Cerium oxide nanoparticles inside carbon nanoreactors for selective allylic oxidation of cyclohexene

The confinement of cerium oxide nanoparticles within hollow carbon nanostructures has been achieved and harnessed to control the oxidation of cyclohexene. Graphitised carbon nanofibres (GNF) have been used as the nanoscale tubular host and filled by sublimation of the Ce(tmhd) 4 complex (where tmhd = tetrakis(2,2,6,6-tetramethyl-3,5-heptanedionato)) into the internal cavity, followed by a subsequent thermal decomposition to yield the hybrid nanostructure CeO 2 @GNF, where nanoparticles are preferentially immobilised at the internal graphitic step-edges of the GNF. Control over the size of the CeO 2 nanoparticles has been demonstrated within the range c.a. 4 to 9 nm by varying the mass ratio of the Ce(tmhd) 4 precursor to GNF during the synthesis. CeO 2 @GNF were effective in promoting the allylic oxidation of cyclohexene, in high yield, with time-dependent control of product selectivity, at a comparatively low loading of CeO 2 of 0.13 mol%. Unlike many of the reports to date where ceria catalyses such organic transformations, we found the encapsulated CeO 2 to play the key role of radical initiator due to the presence of Ce 3+ included in the structure, with the nanotube acting as both a host, preserving the high performance of the CeO 2 nanoparticles, anchored at the GNF step-edges, over multiple uses, and an electron reservoir, maintaining the balance of Ce 3+ and Ce 4+ centers. Spatial confinement effects ensure excellent stability and recyclability of CeO 2 @GNF nanoreactors. in (b), GNF are topologically complex nanoscale structures comprising a series of stacked cups encased within concentric tubes. The internal diameters are much larger than the dimensions of simple molecular species (typically by more than an order of magnitude) and the termini are always open, thus facilitating highly efficient transport of molecules through the internal volume. Furthermore, the 3-4 nm high folds on the interior surfaces of GNF formed by rolled-up sheets of graphene provide exemplary anchoring points for guests, such as molecules and nanoparticles, as shown in (b), due to maximised van der Waals interactions. This creates localised nanoscale reaction environments, different to the bulk phase, which impart spatial restrictions and thus can significantly impact the yields and distributions of products afforded across a range of preparative chemical transformations, as demonstrated by us and others. 37-48


INTRODUCTION
Lanthanide compounds continue to attract significant attention owing to their unusual magnetic, redox and optical properties, different to d-block elements, which is attributed to the nature of the 4f orbitals, buried deep inside the atom and thus shielded from the external environment by the overlying 5s and 5p orbitals. This contracted nature of the 4f orbitals gives lanthanides their unique physicochemical properties and underpins a wide range of applications in catalysis, 1 magnetism 2 and pharmacology. 3 Among these, catalysis alone represents nearly 75% of the total applications of the lanthanides. As a catalyst, cerium (Ce), unlike most of the other lanthanides, 4 is utilised in the form of its oxide (CeO2), also known as ceria, with a wide range of ceria-based nanocatalysts, including spheres, 5 rods, 6 tubes, 7 wires, 8 shuttle-shaped 9 and flower-like nanoparticles, 8 reported over the last decade. The remarkable catalytic behaviour of ceria observed ubiquitously throughout these nanostructures is due to the interconversion between Ce 4+ and Ce 3+ , controllable by the reaction conditions, with CeO2 and Ce2O3 observed under oxidising and reducing conditions, respectively. As a consequence of this interconversion, CeO2 can reversibly release oxygen from its structure at moderate temperatures (<600 o C) 10 leading to the formation of oxygen vacancies in the crystal lattice and simultaneous reduction of Ce 4+ to Ce 3+ . This behaviour makes ceria extremely useful for oxygen storage and release applications, 11 as well as in other important chemical reactions, such as the preparation of imines, 12 the oxidation of CO to CO2 13 and the hydrogenation of CO2 to methanol. 14 However, a less explored and often disregarded application of ceria is its ability to initiate radical-mediated organic transformations, which often commence in the homolytic cleavage of organic peroxides by Ce 3+ , as demonstrated in the oxidation of cyclohexene, 15 the products of which represent important intermediates in the chemical industry. 16 Akin to the structure-function relationships observed in ceria nanocatalysis, the shape and size of ceria nanoparticles are expected to modulate its efficiency as an initiator; 17,18 yet, further research is critically required to probe how the physical and chemical properties of CeO2 influence the yields and distribution of products of radical-based chemical reactions.
