Endohedral fullerenes (endofullerenes) are stable host-guest complexes in which atoms, ions or molecules are trapped inside the cavity of a fullerene1, and are usually sub-grouped as endohedral metallofullerenes2 and non-metal endofullerenes3.

To date, non-metal endofullerenes in which atomic nitrogen, phosphorus, noble gases, or a small molecule are encapsulated by C60 or C70 (most commonly) have been prepared, and are denoted A@C60 (e.g.,) where ‘A’ represents the trapped endohedral species. They are compounds of enormous interest in several areas; for spectroscopic study of the quantised energy level structure of the trapped species4,5,6,7, for study of the internuclear (host-guest) interactions resulting from encapsulation and validation of predictive models of these interactions8,9, for study of the effect of an encapsulated species upon the properties and reactivity of the cage10,11,12,13, and for the materials applications that arise in each of these areas. Currently, there are few reviews in these fields, since the availability of non-metal endofullerenes at macroscopic (multi milligram) scale has developed apace only recently.

In early mass spectrometry experiments, collision of accelerated C60+• or C70+• with helium gas resulted in encapsulation of a single helium atom by the fullerene cage14,15,16,17, and lead to the development of the first methods for preparation of non-metal endofullerenes by high energy direct insertion. Exposure of C60 under high temperature and high pressure of a noble gas, leads to ~0.1% insertion of He, Ne, Ar or Kr, or 0.03% of Xe18, and the level of incorporation is improved in the presence of KCN, to 1% for He and ~0.3% for Ar, Kr, or Xe19,20,21. Substantial enrichment of the noble gas content, by removal of empty C60 using many cycles of preparative HPLC, is achieved for the heavier endofullerenes, but mass recovery is low (Ar@C60, 1.3 mg, 98% filled21,22; Kr@C60, 1.0 mg, 99% filled23; Xe@C60 0.32 mg, 50% filled20). Similarly, detection by mass spectrometry of N2@C60/70, CO@C60 and HeNe@C60 results from high temperature exposure of the fullerene to a high pressure of the corresponding gas24. N2@C60/70 is also obtained from ion implantation under glow discharge25, and the atomic endofullerenes N@C60/70 and P@C60 have been prepared by ion implantation techniques25,26,27,28,29. Pure N@C60 has been isolated by exhaustive HPLC enrichment30, although the material yield (microgram scale) from these direct insertion methods is too low for many spectroscopic applications30,31,32.

In the mid 1990s, Fred Wudl and Yves Rubin proposed the synthesis of endofullerenes from a multi-step procedure in which an opening in the fullerene is created, of suitable size to allow entry of a guest species into the cavity before a series of reactions to repair the cage-opening—restoring the original fullerene with the endohedral guest species trapped inside33,34. For realisation of this approach, controlled methods to create an opening in the fullerene cage would be required and Wudl accomplished the earliest cage-opening of C60 by 1,3-dipolar cycloaddition of an alkyl azide—involving scission of a single bond of C60 (Fig. 1a). Furthermore, controlled expansion of the orifice size was demonstrated by a contiguous oxidative cleavage, via [2 + 2] cycloaddition with singlet oxygen33,35. A strategy of two-step saturation of a six-membered ring of the fullerene cage, followed by [2 + 2 + 2] ring-opening, was explored by Rubin (Fig. 1a)36,37,38, after both authors showed that C60 participates as the 2π component in [4 + 2] cycloadditions (achieving partial saturation, a good first step)39,40. Interruption of the intended sequence by reaction of the first cycloadduct, 4, with 1O2 lead to an open-cage derivative (5) of C60 with the largest orifice then known, able to accommodate H2 (5 mol%) or an atom of 3He (1.5 mol%) when heated under a high pressure of either gas (475 atm 3He, or 100 atm H2) (Fig. 1b)41.

Fig. 1: Early studies of C60 cage-opening.
figure 1

a Wudl’s pioneering one-bond scission and regioselective oxidative cleavage of C6033, and Rubin’s general strategy to create an opening in C60 by two-step saturation of a six-membered ring (saturating substituents on 2 are not shown) followed by [2 + 2 + 2] ring-opening38. Open-fullerene 3 has a 15-member orifice, but this elegant route was not realised practically. b Controlled partial saturation by a one-pot bis-azide addition to C60 in which the fullerene acts as the 2π component in a [4 + 2] Diels-Alder cycloaddition leading to 4, followed by cycloaddition of 1O2 and [2 + 2 + 2] ring-opening. Bis-lactam 5 has an orifice that allows entry of He or H2 into the cage41. c Cycloadduct 6 undergoes photochemically induced [4 + 4] rearrangement, before thermal [2 + 2 + 2] cycloreversion of 7, an unstable intermediate42. The isolable product 8 is of a general structure common to all subsequent reports of C60 cage-opening that have been applied for synthesis of non-metal endofullerenes A@C60.

Preparation of these first ‘open’ endohedral fullerenes, H2@5 and 3He@5, was just one of two early milestones achieved by the Rubin group. In 1996, the product, 6, of Diels-Alder [4 + 2] cycloaddition between C60 and 1-((trimethylsilyl)oxy)-1,3-butadiene, acidic cleavage of the silyl ether and dehydration, was found to undergo photochemical [4 + 4] rearrangement followed by [2 + 2 + 2] cycloreversion of the unstable intermediate 7 (Fig. 1c)42. The isolable ethene-bridged product 8 has an eight-membered opening too small for the entry of guest species into the cavity. Importantly however, this sequence of a Diels-Alder cycloaddition of C60 coupled with elimination to form an intermediate of core structure 6, followed by the [4 + 4] and [2 + 2 + 2] pericyclic steps, would become the initial cage-opening process of all subsequent syntheses of non-metal endofullerenes—every example of which involves an intermediate of general structure 8 as the first stable C60 (or C70) derivative in the reaction sequence.

