Synthesis and Properties of Open Fullerenes Encapsulating Ammonia and Methane

Abstract We describe the synthesis and characterisation of open fullerene (1) and its reduced form (2) in which CH4 and NH3 are encapsulated, respectively. The 1H NMR resonance of endohedral NH3 is broadened by scalar coupling to the quadrupolar 14 n nucleus, which relaxes rapidly. This broadening is absent for small satellite peaks, which are attributed to natural abundance 15N. The influence of the scalar relaxation mechanism on the linewidth of the 1H ammonia resonance is probed by variable temperature NMR. A rotational correlation time of τc=1.5 ps. is determined for endohedral NH3, and of τc=57±5 ps. for the open fullerene, indicating free rotation of the encapsulated molecule. IR spectroscopy of NH3@2 at 5 K identifies three vibrations of NH3 (ν 1, ν 3 and ν 4) redshifted in comparison with free NH3, and temperature dependence of the IR peak intensity indicates the presence of a large number of excited translational/ rotational states. Variable temperature 1H NMR spectra indicate that endohedral CH4 is also able to rotate freely at 223 K, on the NMR timescale. Inelastic neutron scattering (INS) spectra of CH4@1 show both rotational and translational modes of CH4. Energy of the first excited rotational state (J=1) of CH4@1 is significantly lower than that of free CH4.


Introduction
The potentialf or encapsulationo fa na tom or small molecule within the cavity of spherical fullerenes, has long been recognized. [1] The inert and highly symmetric, three-dimensionale nvironmentoft he cavity,means that enclosed (endohedral) species are expectedt ob ehave much as they would in the very low pressure gas state, with preservation of free rotationd own to cryogenic temperatures. [2] Although direct synthesis of endohedral metallofullerenes, [3] and fullerenes containing individual atoms (nobleg as@C 60 [4] and the remarkable N@C 60 [5] )i s possible in very low yield, currently the only high-yielding route to small molecule endofullerenes is via the process of "moleculars urgery" [6] wherebyc hemical transformationsa re used to open ah ole in the fullerene, am olecule is inserted, and af urthers eries of reactions is then used to suture the openinga nd reform the pristine fullerenes hell. To date, this has only been achieved for the incorporation of H 2 , [7] H 2 O [8] and HF, [9] as well as their related isotopologues. Encapsulated H 2 and H 2 Oh ave provenp articularly interesting due to the interaction of their nuclear spin and rotational states. Due to the Pauli principle, the ortho-( nuclear spins aligned) and para-( nuclear spins opposed)a llotropes are limited to odd and even rotational states, respectively.F or H 2 O@C 60 ,i solation contributes to long lived ortho-a nd paraspin states and has allowed their interconversion and physical phenomenon such as spin dependente lectric polarizabilityt o be measured. [10] Ar ange of larger molecules including N 2 , [11][12][13] O 2 , [14] CO, [13,15] NH 3 , [16] CO 2 , [11] CH 4 , [17] CH 2 O, [18,19] CH 3 OH, [18] HCN, [19,20] and HCCH [20] have been incorporated into open fullerenes via a larger openingi nt he shell, but closure to reform the pristine fullerenec age has not yet been achieved for these examples. Althought he high symmetry of the closed cage is lost in these open derivatives, many other properties including isolation are retained. [21] Encapsulated NH 3 and CH 4 are of particulari nterest because (as with H 2 Oa nd H 2 )t heir symmetry leads to interaction between the rotational( J)a nd nuclear spin (I)q uantum states. Methane exists as three nuclear spin isomers (with overall spin We describe the synthesis and characterisation of open fullerene (1)a nd its reducedf orm (2)i nw hich CH 4 and NH 3 are encapsulated, respectively.T he 1 HNMR resonance of endohedral NH 3 is broadened by scalar couplingt ot he quadrupolar 14 N nucleus, which relaxesr apidly.T his broadening is absent for small satellite peaks, whicha re attributed to natural abundance 15 N. Thei nfluence of the scalar relaxation mechanism on the linewidth of the 1 Ha mmonia resonancei sp robed by variable temperature NMR. Ar otational correlation time of t c = 1.5 ps. is determined for endohedral NH 3 ,a nd of t c = 57 AE 5ps. for the open fullerene, indicating free rotation of the encapsu-lated molecule. IR spectroscopy of NH 3 @2 at 5K identifies three vibrations of NH 3 (n 1 , n 3 and n 4 )r edshifted in comparison with free NH 3 ,a nd temperature dependence of the IR peak intensityi ndicates the presence of al arge number of excited translational/ rotationals tates. Variable temperature 1 HNMR spectra indicate that endohedral CH 4 is also able to rotate freely at 223 K, on the NMR timescale. Inelastic neutron scattering (INS) spectra of CH 4 @1 show both rotational and translational modes of CH 4 .E nergy of the first excited rotational state (J = 1) of CH 4 @1 is significantly lower than that of free CH 4 .

