40–50 % selectivity to the ketone product, cyclohexene‐2‐one. Control experiments in which we used free‐radical scavengers show that the oxidation follows the expected free‐radical pathway in almost all cases. Surprisingly, the catalytic cycle in the presence of Cu/N:C does not involve free‐radical species in solution. Optimisation of this catalyst gives >85 % conversion with >60 % selectivity to the allylic ketone at 70 °C and 10 bar O2. We used SEM, X‐ray photoelectron spectroscopy and XRD to show that the active particles have a cupric oxide/cuprous oxide core–shell structure, giving a high turnover frequency of approximately 1500 h−1. We attribute the high performance of this Cu/N:C catalyst to a facile surface reaction between adsorbed cyclohexenyl hydroperoxide molecules and activated oxygen species.


Introduction
The allylic oxidation of alkenes is an important chemical reaction. It allows us to keep the double bond and at the same time create an ew alcohol or carbonyl function. [1,2] As such, it is useful across the board, from bulk chemicals anda grochemicals [3,4] all the way to fine chemicals and fragrances. [5][6][7] In theory,a llylic oxidation is as traightforward exothermic reaction. It requires only O 2 ,afree, eco-friendly and widely available reagent. However, there is at rade-off:O 2 hasahigh activation barrier because of its resonance stabilisation. [8] Once this barrieri so vercome, the active oxygen specieso ften react with hydrocarbonsv ia free-radical intermediates. These wreak havoc in solution and cause side reactions that all-too-often lead to unwanted over-oxidation products. [9][10][11] Recently,wes howed that this problem can be solved for the specific case of the oxidationo fa ctivateda lcohols to aldehydes and ketones by using ab ifunctional catalyst. [12] Yett his oxidation was "only" ad ehydrogenation reaction. It involved the transfer of protons and electrons without the addition of a new Oa tom to the substrate. Allylic oxidation is tricky because it is an oxygenation that involves the cleavageo fa tl east one CÀHbond and the creationofnew CÀOorC =Obonds. This re-quires ad irect interactionb etween the substrate and an active oxygen species, which mustt hen be stopped at the allylic alcohol/ketone stage before "burning" further to carboxylic acids, CO and CO 2 .
Here we examine the catalytic oxidationofc yclohexenewith O 2 under mild conditions [Eq. (1)].C yclohexene is ag ood modelc ompound for two reasons:f irst, it is as mall and symmetric molecule, similar to many startingcompounds in chemical synthesis. Second, it is itself industrially important andp articipates in the synthesis cycles of key C 6 chemicals such as adipic acida nd caprolactone. [13,14] Buildingo no ur preliminary communication on alcohol oxidation, [12] we designed as et of metal oxide catalysts supported on N-doped carbons. We used no noblem etals and we focussed on abundantt ransition metal oxidesasc atalysts.
There are several reportso nt he allylic oxidation of cyclohexene catalysed by abundant transition metals. In general,t he use of O 2 as an oxidant requires additional activation, either by the addition of 5-10 %o fafree-radical initiator such as H 2 O 2 or by using elevated temperatures. Zhang and Ta ng [15] reported aC uc atalysto ne xpanded graphite that gave 99 %c onversion and 65 %s electivity to 2-cyclohexene-1-one. Yine tal. [16] and Rossi et al. [17] used Co-based catalysts to obtain 94 %c onversion and 44 %s electivity,a nd 90 %c onversion and 61 %s electivity,r espectively.P eng et al. [18] used metal-free N-doped We study the allylic oxidation of cyclohexenew ith O 2 under mild conditions in the presence of transition-metal catalysts. The catalysts comprise nanometric metal oxide particles supported on porous N-doped carbons (M/N:C, M = V, Cr,F e, Co, Ni, Cu, Nb, Mo, W). Most of these metal oxidesg ive only moderate conversions, and the majority of the products are overoxidation products.C o/N:C and Cu/N:C, however,g ive 70-80 % conversion and 40-50 %s electivity to the ketone product, cyclohexene-2-one. Control experimentsi nw hich we used freeradicals cavengers show that the oxidationf ollowst he expected free-radical pathway in almost all cases. Surprisingly, the catalytic cyclei nt he presence of Cu/N:C does not involve freeradicals pecies in solution. Optimisation of this catalystg ives > 85 %c onversion with > 60 %s electivity to the allylic ketone at 70 8Ca nd 10 bar O 2 .W eu sed SEM, X-ray photoelectron spectroscopy and XRD to show that the activep articles have a cupric oxide/cuprouso xide core-shell structure, giving ah igh turnover frequency of approximately 1500 h À1 .W ea ttribute the high performance of this Cu/N:C catalystt oaf acile surface reactionb etween adsorbed cyclohexenyl hydroperoxide molecules and activatedo xygen species. carbon nanotubes as catalysts. They tested 22 organic solvents and found that acetonitrile gave the best results of 60 %c onversion and 39 %selectivity to the 2-cyclohexene-1-one.