Despite the breadth of ceria-based nanostructures synthesised to date, the controllable preparation of monodisperse ultra-small CeO2 nanoparticles remains a significant challenge.
Moreover, preservation of their functional properties by preventing deactivation under harsh reaction conditions is essential for their further development. One recent approach to address the latter concerns the encapsulation of CeO2 nanoparticles in core-shell and yolk-shell structures, where CeO2 is protected by an outer shell, which was shown as an efficient way to stabilise CeO2 nanoparticles and explore the effect of spatial confinement at the nanoscale. 19,20 However, carbon nanotubes (CNTs) as a host matrix for the confinement of CeO2 nanoparticles have the potential to solve both of these challenges. Besides being a container with two dimensions on the nanoscale and thus the capacity to template the formation of particles with nanoscopic sizes, the remarkable mechanical (tensile strength much higher than that of steel), thermal (stable up to ca. 600 o C in air and up to 2800 o C in vacuum or inert atmosphere) and chemical stability of CNTs makes them an excellent candidate support material for nanoparticles under a wide variety of reaction conditions. 21 In addition, the curvature of the CNT walls causes re-distribution of the π electron density of graphene layers from the concave inner to the convex outer surface resulting in an electron density gradient. As a consequence, encapsulated metal nanoparticles inside carbon nanotubes often exhibit properties distinct from those adsorbed on the CNT outside surface. 22 Therefore, the encapsulation of metal or metal oxide nanoparticles inside CNTs opens a pathway towards active nanoreactors with tuneable properties, such as those reported by our group 23,24 and others previously. 25,26 5 While several examples of the successful deposition of CeO2 nanoparticles onto the outer surface of carbon nanotubes (CeO2-CNT) have been reported and applied in catalysis, 27,28 biosensing 29,30 and as a fuel additive, 31 the preparation of ceria nanoparticles inside carbon nanotubes remains largely unexplored, with only one example to date reported for CeO2 nanoparticles inside carbon nanotubes of inner diameter 4-8 nm. 32 Such narrow nanotubes serve as excellent model systems for catalyst supports, but their applications for preparative scale reactions are limited. In this study, we utilised graphitised carbon nanofibres (GNF) that have many of the attractive properties of carbon nanotubes, but with the added benefits of (i) large internal diameters (typically larger than 60 nm) enabling the efficient mass transport of reactants and products to and from the internal cavity and (ii) corrugated internal surfaces resulting in the effective stabilisation of nanoparticles, as demonstrated in our previous studies ( Figure 1). [33][34][35][36] Herein, successfully prepared ceria nanoparticles within GNF (CeO2@GNF) were analysed by a variety of structural and chemical characterisation techniques, and applied in the preparative-scale reaction of cyclohexene oxidation, revealing key features of nanoscale confinement in carbon nanoreactors important for sustainable organic synthesis. The internal diameters are much larger than the dimensions of simple molecular species (typically by more than an order of magnitude) and the termini are always open, thus facilitating highly efficient transport of molecules through the internal volume. Furthermore, the 3-4 nm high folds on the interior surfaces of GNF formed by rolled-up sheets of graphene provide exemplary anchoring points for guests, such as molecules and nanoparticles, as shown in (b), due to maximised van der Waals interactions. This creates localised nanoscale reaction environments, different to the bulk phase, which impart spatial restrictions and thus can significantly impact the yields and distributions of products afforded across a range of preparative chemical transformations, as demonstrated by us and others. [37][38][39][40][41][42][43][44][45][46][47][48] EXPERIMENTAL SECTION

General