Synthesis of CH 4 @O pen Fullerene
Heating the open-cage fullerene 1 [12] at 200 8Cf or 68 h, under 153 atm of methaneg ave CH 4 @1 with 65 %e ncapsulationo f the endohedral molecule in quantitative yield (Scheme 1). The filling factor of CH 4 @1 was established by comparison of integrated resonances for CH 4 and an exohedral alkene proton in the 1 HNMR spectrum.F or comparison, CH 4 @3 wasp repared by Iwamatsu and co-workers in 20 %y ield and 39 %i ncorporation, after 20 hat200 8Cunder 190 atm of CH 4 . [17] The rate of loss of CH 4 from CH 4 @1 was measured at several temperatures between 428 and 448 Ka nd displayed the expected 1 st orderk inetics. From the linear Arrhenius and Eyring plots (Figure 2), an activation energy of 134.6 AE 5.0 kJ mol À1 and pre-exponential factor log(A) of 10.9 wasd etermined. The latter is low for au nimolecular reaction, but comparable for those previouslyo bserved for loss of endohedral atoms and molecules from endohedralf ullerenes. [13] The enthalpy and entropy of activation for CH 4 loss were determinedt ob eDH°= 131.0 AE 5.0kJmol À1 and DS°= À47.0 AE 11.2 JK À1 mol À1 giving a DG°at 165 8Co f1 51.5 AE 0.1 kJ mol À1 .T he values are in good agreement with those calculated by DensityF unctionalT heory methods( DH°= 135.9 kJ mol À1 ; DS°= À31.5 JK À1 mol À1 ; DG°= 149.7 kJ mol À1 at 165 8C). The negative entropyo fa ctivation is unusualf or ad issociative reaction, but reflects the loss of rotational and translational degrees of freedom as the endohedral molecule is constrained by its passage through the orifice.

Physical Properties of CH 4 @Open Fullerene
The 1 Hr esonanceo fe ndohedral CH 4 in CH 4 @1 appearsa sa sharp singlet at d = À12.59 ppm (500 MHz, 1,2-dichlorobenzene-d 4 )o rÀ12.33 ppm (500 MHz, CDCl 3 ), compared with d = 2.17 ppm for gaseous CH 4 . [22] The 1 Hs pectrum was acquired with ap ulse delay (d1) of 30 s. followingm easurement of the experimental 1 Hs pin-lattice relaxation curve for CH 4 @1 which shows single exponential decay with a time constant of T 1 = 7.53 AE 0.03 s. (see Supporting Information). Variable temperature proton NMR showedn ol ine broadening down to 223 K ( Figure 3) indicating free rotationo ft he endohedralm ethane on the NMR timescale. Te mperature-dependence of the 1 H chemicals hift is unexplained, but may be due to one or more configurations of different energy,i nf ast exchange on the NMR timescale.