Herein we give a succinct review of the synthesis of non-metal closed endohedral fullerenes via open-cage intermediates, routes that have become known as ‘molecular surgery’. In our discussion of open-fullerenes, we include only those for which encapsulation of an atom or molecule has been demonstrated, and we further limit these to the examples whose synthesis contributed methods to the development of completed routes to A@C60/70. An excellent, comprehensive review of open-cage fullerenes is already available43. More recent studies of the encapsulation of small molecules by open-fullerenes derived from fullerene-mixed peroxides, towards applications of the open host-guest complex by selective trapping/release of the guest, were conducted by the group of Liangbing Gan and lately reviewed44.

Synthesis of closed endofullerenes will be categorised according to two main synthetic routes for encapsulation of (i) ‘small’ guest species He, Ne, H2, HF or H2O in the fullerene cage, and (ii) all larger noble gas atoms and small molecules. Current challenges are discussed in a final ‘outlook’ section.

Open-cage fullerenes

Concurrent methods to obtain an open-fullerene with core structure 8 in one-pot from C60, followed by regioselective oxidative cleavage to widen the cage-opening, were developed by the groups of Shizuaki Murata45,46 and Koichi Komatsu47,48,49.

Murata reported the cycloadduct (10) of Diels-Alder reaction between C60 and palladacyclopentadiene complexes, 9, to undergo photoinduced [4 + 4] rearrangement and [2 + 2 + 2] cycloreversion, identically to the Rubin sequence, to give open-fullerene 11 in ~70% yield (based upon 9 = dimethoxyglyoxime complex, Fig. 2a)46. The HOMO of ethene-bridged open-fullerenes with core structure 8 is localised at the (two) double bond(s) C(1)-C(2)48, and oxidative cleavage of 11 occurs regiospecifically at this position upon irradiation of a toluene or CHCl3 solution of 11 in air, since formation of 1O2 is sensitised by the fullerene itself45. The resulting diketone, 12, has too small an opening for the entry of a guest molecule but was found to undergo an unusual reaction with either substituted hydrazines or hydrazones50, or o-phenylenediamine51 reagents—each involving clean and highly selective scission of the C(3)-C(4) bond by hydroamination. Open-fullerenes, 13 and 14 were obtained respectively, and 14 was found to participate in further selective widening of the cage in the presence of additional amine base. Similarly, widening of the orifice of Wudl’s ketolactam open-fullerene 1 was achieved with o-phenylenediamine and excess pyridine, to furnish 16 (Fig. 2b)52. S-i. Iwamatsu and S. Murata have reviewed the elegant characterisation work which they carried out to elucidate the structures of 131653, and demonstrated the encapsulation of molecular guests (Fig. 2c). Encapsulation of molecular hydrogen by open-fullerenes 13a-c occurs under 13.5 MPa H2 at 100 °C, with up to 83% ‘filling’ of the cage50. The openings of 15 or 16 are too large to prevent the rapid escape of H2, but allow larger guest species to be accommodated and retained. Water enters both 15 and 16a/b under ambient conditions, and temperature-dependent loss of water from H2O@16a/b is much slower than from H2O@15 which has a bigger cage opening51,52. Accordingly, encapsulation of CO, NH3 and CH4 by 15 was accomplished under conditions of moderate pressure to furnish stable endohedral fullerene products, of which NH3@15 is reported to undergo a (slow) partial loss of the guest molecule54,55,56.

Fig. 2: Iwamatsu open-cage derivatives of C60 and their encapsulation of molecular species.
figure 2

a An open-fullerene (11) with the core ethene-bridged structure of 8 is obtained in one-pot from C60, and undergoes regioselective oxidative cleavage. Widening of the opening is achieved by further reaction with a hydrazine or 1,2-diamine. b Wudl’s ketolactam 1 is also a substrate for orifice-widening with an o-phenylenediamine. c Conditions for molecular guest encapsulation by open-fullerenes 13, 15 and 16; aTCE = 1,1,2,2-tetrachloroethane, bApprox. 0.85 MPa (i.e., vapour pressure of NH3 at room temp.), cPartial loss of NH3 occurs slowly, during 6 months storage at −10 °C, dH2O encapsulation by 15 occurs under ambient pressure and the % filling shows a temperature-dependent entry/escape equilibrium (var. = variable).

Akin to the pathway described by Murata and Iwamatsu, a further example of one-pot preparation of an open C60 derivative with core structure 8 was reported by Komatsu (Fig. 3). Upon heating C60 with phthalazine, in 1-chloronaphthalene solution near reflux, open-fullerene 18 was obtained from [4 + 2] cycloaddition, loss of N2 from an unstable intermediate 17, and the [4 + 4] addition / [2 + 2 + 2] reversion sequence already described47. Oxidative cleavage of C(1)-C(2) is regiospecific, although diketone 19 was obtained in modest yield (Fig. 3a)48 cf. the comparable diketone 12. Further examples of [4 + 2] cycloaddition between C60 or C70 (as the 2π component) and ‘diene’ partners embedded in a pyridazine core structure, like the reaction with phthalazine, would lead to the general adoption of this method for the initial cage-opening in synthesis of endofullerenes. These have been reviewed recently57. Of the first examples (Fig. 3b), a substituted 1,2,3-triazine, 20, was partnered with C60 to confer solubility of the open-fullerene product(s) in common organic solvents58. The cycloadduct 21 is an imine-bridged asymmetric analogue of the ethene-bridged compounds 8, 11 or 18 and, from DFT calculations, the HOMO of 21 is localised at the C(2)-C(3) and C(4)-C(5) double bonds (similarly to the ethene-bridged examples). Accordingly, oxidative cleavage of either the C(2)-C(3) or C(4)-C(5) double bond of 21 using 1O2 leads to a separable mixture of 22 and 23, respectively. The major product, 22, was obtained in 61% yield and its orifice can be widened by: (i) Iwamatsu’s regioselective addition of an aromatic hydrazine or hydrazone59, or o-phenylenediamine60 to give 24; or, (ii) sulfur atom insertion using S8 in the presence of a single-electron reductant, tetrakis(dimethylamino)ethylene (TDAE), from which 25 is isolated in good yield58.