Our initial hypothesis wast hat the allylic oxidation reaction would follow ap athway similar to alcohol oxidation with oxygen activation at the support surfacef ollowed by ar eaction at the oxide particle. Based on previousr eports, we expected the reaction to involve free-radical intermediates. [10,17,19] Surprisingly,w efound that at least in one case, namely,i fw e used copper oxide particles supported on N-doped carbon, there are no free radicalsi nt he solution.I nt his study,w etry to resolve the different pathways that lead to allylic oxidation, with the goal to gain ab etter understandingo ft his important reaction.

Results and Discussion
Catalyst synthesis and testing We began by preparing and testing as et of nine d-block metal oxides supported on the same batch of hierarchically porous N-doped carbons [20] (1.2 mmolg À1 M/N:C, M = V, Cr,F e, Co, Ni, Cu, Nb, Mo, W). The catalysts were prepared using vacuum pore impregnation( see Experimental Section for details). To this set of catalysts, we added two blanks:aclean N:Cs upport and ac arbon prepared from ac itric acid precursor (denoted C cit ), whichh as as imilar surface area to N:C( % 1500 m 2 g À1 ) but contains no N. All catalysts were then tested in cyclohexene oxidation by using an autoclave under 10 bar O 2 and 55 bar Ar,w ithin safe explosion limits.T ypically,e ach autoclave was charged with approximately 25 mmol of cyclohexene, 10 mg of catalyst( an ominal substrate/metal oxide ratio of 2000:1) and 15 mL of MeCN as solvent. Reactions were stirred for 16 hat1 000 rpm and analysed by GC.
Cyclohexene is oxidised to four main products (Table 1): 2cyclohexene-1-one (A), cyclohexeneoxide (B), 2-cyclohexene-1ol (C)a nd 2-cyclohexene-1-hydroperoxide( D;h erein the ketone, epoxide, alcohol and hydroperoxide, respectively). The rest of the products were over-oxidation products,m ainly CO and CO 2 .P roducts A-C were determined directly by GC. The hydroperoxide D could not be observed by GC and wasq uantified by reactinge ach sample with PPh 3 (see Experimental Sectionf or details). Control experimentsc onfirmed that the internal standard, cyclohexane, showedn oc onversion under these reactionc onditions. Further, in the absence of any catalyst, the background reactiona t7 08Cg ives only 22 %c onversion, most of it to CO and CO 2 ( Table 1, entries 1-3). The addition of porousc arbon does not change the conversion but reduces the amount of over-oxidation slightly,p ossibly because of radical-scavengingb yt he carbon surface sites. [21] In the presenceo fp ristine N:C, the conversion more than doubles to approximately 50 %. Moreover,t he selectivity to the ketone A increases to 28 %, at the expense of the hydroperoxide D. Indeed, we showed recently that these porousN :C materials are excellent oxygen reduction catalysts, [20] yett hese results also point to aN:C-catalysed route from D to A (vide infra).
The addition of W, Ni, Mo, Fe or Nb does not change the results significantly (Table 1, entries 4-8). For some of thesec atalysts, the selectivity to A is lower than that of the pristine N:C support, which may reflect the blockingo fl abile sites on the support by metal oxide particles. However,t he catalysts that contain V, Cr,C ua nd Co showed as ignificant increasei nc onversion (entries 9-12). Vanadium oxide (V/N:C), which is known as ag ood epoxidation catalyst, [22,23] gives ah igh selectivity to the epoxide B.T he remaining three catalysts are interesting: they are the only ones to give measureable yields of the alcohol C.A ll three give less hydroperoxide compared with the blanks, which indicates ap athway from D to C.C obalt oxide gives the highest conversion. However,c opper oxide gives the highest selectivity to the ketone A with ar emarkably low amount of over-oxidation products. Even at this un-optimised stage, the Cu/N:C catalyst gives ac ombinedk etone+ +alcohol yield of nearly 45 %w ith am inimum turnovern umber (TON) > 1400 andt urnover frequency (TOF) > 88 h À1 (the actual TON and TOF per site are much higher because most of the copper oxide is not accessible, see discussion below). Therefore, we focussed our investigation on these two catalysts.
