Pressure is a fundamental thermodynamic variable that can dramatically alter the atomic and electronic structures of materials, resulting in concomitant changes to their properties.¹ The C=C bonds in fullerenes allow them to form covalent intermolecular bonds when treated at high pressure and high temperature (HPHT). The preparation path based on the reaction diagram and the treatment time greatly influences the completeness of transformation and the degree of polymerization. In addition, the hydrostatic nature of the compression environment plays an important role in bonding orientation and polymerization ordering. As a result, dimers, ordered one-dimensional (1D), 2D, and 3D polymers, and disordered graphite-like, diamond-like phases, etc., have been reported with a variety of novel polymerized fullerene architectures and a range of unusual properties and potential applications.^2–11

The controllable polymerization of fullerenes to create new materials by using molecular confinement or co-intercalation by template molecules in the reactant has been extensively studied under a wide range of conditions of pressure P and temperature T. Solvated fullerenes, in which solvent molecules separate C₆₀ and C₇₀ molecules, are a series of crystalline materials with high stability, tunable metrics, and functionality. The intercalated solvent molecules play an important role in the construction of solvate lattices with different structures and further influence the phase transition at high P.^12–17 In addition, several metastable doped fullerene polymers have been synthesized under HPHT, such as C₆₀/Ni(OEP),^18,19 C₆₀/CNT,²⁰ and Na₄C₆₀.²¹ The unique initial lattice structures and the different degrees of spatial confinement provided by the dopants lead to different polymeric phases accompanied by distinctive physical and electrical properties.

In 2001, Buga et al.²² reported the P- and T-induced polymerization of the mono-atom endohedral fullerene La@C₈₂ for the first time. Later, Liu and co-workers studied the structural transformations of Sm@C₈₈ (Ref. 23) and solvated Sm@C₉₀ (Ref. 24). The HOMO–LUMO gap decreased with increasing P until the carbon cage deformed, but no phase transition occurred. Pei et al.²⁵ observed for the first time that Lu₃N@C₈₀ underwent a reversible n- to p-type inversion under high pressure, with the p-type semiconductor being stable up to 25 GPa. The role of the intercage metal atom is to restrain the compression of the adjacent bonds and protect the endohedral fullerene against collapse, as well as decreasing and postponing the change in band gap. To date, however, the structure and properties of endohedral fullerenes under high P remain undefined.

This paper focuses on the properties and phases of fullerenes and their derivatives under HPHT. Four topics will be discussed: (i) how fullerenes are covalently incorporated into a polymer phase under HPHT; (ii) the correlation between the structural evolution and the novel properties of the polymeric phase; (iii) the effect of pressure on the polymerization behavior of doped or intercalated fullerenes; (iv) the polymeric fullerenes and their derivatives as new and versatile building blocks to prepare a wide variety of new fullerene-based materials governed by weak intermolecular interactions, showing unprecedented architectures and a wide range of potential applications.

To date, many different strategies have been exploited to prepare polymeric fullerenes. Chemical reactions can link C₆₀ cages either covalently or supramolecularly, and the polymeric phases exhibit outstanding structural, electrochemical, and photophysical properties.²⁶ Photopolymerization,²⁷ pressure- and/or temperature-induced polymerization,^2–4 charge-transfer polymerization mediated by metals,²⁸ electron-beam-induced polymerization,²⁹ and plasma-induced polymerization^30,31 can be used to prepare novel polymeric fullerenes. In this section, we will present a general “panoramic” view of all-carbon polymeric fullerenes under high P and/or high T. Since the discovery of polymeric fullerene synthesized at high P in 1994,³² there has been intense interest in the characterization of the polymeric phase products and in clarification of the mechanisms underlying the process of polymerization, because of the potentially useful physical, biological, and chemical properties of these products.^33,34 In particular, there have been extensive investigations of the influence of P and T treatment on polymerization,³⁵ on infrared spectroscopy (IR) and Raman vibrational properties,³⁶ and on photoluminescence³⁷ and magnetism.³⁸