All reagents and solvents were purchased from Sigma-Aldrich, including cyclohexene (>99%) and were used as received. Graphitised carbon nanofibres (GNF) and tetrakis(2,2,6,6-tetramethyl-3,5heptanedionato) cerium (IV) (99.9% Ce) were purchased from Pyrograf Products, USA and Strem Chemicals, USA, respectively. Graphite flakes (Code G/0900/60, Batch 043775) and activated carbon (Activated Charcoal, DARCO) were purchased from Fisher Scientific. All glassware was cleaned with a mixture of hydrochloric and nitric acid (3:1 v/v, "aqua regia") and rinsed thoroughly with deionised water, cleaned with potassium hydroxide in isopropyl alcohol and finally rinsed thoroughly with deionised water before use. Prior to use, GNF was ball milled using a Retsch  Graphitised carbon nanofibres (10 mg, PR19-XT-HHT, ball-milled) and Ce(tmhd)4 (1, 2 or 3 mg) were sealed in a Pyrex glass ampoule under vacuum (2.9 x 10 -5 mbar) and heated at 150 o C in an oil bath for 72 hours to allow for sublimation of the cerium complex and diffusion inside the channels of GNF. After 72 hours, the ampoule was quickly cooled by placing into ice cold water for 10 minutes before opening. The obtained bulk solid was then placed in another Pyrex glass ampoule, evacuated at 1.6 x 10 -2 mbar and backfilled with argon (this procedure was repeated three times to ensure removal of molecular oxygen and atmospheric moisture), and the ampoule was then sealed and heated at 600 o C for 2 hours to decompose the cerium complex to CeO2, followed by slow cooling for 8 hours to a final temperature of 20 o C to yield the resultant CeO2@GNF-X material (where X = 1, 2 or 3 corresponding to the mass of cerium complex used in the initial step).
The CeO2@graphite and CeO2@AC nanocomposite was prepared using an analogous method with graphite powder or activated carbon (AC) and Ce(tmhd)4 in a mass ratio of 10:2 (mg), to serve as a reference for CeO2@GNF-2.

Characterisation of CeO2@GNF, CeO2@graphite and CeO2@AC
High resolution transmission electron microscopy (HRTEM) was performed using a JEOL 2100F transmission electron microscope (field emission electron gun source, information limit 0.19 nm).
Energy dispersive X-ray (EDX) analysis was performed using an Oxford Instruments INCA 560 X-ray microanalysis system. TEM samples were prepared by dispersing small quantities in methanol with bath sonication for 30 seconds to 1 minute, before drop-casting onto a copper TEM grid coated with a lacey carbon film (Agar Scientific UK) and drying in air. Particle size and dspacing were determined using Gatan Digital Micrograph software. Thermogravimetric analysis (TGA) of samples was carried out using a TA Q500 Thermogravimetric Analyser. All samples were deposited onto platinum pans and analysed in air at a heating rate of 10 o C/min from 20 to 8 1000 o C with an isotherm for 10 min at 1000 o C, and airflow of 60mL/min. Raman spectroscopy was performed using a Horiba-Jobin-Yvon LabRAM HR spectrometer. Spectra were acquired using a 532 nm laser (at 0.3 mW power) and a 600 lines mm -1 rotatable diffraction grating, conferring a spectral resolution of better than 1.8 cm -1 . Spectra were baseline corrected using a linear fitting model. The powder X-ray diffraction (PXRD) patterns were recorded using a PANalytical X'Pert Pro diffractometer equipped with a Cu K(α) radiation source (λ = 1.5432 Å, at 40kV and 40mA) in Bragg-Brentano geometry using a Si zero background holder. A few drops of isopropyl alcohol were added to all samples for good adhesion to the sample holder. The parameters for a typical measurement were as follows: start angle 5 o , stop angle 120 o , step size 0.0525 o , time/step 6080s, scan speed 0.002200/s. X-ray photoelectron spectroscopy (XPS) measurements were performed using a Kratos AXIS Ultra DLD instrument. The chamber pressure during the measurements was 5×10 −9 Torr. Wide energy range survey scans were collected at a pass energy of 80 eV in hybrid slot lens mode and a step size of 0.5 eV over 20 min. The charge neutraliser filament was used to prevent the sample from charging over the irradiated area. The Xray source was a monochromated Al Kα emission, run at 10 mA and 12 kV (120 W). The energy range for each 'pass energy' (resolution) was calibrated using the Kratos Cu 2p3/2, Ag 3d5/2 and Au 4f7/2 three-point calibration method. The transmission function was calibrated using a clean gold sample method for all lens modes and the Kratos transmission generator software within Vision II. The data were processed with CASAXPS (Version 2.3.17). The high resolution data was charge corrected to the reference C 1s signal at 284.5 eV.