Although CH 4 @1 is stable to loss of CH 4 at room temperature, the "empty" open-cage fullerene component 1 (i.e. 35 % of the material) was found to readily encapsulate water upon exposure to the atmosphere.R apid exchange of aw ater molecule between the endo-a nd exohedral environments has been reported by Murata and co-workers [12] and is characterized by av ery broad 1 Hr esonance which we detect at d = À11.62 ppm in 1,2-dichlorobenzene-d 4 .D ry samples of the inseparable 65:35 mixture of CH 4 @1:1 were most readily obtainedb yr emoval of water as its azeotrope with THF. The CH 4 molecule is confined within the small space defined by the fullerene cage, so in addition to rotationald egrees of freedom, the molecule also exhibits quantised translation. Analogously to INS investigations on other small molecule endofullerenes such as H 2 @C 60 and H 2 O@C 60 , [24] it is expected that the INS spectra of CH 4 @1 will comprise peaks arising from transitionsa mong the quantised rotational and translational eigenstates of CH 4 confined in its cage. However, unlike H 2 @C 60 and H 2 O@C 60 ,t he open-cage environmento fC H 4 @1 means the cage potential experienced by the CH 4 rotor lacks symmetry and is anisotropic. [25]   As aq uantum rotor possessing four indistinguishable 1 H nuclei (fermions), the permissible eigenstates of CH 4 are classified by the Pauli Principle andw ei dentify nuclear spin isomers for whichr otational and nuclear spin eigenstates are entangled. In the free-rotorl imit (as applies to molecule of CH 4 in the gas phase), the ground rotational state with rotational quantum number J = 0h as A 1 symmetry and total nuclear spin I = 2. Thef irst rotational excited J = 1s tate has F 1 symmetry and I = 1, with energy 1.29 meV.T he second excited state with J = 2h as energy 3.89 meV and comprises ad egenerate pair of states, one with F 2 symmetry and I = 1, the second with Es ymmetry and I = 0. The J = 3r otational state has energy 7.80 meV and comprises at rio of degenerate states with I = 1a nd I = 2. [26] While INS peaks of CH 4 @1 are observed in the complementaryenergy range below 10 meV (Figures 4-6), it is evident that they do not conform to those of af ree-rotor. Notably,t he 0.4 meV excitation observed on IN6 ( Figure 4) is much lower than the J = 1l evel of the free-rotor.I ndeed by analogy with investigations of hindered CH 4 rotors, [26,27] where the energy of the J = 1s tate is significantly reduced from its free-rotorv alue, we may assign the pair of j DE j = 0.4 meV peaks to the J = 0t o 1r otational transition. The peaks at DE = 2.95, 4.65 and 6.85 meV (Figures 5a nd 6) are consistent with examples in the literatureo fJ = 0t o2 ,3 transitions for hindered CH 4 . [26] However,afull assignment of the highere nergy rotational peaks is beyondt he scope of this preliminaryr eport and will be the subjecto fafuture publication. Nevertheless, the pattern of low energy peaks appearst ob es imilart ot hat found for CH 4 in other hindered environments. [26,27] The ground to first excited state translationale xcitations of H 2 ,H 2 Oa nd HF confined in C 60 are observed in the range 9 DE 25 meV. [9a, 24d, 28] The precise values depend sensitively on the rotor mass and the cage potential. Additionally,g iven the latter is anisotropic for this open endofullerene, the degeneracy of the translational states is lifted andw ee xpect to observe three translational peaks with different energy in the INS spectrum. We tentatively assign the broad band centred on 15 meV to these non-degenerate translationsa nd superimposed are some of the higher energy rotational states.
The anisotropy of the cage potentiall ifts the rotational degeneracy, [25] leadingt oasplitting of the respective mJ substates. Wef ind evidence for this in the IN6 spectrum (Figure 4), where the inelastic peaks are significantly broader than the resolutionf unction, indicating unresolved fine structure. That these peaks are inhomogenous and possess multiple components is indicated by the significant asymmetry of the NE peak at the lowest temperature 1.5 K, comparedw ith the more symmetrics hapeo bserveda th igher temperature. As determined by Boltzmann statistics, at 1.5 Kd ifferent components on the high and low energy side of the 0.4 meV NE gain peak will differ in amplitude by af actor of order 3. This theoretical factor is consistent with the observed NE gain peak shape. At the higher temperatures theseu nresolved components have more equilibrated Boltzmann factors, as observed. Therefore, we tentatively assign the excess width of the J = 0t o1r otational peaks to rotational fine structure.
On the timescale of the INS experiments (hours) we did not notice any significant changes which might be attributable to ac hange in the population of the J = 0( m-CH 4 )a nd J = 1( o-CH 4 )s tates, following thermale quilibration. It hasb een reported that conversionb etween these states for CH 4 in an argon matrix, or in the interstices of C 60 has t 1 / 2 % 1.5-2.5 h. [26,29]

Synthesis of NH 3 @Open Fullerene
DFT calculations gave the activation free energy for entry of ammonia into open fullerene 1 at STP as 62.3 kJ mol À1 ,h igher than that for water (30.7 kJ mol À1 )b ut indicating that both will enter rapidly at room temperature nonetheless. The free energy of binding of ammonia in 1 was calculatedt ob e  25 kJ mol À1 more favorable than that for water.W ew ere pleasedt of ind that exposure of as ample of fullerene 1 in CDCl 3 to a1 6% aqueous ammonia solution led to an 85:15 molar ratio of NH 3 @1 and H 2 O@1 by 1 HNMR,the spectrumd isplaying broad peaks at À12.44 and À11.52 ppm, respectively, indicating selectivee ncapsulation of ammonia in accordw ith the calculation of binding energies. In ordert oa void contamination with H 2 O-containings pecies, we switched to am ethanolic solution of ammonia as methanoli st oo large to enter 1 at room temperature. [18] To our delight, rapid formation of NH 3 @1 was observed by NMR although attempts to isolate NH 3 @1 gave only the empty open fullerene(1). In contrast, isolation of NH 3 @3 (the only previously reported NH 3 @openfullerene species)w as achieved by columnc hromatography with ammonia loss occurring only after several monthsa tÀ10 8C. [16] Our observation of the instability of NH 3 @1 to rapid ammonia loss at room temperature is in accord with Murata's conclusion [11] that 1 behaves as if it has al arger orifice than 3,despite the cage-opening being of nominally the same size, that is, a 17-member ring.