P-induced polymerization under various T and P has been studied extensively in recent years.² When moderate hydrostatic or quasi-hydrostatic P at high T is applied to pristine C₆₀, irreversible polymerization occurs over a limited range of T and P. As shown in the P vs. T phase diagram (Fig. 1), C₆₀ polymers have orthorhombic, tetragonal, rhombohedral, or cross-linked phases. Recently, single crystals of these polymeric phases have been successfully prepared (the orthorhombic phase in Ref. 39, the tetragonal phase in Refs. 40–43, and the rhombohedral phase in Ref. 44). Under ultrahigh P and T, 3D polymers have been discovered, but, as yet, without a uniform structure.

C₆₀, an electron-deficient polyene, tends to react with other molecules via radical, nucleophilic, or cycloaddition mechanisms. Fullerene can be functionalized via [2 + n] cycloaddition (n = 1, 2, 3, 4) reactions.⁴⁵ In P-induced polymerization, a [2 + 2] cycloaddition reaction between two [6,6] double bonds of near-neighbor C₆₀ cages is the favored reaction pathway (Fig. 2). A new cyclobutane ring is formed by covalent linkage in the polymeric product. On the other hand, the formation of C₆₀ intermolecular bonds is a thermally activated process, and the polymerization reaction is very slow at room temperature. Davydov et al.⁴⁶ have shown that the activation energy for the formation of the C₆₀ dimer is approximately 135 kJ mol⁻¹. However, at ambient P and T between 300 °C and 400 °C, the polymeric product is metastable and reverts to pristine C₆₀.³² Hence, P is necessary for the formation of a stable polymeric product because it reduces the intermolecular distance sufficiently for chemical bond formation to take place.

In the 3D C₆₀ polymer, [3 + 3] cycloaddition bonds form between 2D layers to construct a 3D framework, and it is a fragment of the di-(1.5)(6,10)methanol-cyclodecane molecule.^47,48 All the equivalent 3 and 6 atoms of each C₆₀ molecule are bonded in pairs with the corresponding atoms of the neighboring (x,y) layer molecules. The intermolecular bond lengths are 0.151 nm and 0.161 nm for the [3 + 3] and [2 + 2] cycles, respectively.⁴⁷

The distance between the center-of-mass positions of the two nearest-neighbor C₆₀ in face-centered cubic (fcc) bulk C₆₀ is 10.1 Å.⁴⁹ Generally, it is 9.1 Å in C₆₀ [2 + 2] cycloaddition polymers and about 9.4 Å when a single C–C bond links two cages.^50,51 For example, the intermolecular distance along the C₆₀ orthorhombic chains is 9.14 Å, as characterized by single-crystal X-ray diffraction experiments.³⁹ In confined C₆₀, the nearest-neighbor cage distance approaches a value of 8.45 Å near 10 GPa.⁵²

Alkali-metal-doped fullerene polymerizes under high T or HPHT. However, the framework is constructed using single carbon bonds.⁵³ In the case of endohedral fullerenes with C₈₀, [2 + 2] cycloaddition reactions occur at both the [6,6] and [5,6] bonds of the cage. The [6,6]-regioisomer is a kinetic product, while the [5,6]-regioisomer is a thermodynamic product.⁵⁴

Polymerization changes both the microscopic and macroscopic properties of a material. Insolubility of the polymeric materials in those solvents that do dissolve C₆₀ is direct proof of the polymerization. Polymerization is also confirmed by shifts in the vibrational modes and activation of Raman and infrared modes due to the reduced molecular symmetry as well as to the greater degree of freedom, and these also indicate the degree of polymerization (Fig. 3). According to the symmetry of the displacement patterns, there are 46 distinct intramolecular vibrations for C₆₀. Group theory indicates ten first-order Raman-active modes, including two breathing A_g modes related to the symmetric oscillations of the entire molecule and pentagons together with eight H_g modes.^55–60 In addition, four intramolecular T_1u modes are infrared-active at first order.^55,61,62 The Raman (Fig. 3) and infrared (IR) spectra of the pressure-dimerized state, as well as those of the orthorhombic, tetragonal, and rhombohedral pressure-polymerized phases of C₆₀, have been investigated.⁵⁵