Fig. 3: Open-fullerenes prepared from reaction of C60 with pyridazine derivatives.
figure 3

a An open-fullerene (18) with the core ethene-bridged structure of 8 is obtained in one-pot from C60, and undergoes regioselective oxidative cleavage. b Cycloaddition of C60/70 with a substituted triazine or pyridazine. Abbreviations: 2-pyridyl (Py), o-dichlorobenzene (ODCB), tetrakis(dimethylamino)ethylene (TDAE).

The cage-opening of 24 is of the same size as its all-carbon analogue 15, and 24 was similarly shown to encapsulate water under conditions of ambient pressure. The equilibrium between empty 24 and H2O@24 is dependent upon temperature and solvent polarity. Encapsulation of formaldehyde or HCN by 24 also occurs under conditions of ambient pressure; H2CO@24 is observed as the minor component (9%) in an inseparable mixture with H2O@24 (35%) and empty 24 (56%) by passing gaseous formaldehyde through a solution of 24 in chlorobenzene at 100 °C. Treatment of a mixture of 24 and H2O@24 in chlorobenzene with excess HCN at 90 °C results in displacement of water and recovery of HCN@24 with near-quantitative HCN incorporation. Slow thermal dissociation of HCN@24 is reported60.

Steps to repair the cage-opening of the endohedral open fullerenes derived from the hydroamination reactions (A@13, A@15, A@16, A@24) have not been developed, as it is a hugely challenging task to find conditions for reversal of the complex rearrangement steps that follow the initial amine condensation. Instead, reversal of the route by which open-fullerene 25 is prepared i.e., by sulfur extrusion and a McMurry-type reductive coupling of the diketone as first steps, is a practicable approach and would be explored by Komatsu and co-workers (see ‘Synthesis of closed endohedral fullerenes’ below). Of course, it is necessary that the orifice of 25 is big enough for the entry of a guest species, and although the 13-membered ring is smaller than that of any example discussed above (5, 13, 15, 16 or 24) calculation of the activation barrier to entry of small guests He and H2 into 25 (18.9 and 30.1 kcal mol−1 respectively)61 with that for entry to 5 (24.5 and 41.4 kcal mol−1 respectively)41 suggests that encapsulation in 25 could be achieved. Indeed, upon heating a powdered sample of 25 at 200 °C under 800 atm of H2 a quantitative recovery of H2@25 was made. The experimental activation energy for escape of hydrogen from H2@25 is Ea = 34.2 ± 0.58 kcal mol−1, and so the complex is stable to dissociation at room temperature61,62. The barrier to escape of helium from 3He@25 (Ea = 22.8 kcal mol−1)63 is much lower than that for dissociation of H2@25, so after heating 25 at 90 °C under 650 atm of helium, >35 mol% encapsulation is inferred by cooling He@25 to −20 °C and reduction with NaBH4 to form a hemiaminal ether across the cage-opening that blocks the escape64.

A closely alike sequence to that used for preparation of 25 was adopted to obtain 27, an open-cage derivative of C70 (Fig. 3b). Thermal cycloaddition between C70 and 3,6-di(2-pyridyl)pyridazine, then photooxidation under xenon lamp irradiation in air, lead to 26 before sulfur insertion afforded 27. The calculated energy barrier for encapsulation of H2 in the cavity of 27 is 31.2 and 31.0 kcal mol−1 for entry of a first then second molecule, respectively (cf. 30.1 kcal mol−1 for H2 entry into 25), and suggests that the 13-membered cage-opening is of comparable size to that of 25—as might be expected from their structural resemblance. Accordingly, heating a powdered sample of 27 at 200 °C under 830 atm of H2 gave a mixed sample of H2@27 (97%) and (H2)2@C70 (3%)65.

Synthesis of closed endohedral fullerenes A@C60 and A@C70

Synthesis of closed fullerenes containing small guest species, He, Ne, H2, HF or H2O

When Komatsu’s open-cage endofullerene H2@25 was subjected to matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry, molecular ion peaks for H2@25 (m/z 1068), empty 25 (m/z 1066), H2@C60 (m/z 722) and C60 (m/z 720) were observed; indicating that repair of the cage-opening of H2@25 to obtain H2@C60 is feasible although substantial loss of H2 (~70%) occurs with gas-phase laser irradiation61. The half-life for thermal dissociation of H2@25 is t1/2 = 54.4 h at 160 °C, but the complex is stable at room temperature so reaction conditions to reduce the size of the cage-opening must avoid high-temperatures. This was accomplished by the Komatsu group in their landmark synthesis of H2@C60, using methods they also applied for the synthesis of 4He@C60 and shown in Fig 4a66,67,68. Oxidation of H2@25 at room temperature gives exo-sulfoxide H2@28, more stable than the endo-sulfoxide by 8.6 kcal mol−1 presumably as a result of steric congestion between the sulfinyl and carbonyl groups in the endo-form68. Then, constriction of the size of the opening is also achieved under very mild conditions, by a photochemical desulfinylation under visible light irradiation. The contracted product of SO extrusion, H2@22, is thermally stable—no loss of H2 occurs after heating a solution of H2@22 in 1,2-dichlorobenzene-d4 at 190 °C for 3 days, cf. t1/2 = 4.2 h for thermal dissociation of H2@25 at 190 °C. Correspondingly, McMurry reductive coupling of H2@22 was performed without loss of endohedral hydrogen at 80 °C, returning to the imine-bridged intermediate H2@21, before final closure of the cage was accomplished by heating at 340 °C under vacuum. H2@C60 is thereby obtained with 93 mol% encapsulation of H2 in an overall yield of 9% from C60, and pure H2@C60 was recovered after preparative recycling HPLC in a substantial material quantity of ~100 mg66,68.