The selectiver eduction of ac arbonyl group on the orifice of 1 to afford 2 has been used to block the escapeo fO 2 ,N 2 and CO 2 guests from the fullerene cage. [11,14] We therefore sought to develop conditions for the encapsulationo fa mmonia by host fullerene 1,w ith subsequentt rapping of the endohedral NH 3 by in situ reduction of ar im carbonyl group to afford NH 3 @2.
Treatmento fasolution of 1 in 1,2-dichlorobenzene with 10 equiv of a7n solution of NH 3 in methanola t08C, followed by reduction with NaBH 4 ,a fforded NH 3 @2 in 45 %y ield with 92 %e ncapsulationo fN H 3 (Scheme 2). The filling factor was calculated by comparison of the integrated 1 Hr esonanceo f the endohedral molecule, with that of an exohedral alkene proton. We found that the selectiver eduction step worked better in the 1,2-dichlorobenzenes olventr eported by Murata, [11] than in chloroform. The ammonia encapsulation step was found to be sensitivet ot he periodo fe xposure to NH 3 ,w ith an optimal reaction time of 10 min. Lower NH 3 encapsulation resultsf rom as horter period (5 min. or 1min. exposure to NH 3 gives 80 %o r3 0% filling, respectively), but with ar eaction time > 10 min. we observed the formation of multiple fullerened erivatives by 1 HNMR. It is probable that these are hemiaminal by-products since treatment with 1 m HCl (aq.) returnst he mixture cleanly to as ingle compound whose 1 HNMR spectrum matches that of 1.T he formation of hemiaminal products in competition with NH 3 encapsulation is likely to limit the filling by "blocking" the fullerene orifice but, importantly,t he use of 1 m HCl( aq.) to quench the two-step (encapsulation/reduction)p rocedure allows ac lean mixture of NH 3 @2 and starting material 1 to be obtained.U nsurprisingly the encapsulation/reduction was found to be somewhat capricious, giving filling factorsf or NH 3 @2 in the range 74-92 %u nder nominally identical reaction conditions. Pure NH 3 @2 (74-92% filled) was readily obtained by column chromatography.

Physical Properties of NH 3 @O pen Fullerene
The NH 3 proton resonance of NH 3 @2 was observed experimentally as ab road peak at d = À12.35 ppm (500 MHz, [D 2 ]dichloromethane) with a3 8.43 Hz linewidth (Figure 7). The spectralw ings are attributed to natural abundance 15 NH 3 @2 (approx.0 .3 %i ntensity), shifted in frequency by as econdary isotope effect (1.5 ppb) and separated by j J 15NH j = 59.8 Hz, as was confirmed by independents ynthesis of 15 NH 3 @2 according to the method described above.
The broad proton peak of 14 NH 3 ,and the very narrow proton peak of the 15 NH 3 isotopologue, providesc ompelling evidence Scheme2.Synthesis of NH 3 @2. that the line broadening in the 14 NH 3 case is due to rapid quadrupolarr elaxation of the 14 Nc oupling partner.T his broadens the 14 NH 3 protonr esonance through the scalar relaxation of the second kind (SR2K)m echanism. [30] This broadening mechanism is negligible for the 15 Ni sotopologue, since the 15 N spin-lattice relaxation is on am uch longer timescale, due to the absence of an efficient quadrupolar relaxation mechanism for 15 N.