Oxidation of cyclohexene
The oxidation of cyclohexene with tert-butylhydroperoxide (TBHP, 70 wt% in H2O) was carried using either CeO2@GNF, CeO2@graphite or CeO2@AC in a two-necked glass round-bottom flask 9 equipped with a magnetic stirrer and a reflux condenser open to air. In a typical reaction, a mixture of cyclohexene (2.9 mmol), TBHP (5.8 mmol), CeO2@Y (10 mg, where Y = GNF, graphite or AC, containing 0.09-0.18 mol % CeO2) and 1,4-dichlorobenzene (1.4 mmol) as an internal standard were dispersed in acetonitrile (2.5 mL) and stirred at 80 o C. The progress of the reaction was followed using 1 H NMR spectroscopy. To study the reaction kinetics, aliquots (0.01 mL) were taken from the reaction mixture at one hour intervals. Confirmation of the final products was determined using 1 H NMR spectroscopy. For recycling studies, the catalyst was recovered by filtration, washed with methanol and dried overnight before repeating the procedure for oxidation of cyclohexene. The same reaction conditions mentioned above with freshly loaded reactants. The recycling reaction was conducted five times for CeO2@GNF.
Prior to filling, graphitised carbon nanofibres were ball milled to reduce their length from more than 10 µm to around 1.2 µm to reduce the possible impact of length-dependent transport resistance of reactant molecules from the bulk liquid-phase when later applied as a nanoreactor. The  To ascertain the loading of ceria in the CeO2@GNF nanocomposites and to study the resultant GNF thermal stability, thermogravimetric analysis was conducted (Figure 2a). For all three CeO2@GNF samples, a single weight loss, representing ~92-95% of the total weight and corresponding to the combustion of GNF, was observed. The residual weight subsequent to GNF combustion (4.7, 5.9 and 7.7 %) is associated with CeO2 loading and correlates well with the initial amount of cerium precursor added to GNF (1, 2 and 3 mg, respectively). It is interesting to note that the GNF combustion temperature for CeO2@GNF is much lower (~625-675 o C) than that of empty GNF (~750 o C). This decreases with increasing CeO2 loading and indicates a catalytic effect of CeO2 on the oxidation of graphitic carbon in air. 49,50 This observation supports the notion that CeO2 is an effective catalyst of oxidation due to the reversible loss of oxygen and interconversion of Ce 4+ and Ce 3+ , thus facilitating the combustion of GNF at a lower temperature. 31 TGA profiles also indicate a uniform one-step oxidation process, which is consistent with a homogeneous composition of the total CeO2@GNF sample, i.e. negating the independent formation of CeO2 and GNF in isolation.