Control reactions in whichw eu sed equivalent amountso f cobalt oxide supported on g-alumina showedl ower conversions and more side-products, which confirms the importance of the N:Cs upport ( Table 2, entry 2). Copper oxide supported on g-alumina shows ag ood conversion but with al ower selectivity and more side-products than that supported on N:C, making the g-alumina-supported catalystl essf avourable (entry 6). Notably,t he differencei nt he surface area between the carbon and the g-alumina was corrected for by increasing the catalyst amount accordingly.T ob oost the number of free radicals at the start of the reaction, we added H 2 O 2 (13 mol % relative to the substrate, entry 3). [5,7] H 2 O 2 can decompose into water and oxygen under these reactionc onditions. The water With coppero xide, however,t he addition of H 2 O 2 or an increase of the reaction temperature affected both the conversion and the selectivity (entries 7a nd 8). The conversion increasedt o8 5%,a nd the combined selectivity to A+ +C increasedt o7 0%.I mportantly,t his increase in selectivity came at the expense of the over-oxidation products (unlike with Co, with which there was still al ot of over-oxidation products). To our minds, this was counter-intuitive:w ew ould assume that the addition of an initiators uch as H 2 O 2 or an increase of the temperature would lead to more CO and CO 2 .T hese results led us to think that perhaps the coppero xide catalysed reaction is not as imple free-radical process. Previous reports in which the oxidation kinetics of cyclohexene and [D 10 ]cyclohexene are compared show ac lear primary isotope effect (k H /k D = 8.2), which indicates that the rate-determining step involvesC ÀHb ond scission. [10] Moreover,t he addition of a radicals cavenger quenched the reaction. [10,15] To check if this also applies our system, we ran additional control experiments in the presence of 6mol %o f2 ,6-di-tert-butyl-4-methylphenol (BHT;s ee details in the Experimental Section). The addition of BHT to the reactionm ixture that contained the N:Cs upport or the Co/N:C catalyst stopped the reactionc ompletely (Table 3, cf. entries 2a nd 4w ith 1a nd 3). However,i fw ea dded BHT to the Cu/N:C-catalysed reaction, there was only as light decrease in the conversion and selectivity (from 85 to 76 %a nd from 53 to 46 %, respectively;e ntries 5a nd 6). This shows that althought he reactions catalysed by metal-free N:Ca nd by Co/ N:Ca re definitely free-radical processes, the Cu/N:C-catalysed reactioni sn ot affected by free-radical scavengers (these experiments werer epeated multiple times by different people to ensure their repeatability and reproducibility). Therefore, we concludet hat in the Cu-catalysed system, there are no free radicals in solution.
SEM images (Figure 1) of the Cu/N:C catalyst showed spherical coppero xide particles of approximately 200-250 nm in diameterd istributed evenly across the surface. Unlike the support, the particles are non-porous. Consequently,m osto ft he coppero xide is "inside" the particle and unavailable for catalysis. If we considert hat the active outer shell is approximately five atomic layers ( % 2nmi nt hickness), the actual active catalyst comprises only 0.1 wt %. Accordingly, the actual TON of this catalystw ould be > 24 000 with ac orresponding TOF of > 1500 h À1 .
We used X-ray photoelectron spectroscopy( XPS;F igure 2) to show that the impregnation of the N:Cs urface with copper oxide does not affect the Nb inding energy.T his suggests that the coppero xide is not coordinated to surfaceNatoms. The impregnation increases the intensity of the O1sp eak, which indicates ah ighero xygen content in the sample. ForC u, the XPS spectrum shows the typical Cu 2p 1/2 and Cu 2p 3/2 peaks, which can be assigned to both Cu + + and Cu 2+ + .Y et the characteristic CuO peak at ab inding energy of 945 eV is absent, which supports the presenceo fC u 2 O [24] (metallicC ui su nlikely at such low treatment temperatures [25] and if we consider the increasei nt he Os ignal). The carbon peak is not affected by impregnationw ith Cu. However,p owder XRD patterns of the   . These results are consistentw ith aC uO-CuO 2 core-shell structure as XRD measures the entirep article, whereas XPS penetrates only af ew atomic layers. [26] Therefore, we suggest that during the thermal treatment following the impregnations tep, the adsorbed Cu(NO 3 ) 2 precursor is first converted to CuO and NO 2(g) .A st he temperature approaches 300 8C, the Cu 2 Os hell starts to form. Indeed, temperature-programmed reduction measurements indicate the presence of multiple copper oxides (details in Supporting Information).S imilarly,t hermogravimetric analysis of the pristineN:C and the Cu/N:C samples shows that the latter decomposes at alower temperature (400 vs. 500 8C, respectively). This supports the hypothesis that Cu partially oxidises the surfacet oc reate more labile sites (see detailsi nt he Supporting Information).