The 14 Nq uadrupolarr elaxationm ay be treated by assuming isotropic rotational diffusion. We define the quadrupolarc oupling constant by [Eq. (1)]: where I = 1f or 14 N, eQ is the electric quadrupolarm omento f the nitrogen nucleus, and eq is the electricalf ield gradient at the deuterium nucleus. [33] The Frobenius norm of the quadrupole couplingt ensor may be written as [Eq. (2)]: where h is the biaxality (asymmetry) parameter of the electric field gradient tensor.T he spin-lattice relaxation rate constant for quadrupolar relaxation, assumingi sotropic rotational diffusion and the extreme narrowing limit, is given by [Eq. (3)]: [34] T À1 1 ¼ which, for I = 1, is equal to [Eq. (4)]: The nuclear quadrupolec oupling constant for 14 NH 3 has been estimated by microwave spectroscopy to be: w Q /2p = 2.05 MHz, [35] with ab iaxality parameter of h = 0. [36] The 14 N T 1 value of 2.66 ms, as inferred from the lineshape of the 1 HNMR spectrum, leads to an estimate of the rotationalc orrelation time for endohedrala mmonia molecules of t c = 1.5 ps. If the overall tumbling of the fullerenec age is at least one ordero f magnitude slower as expected, [37] this would indicate that NH 3 is (essentially) rotatingf reely.W et herefore calculated the rotational correlation time (t c )o fopen fullerene 2 from the experimental relaxationt ime constant T 1 ( 13 C) = 0.54 AE 0.06 s. of the methine carbon located on the orifice of 2,u sing Equation (5) to define relaxationo ft he 13 C(H)OHn ucleus, as applicable for extreme-narrowing isotropic rotational tumbling, dominated by the 13 C-1 Hd ipolar relaxation mechanism: where w CH is the dipole-dipole coupling constant for the interaction between the carbon and protonn uclei, and t c is the overall rotationalc orrelation time for 2. By assuming an internuclear 13 C-1 Hd istance of 108.9 pm, which corresponds to ad ipole coupling constant of w CH /2p = À23.4 kHz, we obtain an estimate of t c = 57 AE 5ps. for the open fullerene 2,w hich is more than an order of magnitude longert han the rotational correlation time for the endohedral ammonia molecule in NH 3 @2.
Variable temperature solution protonN MR on NH 3 @2 showedt hat the NH 3 line broadens as the temperature is increased ( Figure 8). The increase in linewidth with temperature is consistentwith the SR2K mechanism:atemperature increase leads to as horter rotational correlation time t c ,w hich leads to slower quadrupolar relaxation for the 14 Nn ucleus, according to Equation (4). This leads in turn to al ess effective averaging of the J NH splittings,and hence to ab roader protonpeak.
The experimental 1 Hs pin-lattice relaxation curve forN H 3 @2 shows single exponential decay with at ime constant of T 1 = 5.05 AE 0.02 s, (see Supporting Information for all spin-lattice relaxation curves).
We measured the IR spectra of endofullerene NH 3 @2 and the empty open-cage species 2 between 600 and 9000 cm À1 in the temperature range 5K to 300 K. The spectra of NH 3 @2 show clear peaks present only in NH 3 @2 and not in 2 in three spectralr egions, around 1604 cm À1 ,3 300 cm À1 ,a nd 4700 cm À1 (Figures 9a nd 10). The % 1604 cm À1 region contains ac luster of four well-resolvedp eaks, each of which may be fitted well with aG aussians hape (Figure 9). The other two spectral regions contain peaks at 3196, 3288a nd 3380 cm À1 (Figure 10 a), and 4430 and 4970 cm À1 (Figure 10 b). It shouldb en oted that the spectralr egionsb etween 900-950, 1550-1950 and 3400-3550 cm À1 are obscured by strong fullerene absorption.
The intensity of all peaks decreases rapidlya st he temperature is increased above % 10 K ( Figure 11), with ap articularly strongd ecrease with temperature observed for the % 1604 cm À1 peak cluster.H owever,t he members of this peak cluster alwaysdisplay the same relative intensities.
In the gas phase, NH 3 has two vibrations with hydrogen atoms moving in the triangular plane (n 2 and n 4 ), and two vibrations with out-of-plane motions of hydrogen atoms n 1 and n 3 (Table 1). [38] The n 1 and n 2 vibrations are non-degenerate, while the n 3 and n 4 vibrations are doubly degenerate in the gas phase. The following tentativea ssignmentso ft he NH 3 absorptionp eaks in NH 3 @2 may be made;t he cluster of peaks at % 1604 cm À1 is assigned to the n 4 vibration, redshifted by % 24 cm À1 with respectt ot he same mode in NH 3 gas. The 3196 cm À1 peak is assigned to the fundamental n 1 vibration, redshifted by 141 cm À1 with respect to the same mode in NH 3 gas. The 3288 cm À1 and 3380 cm À1 peaks are assigned to the n 3 vibration. We postulate that the doubled egeneracy of the n 3 mode is lifted by the asymmetric environment of the open cage. The mean frequency of these two peaks is 3334cm À1 ,i ndicatingaredshift of 109 cm À1 with respectt ot he same mode in NH 3 gas. The peak at 4430 cm À1 is difficultt oa ttribute to a knownm ode or feasible combination mode of NH 3 alone,a nd we tentatively assign it to ac ombinationm ode involving coupled vibrations of the endohedralm olecule and am ode of the enclosing cage. The high-frequency peak at 4970 cm À1 is tentatively attributed to ac ombination mode, involvingt he % 1604 cm À1 n 4 vibration and the 3380 cm À1 component of the n 3 vibration.