Fig. 4: Komatsu’s synthesis of H2@C60/70 and 4He@C60/70.
figure 4

a Synthesis of H2@C60 and 4He@C60 via sulfide oxidation and photochemical desulfinylation, each at ambient temperature. The order of steps is altered to avoid an intermediate 4He@25 from which escape of helium is very facile. b Synthesis of H2@C70 and 4He@C70 using identical methods. Tetrahydrofuran (THF), 2-pyridyl (Py), o-dichlorobenzene (ODCB).

As helium escapes rapidly from He@25 at room temperature (in previous work, this species was reduced to a hemiaminal ether derivative in which the cage-opening is blocked, i.e., helium is trapped, see earlier) it proved necessary to re-order the reaction sequence described for synthesis of H2@C60, in order to prepare He@C60 (Fig. 4a). Oxidation of empty open-fullerene 25 was carried out first, and the sulfoxide derivative 28 was then the substrate for ‘filling’ under 1230 atm of helium gas at 115 °C before cooling to ambient temperature whilst pressurised, and rapid photo-desulfinylation for just 1.5 h, at room temperature. Encapsulation of endohedral species (e.g., 4He in this example) under high-pressure conditions well in excess of 1200 atm is cited in many further examples, vide infra, and although we refer the reader to the original literature for details of the bespoke apparatus used therein, typically relies upon hydraulic or manual gas compression following initial pressurisation. 4He@22 was obtained with 30 mol% incorporation of helium and no loss of the endohedral atom occurred during the final two steps, McMurry reductive coupling and thermal closure. Enrichment to a sample of 4He@C60 with 95% helium encapsulation was achieved67.

These cage-closure methods were also applied for the preparation of a separable mixture of H2@C70 + (H2)2@C70 that were each isolated as the pure endofullerene69, and 4He@C70 with 30 mol% helium incorporation67 (Fig. 4b). Helium incorporation was enriched to 60 mol% by recycling HPLC. The endohedral helium dimer was not detected, although it is known that the cavity of C70 can accommodate two helium atoms70.

The molecular endofullerene H2O@C60 is an important target for synthesis by molecular surgery, as a substrate for study of the rich, quantum energy level structure of the isolated water molecule. In order to achieve the synthesis of H2O@C60, an open-cage derivative of C60 with a larger opening than the examples discussed so far was required (Fig. 5). After cage-opening of C60 by reaction with 3,6-bis(6-(tert-butyl)pyridin-2-yl)pyridazine (29) according to the now well-established sequence of [4 + 2] cycloaddition, [4 + 4] rearrangement and retro-[2 + 2 + 2] cycloreversion, photo-oxidative cleavage leads to diketone 3071. The first report of this reaction with 1O2 required irradiation of a solution of the open-fullerene in mixed 1-chloronaphthalene and CS2 solvents for 23 h, as oxygen is passed through the reaction vessel. However, CS2 has flashpoint ca. 30 °C and an auto-ignition temperature of 100 °C, prompting the research groups of both Yasujiro Murata and Richard Whitby to seek safer alternatives; CS2 may be replaced with CCl4 under LED irradiation to obtain 30 in 52% yield72, and switching the co-solvent to toluene leads to the isolation of 30 in an improved yield of 70% after just 1 h under irradiation with a high-pressure sodium lamp (Fig. 5a)73. The isolated yield of 30 is measured over two steps in which the C60 starting material is present in excess—i.e., acting as photosensitizer in the singlet oxygenation. Murata and co-workers found that a second regiospecific oxidative cleavage of the C(1)-C(2) bond of 30 takes place using N-methylmorpholine N-oxide, and the resulting tetraketone, 31, is isolated as its bis(hemiketal) hydrate 32 when the oxidation is carried out in wet THF. The 16-membered cage opening of 31 is large enough for entry of water, so upon heating a solution of 32 in wet toluene at 120 °C, for 36 h under 9000 atm, quantitative recovery of H2O@32 is made via the dynamic equilibrium between 31 and 32, that enables encapsulation of H2O by 31 whilst the water molecule is unable to escape from the hydrate 32, since the 13-membered opening is too small71. Without pressurisation, a solution of 32 in wet toluene equilibrates to 23 mol% endohedral water content after 36 h at 120 °C; and Whitby et al. found that heating a solution of 32 in 1-chloronaphthalene with water to 100 °C in a sealed tube gives H2O@32 with 78 mol% endohedral water, after 48 h (i.e., under some pressurisation but using accessible conditions that require no special apparatus)74.

Fig. 5: Synthesis of H2O@C60.
figure 5

a Open-cage fullerene 31 has a 16-membered orifice and is the substrate for encapsulation of H2O. Encapsulation occurs by in situ dehydration of 32, see the main text for detail. b Double reductive coupling of carbonyl groups on the 16-membered tetraketone cage-opening. Only the orifice atoms are shown. c Final cage-closure. d Synthesis of H2O@C60. 1-chloronaphthalene (1-ClNpth), N-methylmorpholine N-oxide (NMMO), 2-pyridyl (Py), o-dichlorobenzene (ODCB).