The Raman spectra of CeO2@GNF (Figure 2b) are dominated by two characteristic GNF bands: the G (graphitic) band -a carbon-carbon stretching vibration of E2g symmetry, typically observed in graphitic nanocarbons -at ~1580 cm -1 and the D (disorder) band -a ring-breathing mode of A1g symmetry, requiring a defect for its activation -at ~1340 cm -1 . 50 Interestingly, a small increase in the intensity ratio of the D and G bands (ID:IG) in the CeO2@GNF nanocomposites relative to GNF was observed, indicating a change in the structural ordering in the graphitic sidewalls of GNF, consistent with the shortening/opening and thermal treatments employed during the preparation of the nanoreactors. 48 In addition, a very weak band at ~450 cm -1 is observed in some of the Raman spectra of CeO2@GNF, consistent with the vibration of the Ce-O bond, 50  Although the TGA, Raman spectroscopy and PXRD results all suggest the formation of a nanocomposite of CeO2 and GNF, the location of CeO2, i.e. inside or outside GNF, could only be ascertained from transmission electron microscopy (TEM). Interestingly, TEM imaging reveals CeO2 nanoparticles are well dispersed inside the GNF in all samples (Figure 3). At the lowest amount of cerium precursor (1 mg in CeO2@GNF-1), effectively all CeO2 nanoparticles are observed to be embedded inside GNF (Figure 3a); however, at higher precursor loadings (3 mg in CeO2@GNF-3), a noticeable, but still minor, number of nanoparticles were found on the outer surface of GNF (Figure 3c and S2a and d, SI). This indicates that the best level of control over the location of nanoparticles is achieved with the smallest amount of the precursor complex.
Consistent with previous studies, CeO2 nanoparticles are found to be preferentially adsorbed at the internal step-edges of GNF (Figures 1, S2b and c, SI). Furthermore, mean diameters of ceria nanoparticles of 4.6±0.2, 6.2±0.6 and 8.5±0.3 nm were determined by TEM for CeO2@GNF-1, -2 and -3, respectively (insets in Figures 3a-c), the latter of which matching well with the size estimated from PXRD, clearly showing the positive correlation between nanoparticle size and precursor concentration. 54 High resolution TEM images (Figure 3e) (Figure 3d). However, more detailed analysis by XPS (Figure 2d) indicates the existence of an oxygen-depleted surface containing both Ce 4+ (78.8 at%) and Ce 3+ (21.2 at%) based on deconvolution of the Ce 3d signal ( Figure S3 and Tables S1-2, SI). 55,56 The presence of Ce 3+ is significant for our reaction, as Ce 3+ is known to be key for initiation of radical-mediated organic transformations. 15

CeO2@GNF in the oxidation of cyclohexene
The CeO2@GNF composites were tested as initiators in the oxidation of cyclohexene by tertbutylhydroperoxide. This specific reaction was chosen as it is known to proceed through two different routes, either via allylic or olefinic oxidation, leading to the formation of five main products [57][58][59][60] , with the ratio of these products influenced by the conditions of the reaction, including the oxidant, temperature, time, pressure and solvent. 61,62 Under our experimental conditions of 80 o C for 9 hours in acetonitrile, the control reaction (entry 1, Table 1) gave 18% conversion for cyclohexene with the formation of products via both allylic (2-cyclohexenylhydroperoxide (1)) and olefinic (cyclohexaneepoxide (4)) oxidation pathways in a 50:50 ratio of (1):(4) indicating low selectivity of this reaction. In the presence of empty GNF (entry 3, Table 1), the conversion increased to 45% with the formation of only allylic oxidation products (1), (2) and (3). The observed higher conversion is consistent with the known effect of decomposition of peroxy compounds, here TBHP, in the presence of graphitic carbon surfaces generating reactive oxygen species, i.e. hydroxyl radicals. 