Gaseous NH 3 has a n 2 vibrational mode at 950 cm À1 .Noanalogousp eak is observedi nt he spectrum of NH 3 @2.T he absence of this peak is probablyd ue to the strong fullerene absorptionint his spectral region.
No obvious fine structure is observed on the IR peaks, with the exception of the % 1604 cm À1 peak cluster.T he absence of rotational fine structure indicates that the potential generated by the confining asymmetric cage quenches the rotational Figure 9. IR spectra of NH 3 @2 recorded between5and 250 K, in the regions around 1600 cm À1 . Figure 10. IR spectra of NH 3 @2 recorded between5and 250 K, in the regions around( a) 3300 cm À1 and (b) 4700 cm À1 . Figure 11. Te mperature dependence of NH 3 @2 normalized IR absorptionl ine areas. Peaks are labeled by their frequencies. The normalized area of the 1604 cm À1 peak is the normalized sum of four peak areas from Figure9. Table 1. Normal modes of NH 3 ,t heir irreducible representations in point group C3n,f requencies in the gas phase and measured frequencies in NH 3 @2.

Mode
Gi (C 3 n) w gas (cm À1 ) [39,40] w NH 3 @2 (cm À1 ) (w NH 3 @2Àw gas)/ w gas The strong decrease in peak intensities with increasing temperature( Figure 11)i mplies the existence of many levels above the ground state which become thermally populateda tthe expense of the ground state population. These higherl evelsp resumably have ac omplexr otational/translational fine structure. As ar esult, the spectrala bsorptiona te levated temperature is distributed over av ery large number of unresolved spectral peaks, so that all identifiables pectral features disappear at high temperature.
An exception to the absence of fine structure is the peak cluster at % 1604 cm À1 ,w hichi st entatively assignedt ot he n 4 vibration (Figure 9). Since the relative intensities of the sharp components appear to be temperature-independent, the fine structure must be due to as plitting of the excited state energy levels,w ith the ground state remaining unsplit.T he origin of this splitting is unknown, but might be due to rotational fine structure, or at unnelling process between two or more potential minima. However,i ti su nclear why such structure is not clearlyd isplayed for all of the other peaks and, at present,w ed on ot have ad efinitive explanation for the fine structure around % 1604 cm À1 .

Conclusions
We have prepared and characterizedo penf ullerenese ncapsulating ammonia and methane. The encapsulatedm ethane and ammonia display long 1 Hs pin-lattice relaxation times at room temperature, of 7.5 and 5.1 s. respectively.The variable temperature 1 HNMR spectra indicate that both endohedralm olecules rotate freely within the cages at 223 K, on the NMR timescale.
In the INS spectra of CH 4 @1 we find the first rotational peak at 0.4 meV and translational energy at approximately 15 meV for CH 4 .T his is in qualitative agreement with CH 4 entrapped in interstitial sites of the C 60 crystal, [26] where the J = 0t o1rotational transition is centred on 0.6 meV,a nd the translational peak is centred on 10.9 meV.T he differences are indicative of a strongerc rystal field potential, and stronger confinement of the methane in CH 4 @1.W ed id not observe properties resulting from the entanglement of nuclear spin and molecule rotation/ vibration. The broad NH 3 resonance in the 1 HNMR spectrum of NH 3 @2 is associated with the quadrupolar relaxation of the 14 Nn ucleus, and is interpreted in terms of the 14 Ns pin lattice relaxation time: T 1 ( 14 N) = 2.66 ms. The scalar relaxation of the second kind mechanism is verified by broadening of the 1 Hl inewidth with increasing temperature. Ar otational correlation timeo f t c = 1.5 ps. is estimated for endohedralN H 3 .A ne xperimentally determined rotational correlation time of t c = 57 AE 5ps. for the open fullerene 2 confirms free rotation of the encapsulated molecule.