Repair of the cage-opening of 32 involves dehydration to return to the tetraketone 31, then sequential reductive couplings of the ‘paired’ carbonyl groups that were formed in the sequential oxidative cleavage steps during cage-opening (Fig. 5b). Reductive coupling is achieved upon reaction with trivalent phosphorus reagents, by a mechanism that involves initial formation of an intermediate β-oxo-phosphorus ylid 33, then intramolecular Wittig reaction that returns the cage-opening to the diketone 3073,75. Formation of the β-oxo-phosphorus ylid 33 could take place via attack of phosphorus at the carbonyl carbon followed by [1, 2]-phospha-Brook rearrangement, by electron transfer to the fullerene then attack of phosphorous directly at oxygen, or by Kukhtin-Ramirez addition76,77,78; elimination of R3P = O occurs with another phosphine/phosphite addition in each case. Diketone 30 is then subject to the same reduction sequence, although calculations support the formation of an epoxide intermediate 34 from the first +PR3/−R3P = O step, rather than a mechanism involving phosphorus ylid formation and intramolecular Wittig reaction79. With an excess of the phosphorus reagent, the stable ethene-bridged derivative 35 is obtained. The final step of the closure sequence leads to C60, and involves sequential [4 + 2] intramolecular cycloaddition, radical cleavage of the strained intermediate 36 (formally a retro [4 + 4] cycloaddition) and [2 + 2 + 2] cycloreversion (Fig. 5c)68. In their syntheses of 4He@C60, H2@C60, H2@C70 and (H2)2@C70 described earlier (Fig. 4), Komatsu and Murata employed vacuum pyrolysis for this step.

So, from their sample of pure H2O@32 obtained by high-pressure filling, the Murata group achieved the first synthesis of H2O@C60—effecting dehydration to H2O@31 and the sequential reductive couplings with excess P(OiPr)3 in refluxing toluene, before vacuum pyrolysis of alumina-supported solid H2O@35 to complete the closure. H2O@C60 was obtained in 15% over these steps (Fig. 5d)71.

The yield of the first reductive coupling (of H2O@31) using alkyl phosphite reagents is compromised by unwanted formation of an α-hydrophosphate side-product, but clean reduction occurs with trialkyl phosphines74. From their sample of H2O@32, with 78 mol% endohedral H2O, the Whitby group carried out dehydration of the bis(hemiketal) to obtain H2O@31 with in situ clean reduction to H2O@30 using excess PPh3 in refluxing toluene. The second reductive coupling was then conducted with P(OiPr)3, and the final pyrolysis was adjusted to follow a lower energy pathway in the presence of N-phenylmaleimide39, which reacts with intermediate 37 in a [4 + 2] Diels-Alder reaction. The cycloadduct 38 reverts to C60 via a retro [4 + 2] cycloaddition (Fig 5c) and H2O@C60 (78 mol% H2O) was obtained in an improved yield of 51% from H2O@32 (Fig. 5d)74.

Finally, to optimise the synthesis of H2O@C60, Murata recently reported theoretical modelling of water encapsulation and cage closure steps for structural analogues of tetraketone 31, choosing substituent patterns around the orifice that could be readily accessed according to the choice of azine used as the ‘diene’ 4π partner to C60 (the 2π component) in the first [4 + 2] pericyclic cage-opening step of the molecular surgery route. The optimal open-cage derivative was predicted to be the one formed from oxidative cleavage of Komatsu’s diketone 22 (Fig. 3), which was therefore prepared from 22 using N-methylmorpholine N-oxide in wet THF, and isolated as its bis(hemiketal) hydrate 39. Near-quantitative water encapsulation was achieved under high-pressure to give H2O@39 (i.e., similarly to high-pressure quantitative water ‘filling’ of 32 via the equilibrium with its dehydrated form), and in situ dehydration of the bis(hemiketal) then sequential reductions with P(p-tolyl)3 and (P(OiPr)3 were effected in a single pot, before vacuum pyrolysis gave H2O@C60 (98 mol% H2O) with 87% isolated yield in the pyrolysis step and in 70% from H2O@39 (Fig. 5d)79.

The same methods have also been applied for synthesis of H2O@C70. Initial [4 + 2] Diels-Alder cycloaddition of 3, 6-bis(6-(tert-butyl)pyridin-2-yl)pyridazine (29) occurs at the α-bond or β-bond of ellipsoidal C70 to yield isomeric products in 42% and 6% (from α- and β-bond scission, respectively) after the 4 + 4] rearrangement and retro-[2 + 2 + 2] cycloreversion sequence. Widening of the cage-opening of each isomer can be achieved using the methods already described; photo-oxidative cleavage with 1O2 followed by a second regiospecific oxidative cleavage with wet 4-dimethylaminopyridine N-oxide yields isomeric tetraketone open-fullerene derivatives of C70 each with a cage-opening of identical structure to the C60 analogue 31. However, only the C70 open-fullerene tetraketone derived from initial scission of the β-bond is easily able to accommodate entry of water, presumably since there is more strain release associated with β-bond scission (or a C60 bond scission) which corresponds to a bigger resultant orifice80. Correspondingly, the β-bond scission isomer is the minor product of C70 cage-opening, but is the required intermediate to achieve the encapsulation and closure steps. These have been carried out as already discussed for synthesis of H2O@C60 (Fig. 5); i.e., water uptake in wet toluene at 120 °C, for 40 h under 9000 atm, double reductive coupling with excess P(OiPr)3 in refluxing toluene, and thermal closure in the presence of N-phenylmaleimide. Pure H2O@C70 is obtained by preparative single-stage HPLC, from a separable mixture containing ca. 18% empty C70 and trace (H2O)2@C7081. Interestingly, encapsulation of H2O into the C70 open-fullerene tetraketone derived from initial α-bond scission does occur in the presence of HF; treatment with wet HF-pyridine (70% w/w, 0.5 molar equiv.) at 9000 atm. and 120 °C for 18 h, results in encapsulation of HF (32 mol%), H2O・HF (11 mol%) and H2O (27 mol%). After cage closure, pure (H2O・HF)@C70 can be isolated from an inseparable mixture of H2O@C70 and HF@C7082.

Calculation of the binding energy and activation energies of entry/exit for encapsulation of guest species into open-cage fullerenes is a vital tool to inform the conditions of ‘filling’ and subsequent steps, towards synthesis of A@C60. This approach has been applied for syntheses of H2@C60, HF@C60 and the smaller noble gas endofullerenes He@C60 and Ne@C60 (Fig. 6).