Mechanistic considerations
From these results, we propose two alternatives for the catalytic allylic oxidation of cyclohexene with O 2 .T he first follows the traditional free-radical route and pertains to the Co, Fe, Cr,M o, V, Ni, Nb and W/N:C catalysts. Here, O 2 is either activated thermally or in ar edox process on the N:Cs urface. The insertion of this activatedoxygen into the allylic CÀHb ond gives the cyclohexenyl hydroperoxide D.T his can then either rearrange to give the ketone A or undergo scission to give oxo and peroxo radicals that propagate ac hain oxidation reaction. [2,15] Accordingly,t his pathway,w hich involves free radicals in the bulk solution, is quenched readily if BHT is added. Conversely,int he presence of Cu/N:C, there are no free radicals in the bulk solution.O xygen can still be activated at the N:Cs urface sites but now there are two options:t he small amountso fs hort-lived activated oxygen species (e.g.,O 2 À C radical anions) that travel into solution will be quenched by BHT (Figure 3a,c f. also the difference in conversion between entries 5a nd 6i nT able 3). The BHT molecules are too bulky to enter the micropores. Therefore, they will quench only the radicals in the solution. Conversely,t he activated oxygen species that are close enough to diffuse to as upported copper oxide particle [27] can react there with cyclohexene to form an adsorbedh ydroperoxide ( Figure 3b). This adsorbed hydroperoxide can undergo two reactions. The first is rearrangement and dehydration to give the ketone A and amolecule of water (Figure 3c). [9,28] The second is ad isproportionation reaction with another cyclohexenem olecule to give two molecules of cyclohexene-1-ol C (Figure 3d). Compared with the other metal oxides, the scission of the ROÀOH bond on the copper oxide surface is apparently much slower.T his means that fewer free radicals are released into the solution,w hich gives enough time for the rearrangement and disproportionation reactions. Interestingly,t here is am arked difference between the oxidation of activated alcohols, whichw er eported earlier, [12] and that of cyclohexene. With an activated alcohol substrate such as cinnamyl alcohol, the N:Cs upport is required for oxygen activation.T here, no reaction was observed for coppero xide particles supported on C cit ,a na nalogousp orous carbon with no Nd opants. Cyclohexene oxidation, however,d oes proceed in the presence of Cu/C cit ,w hich showst hat the allylic oxidation in this case is easier.T his is supported by the results of Gray and co-workers [10] who showed that the allylic CÀHb onds cission is rate-determining and by the fact that this bond is weaker than the alcohol CÀHb ond (83 and 96 kcal mol À1 ,r espectively [29,30] ).
In all cases, the epoxide B probably forms by another pathway. [4,31] Cyclohexene molecules can interactw ith M=Og roups on the particle surface to give the epoxidea nd al abile surface site that is then re-oxidised by incoming oxygen. [24] Conclusions The catalytic oxidation of cyclohexene with O 2 can follow different pathways that depend on the type of catalyst. In the presence of transition metal oxide nanoparticles supported on N-doped carbons, the key step is the insertiono fO 2 into the allylic CÀHb ondt og ive the cyclohexenyl hydroperoxide. This reactionc an be enhanced by oxygen activation att he Ndoped carbon surface. In most cases, the allylic oxidation follows af ree-radical pathway.H owever,i nt he presence of Cu/ N:Ct he reactiond oes not releasef ree radicals into solution. This enables am ore selectiver eaction at the copper oxide surface, whichprobably involves cuprous oxide sites.