The difference IR spectrum of NH 3 @2 and 2 at 5K,i dentified three vibrations of NH 3 (n 1 , n 3 and n 4 )r edshifted in comparison with free NH 3 .Arapid decrease in IR peak intensityw ith increasingt emperature indicatest he presenceo falarge number of excited translational/ rotational states which are populated above 50 K. The structures of these states is complex, so that no resolved IR peaks are observed. Nevertheless, rotational freedomi sp ossible, in accordance with NMR observations.

Experimental Section
Synthesis and Characterization of Open-Cage Fullerene (OCF) Derivatives Reactions requiring dry conditions were conducted under an argon atmosphere using standard Schlenk and syringe techniques with freshly distilled solvents. All apparatus was dried in ah ot in oven (> 140 8C, 12 h) before being cooled in as ealed dessicator over silica gel or assembled while hot and cooled under vacuum (0.1 mm Hg). 1,2-Dichlorobenzene was distilled from CaH 2 at 55 8C under av acuum of 15 mm Hg. Ethanol was dried over 3 molecular sieves. All other reagents, solvents or gases were used as received from commercial suppliers. High-pressure reactions were conducted in aP arr pressure vessel of 75 mL volume and 1i nch i.d.,s ealed with PTFE or graphite gasket. High-pressure reactions were heated using an external oil bath and temperature monitoring was conducted using an external thermostat. NMR spectra were recorded on Bruker AVII400, AVIIIHD400 or AVIIIHD500 FT-NMR spectrometers in the indicated solvent at 298 K. 1 Hc hemical shifts are reported as values in ppm referenced to residual solvent. Spectra collected in 1,2-dichlorobenzene-d 4 are referenced to residual solvent at d H = 7.19 ppm, d H = 6.94 ppm;t his solvent assignment is referenced to TMS (d H = 0ppm). The following abbreviations are used to assign multiplicity and may be compounded:s= singlet, d = doublet, t = triplet, q = quartet and m = multiplet. Coupling constants, J,a re measured in Hertz (Hz). 13 Cs pectra are proton decoupled and referenced to solvent. 13 Cc hemical shifts are reported to 2d .p. in order to distinguish closely neighbouring resonances. Low-resolution mass spectra were recorded using a MaXis mass spectrometer (Bruker Daltronics) equipped with Time of Flight (t.o.f.) analyzer using positive electrospray ionization. Samples were infused via as yringe driver at ac onstant flow rate of 3 mLmin À1 .H igh-resolution mass spectra were obtained using as o-lariX FT-ICR mass spectrometer equipped with a4.7T superconducting magnet, using positive electrospray ionization. Values of m/z are reported in atomic mass units.

Open Fullerene 1
Open fullerene 1 was prepared according to the method of Murata, [12] our procedure differing only in purification which was carried out by column chromatography over SiO 2 eluted with a gradient of 2% ! 5% EtOAc in toluene. Spectroscopic data were consistent with the published data.

CH 4 @1
AP arr pressure vessel equipped with glass reactor insert was charged with as olution of open fullerene 1 (305 mg, 0.27 mmol) in 1-chloronaphthalene (15 mL of ! 85 %t echnical grade containing % 10 %2 -chloronaphthalene). The reactor vessel was sealed and flushed with CH 4 before charging with CH 4 to 101 atm at room temperature. The vessel was heated to 200 8C( external oil bath temperature) and stirred at this temperature with an internal pressure of 153 atm for 68 h, then cooled to room temperature and the pressure slowly released. The residue was diluted with 1-chlor-

NH 3 @2
Open fullerene 1 (26 mg, 0.023 mmol) was dried at 140 8C( external oil bath temperature) for 2h under av acuum of 0.3 mm Hg, before cooling under argon and addition of degassed 1,2-dichlorobenzene (4 mL). The solution was cooled to 0 8C( ice/salt bath) and NH 3 (33 mLo fa7 n solution in methanol, 0.23 mmol) was added drop-wise. The resulting mixture was stirred at 0 8Cf or 10 min. before addition of NaBH 4 (0.2 mL of af reshly prepared 58 mm solution in EtOH, 0.011mmol) and stirring at 0 8Cf or 15 min. further. 