Fig. 6: Whitby’s synthesis of closed C60 endofullerenes containing a small endohedral species.
figure 6

Open-cage fullerenes 31 and 40 each have a 16-membered orifice able to accommodate the entry of guest species. Solution-phase encapsulation of H2 or HF by 31 occurs via dehydration of bis(hemiketal) 32 for synthesis of H2@C60 and HF@C60. Solid-state filling of 40 was performed under the tabulated conditions for optimised synthesis of H2@C60 and He@C60 isotopologues, and Ne@C60. High-pressure solid-state filling was carried out in a 100 × 5.2 mm 316 L stainless steel reactor as part of a bespoke apparatus for gas compression using a manual pump.73 1-Chloronaphthalene (1-ClNpth), the ‘Ar’ 5-tert-butylpyridyl substituent structure is shown in Fig. 5a.

The calculated activation energies for both the entry of HF into 31, and its loss from the cage, indicate that encapsulation and release of HF are much more favourable than the corresponding trapping/release of H2O by the same open-fullerene75,83. So, an optimal 50 mol% filling with HF occurs under ambient conditions, by equilibration of a solution of either 31 or 32 in CH2Cl2 with excess HF-pyridine at room temperature83. Conversion of HF@32 to HF@31 occurs simply by stirring with molecular sieves at room temperature, and the β-oxo-phosphorous ylid intermediate of the first reductive coupling closure step, HF@40, is isolated from slow reaction between HF@31 and PPh3—also at ambient temperature. However, the intramolecular Wittig reaction of HF@40 which completes the reduction step requires heating to >100 °C and causes complete thermal dissociation. Loss of HF is minimised using di-(2-furyl)phenylphosphine which effects the reduction of HF@31 at a lower temperature, hence HF@30 is obtained in good yield with 30 mol% remaining HF incorporation from HF@32 (50 mol% HF) (Fig. 6). No loss of HF takes place from the small (12-membered) cage-opening of HF@30, so the second reduction is safely carried out with P(OiPr)3 in refluxing toluene, and thermal closure in the presence of N-phenylmaleimide returns HF@C60 (30 mol% HF)75.

The calculated activation energy for entry of H2 into the cavity of 31 is ca. 12 kJ mol−1 higher than that for entry of H2O83, despite the smaller size of H2 and presumably due to the attractive dipolar interactions of H2O in the cage entrance. Yet, as substantial H2O incorporation into 31 is achieved under very mild conditions (78 mol% using wet 1-chloronaphthalene at 100 °C in a sealed tube—see earlier74), Whitby showed that 60 mol% encapsulation of H2 in 31 takes places under conditions of only moderate pressure—under 120 atm H2 at 120 °C bis(hemiketal) 32 undergoes in situ dehydration, accelerated with molecular sieves, to form H2@31. Heating H2@31 with PPh3 then induces a contraction of the cage-opening by the first reductive coupling, but has to be conducted under the same H2 pressurisation to avoid loss of the endohedral molecule (Fig. 6). A second reduction with P(OiPr)3, then thermal closure in the presence of N-phenylmaleimide, completed the synthesis of H2@C60 (60 mol% H2) in 51% yield from 3174.

This route complements the synthesis of H2@C60 by Komatsu (Fig. 4), although H2@C60 was obtained with 93 mol% filling via Komatsu’s more forcing conditions for H2 encapsulation by 25 (800 atm H2 at 200 °C). After showing that the β-oxo-phosphorus ylid 40 is an isolable intermediate Whitby surmised that, if entry of H2 (or another species) into 40 could occur at a temperature lower than that required for the following intramolecular Wittig reaction, it would be possible to ‘fill’ the phosphorus ylid 40 then induce the Wittig closure that traps the endohedral species simply by raising the temperature73. Calculation of the activation enthalpies for entry of small guests, H2, He and Ne, through the 16-membered openings of 31 and 40 indicates that—in each case—the barrier to entry into 40 is only ca. 10 kJ mol−1 higher than that for entry into 31, and it was found that ‘closure’ of H2@40 (to H2@30) occurs after equilibration of H2 between the fullerene cavity and outside. As the Wittig reaction is unimolecular (cf. overall reductive coupling of 31) it also became possible to conduct the combined encapsulation and Wittig reaction steps without solvent, with significant advantages—a small (ca. 1–5 mL volume) pressure reactor can be used, so that high-pressure conditions can be safely achieved whilst the volume of gas remains low, allowing rare and/or expensive gases to be used. So, solid-state filling of 40 with the isotopologues of molecular hydrogen and helium, H2, HD, D2, 3He and 4He, as well as with Ne, was achieved with in situ thermal contraction of the cage-opening according to the conditions of Fig. 6 (table). Cage-closure of the resulting diketone endofullerenes A@30 (A = H2, HD, D2, 3He, 4He or Ne) was carried out using the usual conditions of a reductive coupling with P(OiPr)3, and N-phenylmaleimide-mediated thermal closure. Notable, are the improved syntheses of H2@C60 (95 mol% H2) and 4He@C60 (50 mol% 4He), syntheses of HD@C60 (83 mol% HD) and 3He@C60 (52 mol% 3He) despite the commercial availability of HD and 3He at only low-pressure, and the first synthesis of Ne@C60 (63 mol% Ne—enriched to 100 mol% by recycling preparative HPLC)73.

Synthesis of closed fullerenes containing larger noble gas atoms or small molecules

The 16-membered cage-opening of 31 (or 40) is too small to achieve entry of molecules larger than H2O, or of noble gas atoms larger than neon.