Experimental Section
Materials and instrumentation GC was performed by using aP erkinElmer Clarus 580 instrument. This system was equipped with af lame ionisation detector and autosampler (G4513A). A3 0m 32 mm I.D. Rxi-5 ms fused silica crossbond diphenyl dimethyl polysiloxane column with af ilm thickness of 0.25 mmw as used. The injector volume was 1 mL, and the flow was 100 mL min À1 of He carrier gas. The temperature program was 40 8C, 20 8Cmin À1 ,1 60 8Cf or 2min. SEM was performed by using aV erios-460 microscope (FEI) at an accelerating voltage of 5kVw ith aw orking distance of 2-5 mm. Powder XRD patterns were obtained by using aM iniFlex II diffractometer by using Ni-filtered CuK a radiation. The X-ray tube was operated at 30 kV and 15 mA with a0 .018 step and 1s dwell time. XPS measurements were performed by using aP HI VersaProbe II scanning XPS microprobe (Physical Instruments AG, Germany) using am onochromatic AlK a X-ray source with ap ower of 24.8 Wa nd ab eam size of 100 mm. The spherical capacitor analyser was set at a4 5 8 take-off angle with respect to the sample surface. The pass energy was 46.95 eV to yield af ull width at half maximum of 0.91 eV for the Ag 3d 5/2 peak. Peaks were calibrated using the C1sp osition. Curve fitting was performed using the XPSPeak 4.1 software package. All chemicals were obtained from commercial sources (> 99 %p ure) and were used as received. Temperature-programmed reduction (TPR) was performed by placing 25 mg of sample sandwiched between two quartz wool plugs in aq uartz tube reactor (4 mm i.d.). After purging with N 2 ,aflow of 5% H 2 in N 2 was applied. The system was allowed to equilibrate and then heated at 5 8Cmin À1 to 800 8C( no hold time). The N:Cs upport samples were prepared following the procedure published by Eisenberg et al. [20] Briefly,n itrilotriacetic acid (NTA) was mixed in a1 :1 ratio with magnesium carbonate. This was dissolved in de-ionised water,s tirred for 10 min at 85 8C, and cooled to RT.T he solid was then precipitated by adding an excess of ethanol and chilling in an ice bath for 2h.T he white solid was scraped out, dried at 40 8Cf or 48 h, and ground into af ine white powder. This powder was then pyrolysed in Ar at 900 8C. The MgO particles were washed with 3 500 mL of 0.5 m citric acid. The resulting crude N:Cs ample was dried at 120 8Cf or 2hand treated under Ar at 1000 8Cf or 1h.

Preparation of M/N:C catalysts
This is am odification of the procedure published by Slot et al. [12] The N-doped carbon support (100 mg) was placed in as mall vial with as eptum. The air was removed carefully by using an eedle, and an aqueous solution of the desired metal precursor salt (0.2 mL, which corresponds to an ominal loading of 1mmol m À2 ) was added to the vial under continuous stirring. The vial was shaken vigorously for 2-3 min to create au niform solid paste, which was then dried at 85 8Cf or 12 h. Each catalyst was then heat-treated at 300 8Cu nder Ar (except for Nb/N:C, which was treated at 700 8C) and cooled to RT.T he different M/N:C catalysts were prepared from their respective precursors salts: Co(NO 3

Catalytic oxidation of cyclohexene
This is am odification of the procedure published by Cao et al. [32] A 75 mL autoclave lined with a5 0mLT eflon insert was loaded with cyclohexene (2.5 mL, 24.7 mmol), cyclohexane (0.5 mL, internal standard), acetonitrile (solvent, 15 mL), catalyst (10 mg M/N:C carbon or 73 mg Co/alumina) and as tirring bar (30 mm). The autoclave was sealed, flushed with Ar and O 2 twice before the final O 2 (10 bar) and Ar (55 bar) atmosphere was applied. The autoclave was then heated to 70 8Cf or 16 hw ith stirring at 1000 rpm. After 16 h, the autoclave was cooled to RT.A cetone (5 mL) was added to the sample, and the reaction mixture was filtered using 0.45 mm PTFE syringe filters and analysed by using GC.
The presence of free radicals in solution was tested by adding BHT (354 mg, 6mol %) to the reaction at t = 0a nd then following the above procedure. Reactions were performed in triplicate, and all GC analyses were performed in duplicate.

Quantification of the cyclohexenylh ydroperoxide D
The hydroperoxide D cannot be measured directly by using GC because of its instability.I nstead, we quantified it by comparing a control reaction sample to one in which triphenylphosphine (PPh 3 , 30 mg, 1mol %) was added. The sample was shaken for 1min, and heat was generated as the PPh 3 reacts with the hydroperoxide D to give PPh 3 Oa nd the alcohol C [Eq. (2)].A fter this reaction, the sample was analysed by GC and compared to its untreated counterpart. The subtraction of the initial amount of the alcohol formed in the control reaction from the amount of the alcohol after the addition of PPh 3 gives the amount of hydroperoxide D in the original sample. [6,33]