1 m HCl (2 mL) was then added and the cooling bath removed. After warming to room temperature, the mixture was stirred overnight before separation of the organic phase and extraction of the aqueous phase with 1,2-dichlorobenzene (1 mL). The combined organic extracts were filtered through as hort SiO 2 column with CHCl 3 (elutes 1,2-dichlorobenzene near the solvent front) followed by EtOAc (elutes NH 3 @2 near the solvent front). The EtOAc filtrate was concentrated in vacuo. Purification by column chromatography (SiO 2 eluted with 94:4:2 toluene:EtOAc:AcOH) gave the title compound as ab rown/black solid ( Data for the minor component of the mixed 1 Hs pectrum in 1,2-dichlorobenzene-d 4 ("empty" open fullerene 2)h as spectroscopic data consistent with that published by Murata et al. [14] Measurement of 1 Hand 13 CS pin-Lattice Relaxation Experimental 1 Hs pin-lattice relaxation curves were measured for CH 4 @1 (17.4 mm)a nd NH 3 @2 (17.6 mm)i nd egassed 1,2-dichlorobenzene-d 4 ,a nd the experimental 13 Cs pin-lattice relaxation curve was measured for the C(H)OH methine carbon on the orifice of NH 3 @2 (7.3 mm)i nd egassed CDCl 3 .A ll spectra were acquired at 11.7 Tand 25 8C, using aB ruker AVIIIHD500 FT-NMR spectrometer. Spin-lattice relaxation times T 1 were estimated using the saturation-recovery pulse sequence. The 908 pulse was calibrated for each sample using Bruker To pSpin and the saturation-recovery sequence employed a1 6-point delay list (0.01, 0.05, 0.1, 0.2, 0.3, 0.5, 0.75, 1, 1.5, 2, 2.5, 5, 7.5, 10, 15, 30 s.). Signals of interest from the saturation-recovery experiments were integrated using Bruker To p-Spin, and the data were fitted using Mathematica. Signal amplitudes was normalized to the last data point. The fitted curves have asingle-exponential form.

IR Spectroscopy
Samples of NH 3 @2 and H 2 O@2 with filling factors of f = 0.9 (NH 3 @2)a nd 0.35 (H 2 O@2)w ere studied. The fraction of empty cages is 1Àf. The sample of NH 3 @2 contained as mall amount of H 2 O@2 so the resonance lines specific to NH 3 were identified by comparing the NH 3 @2 and H 2 O@2 spectra. Some spectral regions were opaque because of the absorption by the open fullerene. The samples were pressed into pellets of 3mmd iameter and of thickness d = 155 mm( NH 3 @2) and 190 mm( H 2 O@2).S pectra were recorded with aV ertex 80v (Bruker Optics) spectrometer between 600 and 9000 cm À1 with al iquid nitrogen cooled HgCdTed etector. Sample temperature was controlled between 5a nd 300 Kw ith a continuous flow cryostat. Transmitted intensity through the sample, I s ,w as referenced to the intensity through a3mm diameter hole, I 0 .T he absorption coefficient a was calculated from the ratio Tr = I s /I 0 as a = Àd À1 ln[Tr(1ÀR) À2 ]w here the factor (1ÀR ) 2 with R = (hÀ1) 2 (h + 1) À2 corrects for two back reflections, one from the sample front and one from the back face. We used h = 2o fs olid C 60 as the refraction index. [41] The spectra in Figures 9a nd 10 are difference spectra a(T)Àa(300 K) with the base line removed.  [24a,b] IN6 is designed for quasi-elastic and inelastic neutron scattering. It operates on ac old neutron source and provides good resolution at low energy transfer,a ccessing both neutron energy (NE) gain and NE loss components of the spectrum. By convention the NE gain is defined with negative values of energy transfer DE. The powdered samples were wrapped in Al foil sachets for mounting in the cryostat. A"blank" sample of open fullerene (1)w ith identical mass to that of the CH 4 @1 sample was employed. In order to remove spectral features arising from fullerene cage modes and scattering from the construction materials of the cryostat, the INS spectra from the "blank" were subtracted from those recorded on the sample containing CH 4 .The spectra were recorded at cryogenic temperatures and the data is openly available [http://doi.ill.fr/10.5291/ILL-DATA.7-04-148].

Computational Experiments
Computational experiments were carried out using the Gaussian 09 software package. [42] Am odel structure (1b)f or open fullerene 1 in which the 6-tert-butyl pyridyl substituents were replaced by methyl substituents, was used. Structures and transition states were optimised using DFT with the M06-2X functional [43] and Dunning's correlation consistent basis set cc-pVDZ. [44] Frequency calculations were carried out for each stationary point to check that the optimised geometry corresponded to am inimum or at ransition state, and to allow the Gibbs free energies and entropies to be calculated at defined temperatures and pressures using the Gaussian freqchk utility.V ibrations were not scaled and low frequency vibrations were not removed. Electronic energies were calculated at the above geometries using M06-2X with the cc-pVTZ basis set and were corrected for basis set superposition errors using the counterpoise method. [45]