63,64 The key effect of the presence of ceria in CeO2@GNF, as compared to empty GNF, is a significant increase of conversion (up to 94%) with 2-cyclohexenone (2) and 2-cyclohexenyl hydroperoxide (1) as major products observed for CeO2@GNF-2 (entry 5, Table 1 and Figure 4a). Comparison of the CeO2@GNF nanoreactors with different loadings of ceria (CeO2@GNF-X where X=1,2,3) did not reveal significant differences in terms of cyclohexene conversion ( Figure S5 and Table S3, SI). Interestingly, we noted a clear dependence on the specific distribution of products as a function of reaction time using CeO2@GNF-2 (Figure   4c). At 7 hours, the yield of (1) is maximal at 89% selectivity (entry 4, Table 1), at a high conversion of 88%. As the reaction proceeds, (1) is consumed to produce both (2) and (3), with a conversion of 100% and moderate selectivity of 56% for (2) observed at 24 hours (entry 6, Table   1). This indicates that the preferred product ((1) or (2)) can in principle be obtained in high yield and moderate-to-good selectivity simply by stopping the reaction at a user-specified time point. Table 1. The effects of CeO2 loading, particle size, reaction time and location in GNF and on graphite and activated carbon on the conversion of cyclohexene and distribution of the afforded products: 2-cyclohexenyl hydroperoxide (1), 2-cyclohexenone (2), 2cyclohexenol (3), cyclohexane epoxide (4) and 1,2-cyclohexanediol (5).  It is important to note that the lack of products from the olefinic oxidation pathway differs significantly from previous reports on the oxidation of cyclohexene with TBHP using CeO2 nanoparticles, where a mixture of (2), (3) and (4) were identified as major products (Table S4,   SI). 65 In addition, Voort et al. 66 reported the formation of both (4) and (1) as products of cyclohexene oxidation by TBHP, suggesting the reaction proceeds through both allylic and olefinic oxidation pathways. Moreover, it should be highlighted that our nanoreactors CeO2@GNF have very small ceria loadings (0.13 mol%) (Table S4, SI) compared to previous reports (by more than two orders of magnitudes in some instances), with up to 95 wt% of CeO2 in CeO2/VO2 -more a stoichiometric reactant than a catalyst -required to bring about effective oxidation. 65 In order to study the stability, and thus recyclability, of CeO2@GNF, we have evaluated the cyclohexene conversion using CeO2@GNF over five uses. After each use, the catalyst was filtered and washed repeatedly with methanol and kept overnight at 65 o C in air. The conversion of cyclohexene was found to subtly increase from the first to the third use, with a very small decrease observed thereafter (Figure 4b). The increase in conversion after the first use can be attributed to the cleaning of the nanoparticle surface under the reaction conditions. 67 TGA profiles of fresh and recycled CeO2@GNF ( Figure S4, SI) shows no significant loss of CeO2 content even after five uses, indicating good stability of CeO2@GNF to leaching.

Entry
The enhanced cyclohexene conversion observed with CeO2@GNF, and its associated stability towards repeat use, can be attributed to the effects of spatial confinement of CeO2 inside GNF.
This has been shown to lead to: (i) well dispersed CeO2 nanoparticles of very small diameter (3-8 nm), protected at the graphitic step-edges from deactivation mechanisms, yet providing a large surface area for the adsorption of reactant molecules; (ii) increased local concentration of reactants inside nanoreactors promoting the effective interaction between reactant molecules and CeO2, thus improving conversion and (iii) easy diffusion of reactant and product molecules to and from the nanoreactor, while maintaining strong local interactions with the CeO2 nanoparticles at the GNF step-edges. These effects are consistent with the spatial confinement observed for other nanoparticles inside carbon nanoreactors, including GNF, that often result in highly selective organic transformations use-to-use. 25,26 The importance of confinement To validate the importance of step-edges and nanoscale confinement inside GNF on the functional properties of ceria, CeO2 nanoparticles were separately deposited onto activated carbon (AC) and graphite flakes ( Figure S6, SI). AC and graphite both provide anchoring sites for nanoparticles, akin to GNF, but no spatial restrictions as inside the GNF cavity. TEM analysis ( Figure S6c-g, SI) indicates that ceria nanoparticles on AC and graphite are uniformly distributed on both carbons and are qualitatively similar in size to those observed in CeO2@GNF. A preference for cubic morphology was noted for CeO2 on graphite. Interestingly, TGA ( Figure   S6a,b) shows clear differences in the combustion temperature of the carbon supports after deposition of ceria: a significant reduction (by more than 200 o C) was observed for AC; only a minimal difference was recorded using graphite. This suggests a higher intimacy of contact between CeO2 and AC, like that observed using GNF, relative to graphite where contact is poorer.