However, the encapsulation of ‘large’ molecules CO, NH3 and CH4 into Iwamatsu’s 17-membered cage-opened C60 derivative 15 (Fig. 2) encouraged Yasujiro Murata to apply the sulfur insertion method that he and Komatsu had earlier developed (for widening the orifice of 22 to 25, Fig. 3) for expansion of the opening of 31. Insertion of a sulfur atom into the rim of 31 was achieved using S8 in the presence TDAE to afford 41, which has a 17-membered opening (Fig. 7)84. Rapid exchange of water in/out of 41 at room temperature in CDCl3 indicates the opening to be larger than that of 15—thermal dissociation of H2O@15 is relatively slow51 —and in order that encapsulation of large species ‘A’ by 41 is a viable route for the synthesis of closed endofullerenes A@C60, it is obviously necessary that sulfur extrusion to contract the cage-opening (returning to A@31) can be performed using conditions under which the endohedral species is not lost from A@41, or from an intermediate in the process of sulfur removal. Development of conditions for synthesis of A@C60 via routes that rely upon encapsulation into 41 has therefore required knowledge of the energetics of the encapsulation and loss, A + 41 A@41.

Fig. 7: Synthesis of C60 endofullerenes containing a large endohedral species.
figure 7

a Open-cage fullerene 41 has a 17-membered orifice able to accommodate the entry of large guest species. Conditions for encapsulation of the guest species ‘A’ are described in the main text. Stable host-guest complexes CH4@41, Ar@41 and Kr@41 are intermediates in the synthesis of CH4@C60, Ar@C60 and Kr@C60; but more labile guests are characterised in the ‘stoppered’ open-fullerene A@42. Conditions for cage-closure of A@42 have not yet been developed. 1-Chloronaphthalene (1-ClNpth), the ‘Ar’ 5-tert-butylpyridyl substituent structure is shown in Fig. 5a.

So, Murata demonstrated pressure-dependent insertion of CH3OH into 41 in chlorobenzene solution at 150 °C, achieving up to 60 mol% encapsulation of CH3OH at 9000 atm, and also noting that contamination of the CH3OH@41 product with N2@41 indicates that partial solubilisation of gaseous species enables their insertion under pressure85. Accordingly, Whitby achieved 65 mol% encapsulation of CH4 into 41 in 1-chloronaphthalene solution at 200 °C, under 153 atm of methane86. Both CH3OH@41 and CH4@41 are stable at room temperature, showing no loss of the endohedral molecule over many months, and confirmed by the experimental kinetic parameters for thermal dissociation of CH4@41; Ea = 134.6 ± 5.0 kJ mol−1 and ΔG = 151.5 ± 0.1 kJ mol−1 at 165 °C. In contrast, insertion of formaldehyde (from 1, 3, 5-trioxane in chlorobenzene solution under 8000 atm, at 150 °C) gave H2CO@41 with 35 mol% H2CO, but more than half of the H2CO is lost from a solution of H2CO@41 in CDCl3 after 30 h at room temperature. To prevent the escape of H2CO from the cage, selective reduction of one carbonyl group C(1)-O(2) from the exo-face acts to ‘stopper’ the opening (Fig. 7), and the alcohol product H2CO@42 suffers no loss of formaldehyde after many months of storage at room temperature85. The calculated free energy for entry of ammonia into 41 (62.3 kJ mol−1) indicates facile entry under ambient conditions, such that solution-phase exposure of 41 to methanolic ammonia under dry conditions (to avoid encapsulation of water) results in rapid formation of NH3@41, but the complex is unstable to loss of NH3 similarly to the instability of H2CO@41 to loss of formaldehyde, and cannot be isolated. Instead, in situ reduction of NH3@41 using NaBH4 affords NH3@42 with >90 mol% NH3 incorporation86. In recent years, several examples of this ‘stoppered’ open fullerene A@42 have been obtained by encapsulation of guests with a low energy barrier to escape from 41, followed by reduction with NaBH4 or BH3・THF; N2@42 (43 mol% N2)87, CO2@42 (76 mol% CO2)87, 3O2@42 (81 mol% 3O2)88, NO@42 (90 mol% NO)89,90 and H2O2@42 (35 mol% H2O2)91. In each case, samples of pure A@42 can be obtained by recycling preparative HPLC—with the exception of H2O2@42 whose complete separation from contaminant H2O@42 is laborious.

Earlier we described the first steps for repair of the cage-opening of Komatsu’s sulfide, 25, via oxidation to the corresponding exo-sulfoxide 28, then photochemical desulfinylation under visible light irradiation ((Fig. 4), and it is straightforward to envisage these key steps applied to contract the cage-opening of 41 (i.e., more importantly, A@41). Indeed, oxidation of 41 using dimethyl dioxirane (DMDO) or mCPBA cleanly furnishes exo-sulfoxide 43 without trace of the endo-sulfoxide or sulfone, but upon attempted photodesulfinylation of a mixed sample of N2@43, H2O@43 and empty 43 using visible irradiation (Xe lamp, benzene, room temp., 21 h)— i.e., conditions comparable to those reported for elimination of SO from H2@28 and 4He@28—the anticipated desulfinylation product(s) N2@/H2O@/empty 31 were not obtained92. Nonetheless, mass spectrometric analysis of 43 implied that SO extrusion is feasible; a peak corresponding to [M+H − SO]+• appears in the atmospheric pressure chemical ionisation (APCI) mass spectrum92 and, significantly, the radical cation [M − SO]+• is the dominant species in the atmospheric pressure photoionisation (APPI) mass spectrum93. Encouraged to pursue the possibility of photo-induced desulfinylation of 43, and noting that the expected product of desulfinylation 31 is unstable under visible light irradiation, Bloodworth and Whitby found that the reaction was facilitated by trapping 31 as its more photo-stable bis(hemiketal) hydrate 32 in situ. CH4@43 was prepared with >99.5 mol% CH4 content by heating powdered fullerene 41 at 190 °C under >1500 atm of methane, before oxidation with DMDO; then, under irradiation at 589 nm with a low-pressure sodium lamp for 35 h in a mixed solvent system of toluene, acetonitrile and acetic acid (10% v/v aq.), CH4@43 successfully underwent loss of SO and hydration to give CH4@32—although in only 13% isolated yield. The <1% content of ‘empty’ 43 carried through this reaction encapsulates water under the aqueous conditions such that a trace of H2O@32 contaminates the CH4@32 product, so endohedral water is removed at 140 °C under a dynamic vacuum (conditions that also effect dehydration of CH4@32 to CH4@31) without loss of methane, before completion of the final cage closure steps. The first of the two sequential reductive couplings (of A@31, then A@30, see Fig. 5b and the earlier discussion) was carried out using PhP(2-furyl)2 at 50 °C, i.e., under the mild conditions originally developed to attenuate loss of HF during reduction of HF@31 (Fig. 6), but now because the temperature is too low for re-entry of water traces. The second reduction (of CH4@30) was safely achieved with P(OiPr)3 in refluxing toluene as the opening of 30 is too small to accommodate water, and the final N-phenylmaleimide-mediated closure step gave pure CH4@C60 after removal of the traces (<1%) of empty C60 by recycling HPLC (Fig. 7)93.