However, 1 H NMR spectroscopy analysis indicates lower conversion in the oxidation of cyclohexene using both CeO2@AC and CeO2@graphite relative to CeO2@GNF (entries 7, 8 and 5, respectively, Table 1). Moreover, whilst cyclohexene conversion is largely retained after the first use of CeO2@AC, this drops significantly when using CeO2@graphite (Figure 3b) indicating low stability, and thus recyclability, of ceria on this carbon support. Thus, whilst we cannot rule out additional factors that may contribute to the observed reactivity and stability of CeO2@GNF, these observations suggest the importance of nanoscale confinement inside GNF nanoreactors, providing a new way to enhance cyclohexene conversion, use-to-use, whilst simultaneously improving the selectivity for allylic oxidation at very low loadings.

Mechanistic considerations
Based on our results and previous literature concerning the TBHP-mediated oxidation in the presence of CeO2 and transition metals [68][69][70] , a plausible mechanism for the oxidation of cyclohexene can be proposed (Figure 4d). Since no epoxide is detected during the reaction, the formation of products through the selective oxidation of the allylic carbon centre by free radicals appears the most viable option. Our XPS measurement demonstrated that ~20% of cerium cations in CeO2@GNF are in oxidation state Ce 3+ . The redox reaction of Ce 3+ centres with t-BuOOH yields t-BuO . radicals, similar to the process described for other nanoscale materials. [71][72][73] The free radical t-BuO . extracts the allylic hydrogen from cyclohexene to form the cyclohexenyl radical, which becomes trapped by atmospheric O2 to form 2-cyclohexenyl peroxide radical. 74,75 The critical role of atmospheric oxygen was ascertained in control experiments, under anaerobic conditions, where negligible conversion was observed (entry 2, Table 1). The 2-cyclohexenyl peroxide radical subsequently abstracts hydrogen from the allylic CH2 of another molecule of cyclohexene to form 2-cyclohexenyl hydroperoxide (1), 76,77 the existence of which is confirmed by 1 H NMR spectroscopy. This homolytic abstraction of hydrogen is also energetically favourable: the bond energies of RO-H and the allylic C-H bond are 90 and 85 kcal mol -1 , respectively. 78 In effect, ceria nanoparticles act as an initiator of the free-radical allylic oxidation cycle which then continues producing (1) which in turn forms (2). The low amount of HO . radicals in this process explain the low yield of product (3), only observed after extensive heating, which forms without ceria in GNF on shorter timescales.

22
Analysis of CeO2@GNF after the reaction offers additional mechanistic insights. XPS shows no significant changes in the ratio between Ce 4+ (78.1%) and Ce 3+ (21.9%), but a shift of ca. 1 eV towards higher binding energy in all Ce 3d components. The shift is likely to be associated with the presence of hydroxyl groups on ceria after the reaction (bi-product of peroxide decomposition; Figure 3d), which corroborates with a large increase of the O 1s peak at ca. 532 eV associated to hydroxyl groups ( Figure S3d, SI). This suggests that GNF acts as an electron reservoir compensating for the loss of electrons in ceria, as observed previously for other metal compounds inside carbon nanotubes, 79 Author Contributions

24
The manuscript was written through the contributions of all authors. All authors have given approval to the final version of the manuscript.

Funding Sources
The work was supported by the University of Delhi, India, the Engineering and Physical Research Council (EPSRC) and the Centre for Sustainable Chemistry, University of Nottingham.

Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENT NA acknowledges the financial support provided by the University of Delhi, India to carry out the work at the University of Nottingham. We also acknowledge the Nanoscale Microscale