The successful (if low-yielding) photo-desulfinylation of CH4@43 has also enabled the method to be applied for preparation of the larger noble gas endofullerenes, Ar@C6094 and Kr@C6095 (Fig. 7). DFT calculations of the barrier to entry and binding enthalpies for encapsulation of the larger noble gas atoms argon and krypton by 41, cf. methane, indicated that filling could be achieved under similar high-pressure conditions. Accordingly stable open endofullerenes Ar@41 and Kr@41 were obtained with near-quantitative incorporation of the noble gas atom under conditions similar to those that gave >95 mol% CH4 encapsulation: ca. 1400 atm of argon, or ca. 1500 atm of krypton, at 180 °C. Completion of the syntheses of Ar@C60 and Kr@C60 was carried out according to the methods described for CH4@C60, with improved isolated yields of 26% and 23% of Ar@32 and Kr@32, respectively, from the key photo-desulfinylation step, suggesting that larger endohedral species’ inhibit the reaction. However, experiments with mixed CH4@43/H2O@43 samples cannot distinguish between an inhibitory effect of methane and a promoting effect of water for example93, and the mechanism by which an endohedral species influences fullerene reactivity in the desulfinylation step remains to be fully understood. As methanol is a larger species than methane but has an electronic structure closer to water, it is of interest to note that of the group of stable complexes A@41 given in Fig. 7, it is the earliest reported example CH3OH@41 that remains an unused intermediate, i.e., the synthesis of CH3OH@C60 has not yet been pursued to our knowledge.


The synthesis of noble gas endofullerenes He@C60/70, Ne@C60, Ar@C60 and Kr@C60, molecular endofullerenes H2@C60/70, (H2)2@C70, HF@C60, H2O@C60/70, (H2O・HF)@C70 and CH4@C60, and isotopologues of several of these, are significant achievements from the research groups of Koichi Komatsu, Yasujiro Murata and Richard Whitby.

The dominant open-fullerenes now employed as key intermediates for guest encapsulation in molecular surgery are 31 (16-membered opening for entry of ‘small’ atoms and molecules), and 41 (17-membered opening for entry of larger species). The classification of endohedral guest species’ as ‘small’ or ‘large’ is not intended to imply that encapsulation depends solely upon their size, as the energies of activation for guest entry/exit and binding inside the cage depend on both steric and electronic interactions. Rather, this grouping reflects a calculated barrier to encapsulation into 31 that informs the authors’ own work.

Many studies of the properties of non-metal endofullerenes have been facilitated by the availability of the materials, although a review of these is sadly beyond our scope here. Similarly, theoretical study of endofullerenes is a very large field and our own motivation for endofullerenes synthesis is both the opportunity for their direct study and also the value of resulting data as a test of theoretical models. To satisfy these needs, many synthetic challenges remain to be addressed.

A low-yielding photochemical ring-contraction step is a constraint of the current method for synthesis of A@C60 via A@41, limiting the yield and material quantities which can be obtained when A is ‘large’ (CH4, Ar, or Kr to date). An understanding of the mechanism and approaches to optimisation of this limiting step, are of great importance—not only to overcome the low yield, but also to inform new routes for ring-closure. Two major targets, not yet achieved, are synthesis of NH3@C60 and O2@C60, which have exciting applications in nuclear hyperpolarisation. These species cannot be accommodated by the smaller cage-opening of 31 but escape rapidly from the larger opening of 41.

A ‘stoppered’ open-fullerene 42 restricts the escape of NH3 and O2 (as well as N2, NO, CO2, H2CO and H2O2) but does not, in our hands, undergo contraction of the cage-opening under similar conditions to the photo-desulfinylation of A@43.

Noble gas endofullerenes are of enormous contemporary interest as the first series of compounds in which it is possible to study internuclear interactions between a noble gas atom and the cage, or in the noble gas dimer. Encapsulation of xenon by 41 is calculated to have ΔHentry = 152 kJ mol−1 and ΔHbind = −56 kJ mol−1, a significantly higher barrier to encapsulation than for krypton (ΔHentry = 87 kJ mol−1 and ΔHbind = −57 kJ mol−1)95, the largest noble gas encapsulated in an open-fullerene so far. In consequence attempted preparation of Xe@41 under >1800 atm pressurisation of xenon gas, at 212 °C for 17 h, results in negligible (<1%) xenon incorporation95, restricting the range of noble gas endofullerenes available for study. Furthermore, the possibility of encapsulating still larger guests (including dimers) in the bigger cavity of C70 has not been realised, in part because a C70 derivative with the ‘large’ 17-membered opening corresponding to the structure of 41, remains elusive80,96.

Solutions to these challenges, e.g., involving new ring-closure methods and alternative ‘large’ cage-opened derivatives of C60/70, informed by both experimental and computational studies, is where much effort is currently directed in the field.