The number of covalent bonds per C₆₀ molecule increases with increasing P and T. Accordingly, the formation of intermolecular bonds shifts the electron density away from the remaining double bonds, which become weaker. In particular, the A_g(2) mode is one of the most sensitive to polymer structure and charge transfer. As shown in Fig. 3, it downshifts from 1470 cm⁻¹ in the monomer to 1465 cm⁻¹ in the dimer, which corresponds to two sp³-like coordinated carbon atoms per fullerene molecule.⁶³ Upon further polymerization, the softening is 12 cm⁻¹ in the linear polymeric chains, 20 cm⁻¹ in the planar tetragonal phase, and 60 cm⁻¹ in the 2D rhombohedral polymers.⁵⁵ Apart from this, the presence of sp³ intermolecular bonds leads to the appearance of new bonds in the range 900–1000 cm⁻¹.

Thermally induced dissociation of the polymeric bonds in C₆₀ polymers can be monitored as a function of time at high T. For example, the band shape and the intensity of the A_g(2) mode change with increasing time.⁶⁴ Novel polymeric C₆₀ phases can also be recognized by Raman spectroscopy measurements. For instance, a branched-chain C₆₀ polymer shows a characteristic mode at 1454 cm⁻¹, a more pronounced splitting of H_g(1), and a new Raman-activated mode around A_g(1).⁶³

The symmetry of C₆₀ is reduced from I_h to D_2h in the orthorhombic and tetrahedral polymers and to D_3d in the rhombohedral phase.⁶⁵ Accordingly, the T_1u(2) mode is primarily polarized in the stretched directions and dramatically downshifts more than 50 cm⁻¹ in the IR spectra of the orthorhombic and rhombohedral polymers. The large distortions of the C₆₀ sphere due to the numbers of cycloaddition connections demand a distorted polymer C₆₀ ball with relaxed atomic positions.

Before discussing the formation and structures of polymeric C₆₀, we briefly review the structural features of C₆₀ at ambient P. C₆₀ molecules exhibit a rotationally disordered fcc structure above 260 K, free rotation at high T, and an orientationally disordered phase at 90 K. Under compression, the crystal undergoes an orientational transition at around 0.4 GPa and ambient T, which transforms into a simple cubic (sc) structure.^2,13,66 During further shortening of the intermolecular distance, [6,6] double bonds open and four-membered rings link nearest neighbor cages. A solid-state mechanochemical method shows that the C4 ring connecting the cages is square rather than rectangular. As shown in Fig. 4, C₆₀ dimers are disordered in position and orientation within an average cubic lattice derived from the monomer.⁶⁷ C₆₀ dimerization begins either within the fcc phase, in which the cage rotates freely, or within the sc structure with nearest neighbors in the P-DB (C=C double bond facing pentagon) or H-DB (C=C double bond facing hexagon) configuration.⁶⁸ This mechanism can be observed even at room temperature at about 1.0 GPa.⁶⁹ However, this thermally activated process is very slow at ambient T. As a result, direct synthesis of pure C₆₀ dimers at HPHT is not yet possible. Typical conditions for synthesizing the C₆₀ dimer are 1.5 GPa, 423 K, and 1000 s, with a yield of about 80%.⁵⁵ When the C₆₀ molecules are linked, the icosahedral symmetry is drastically reduced. As a consequence, the degenerate energy levels near the Fermi level are removed, and thus the optical band gap narrows.⁷⁰

At higher T (<700 K) and P (∼0.7–9 GPa), long linear chains with two four-membered rings per C₆₀ cage form. Parallel straight chains along the original 〈110〉 directions in the ordered orthorhombic structure are observed in powder and single-crystal diffraction results, although disordered branched chains possibly exist as well (Fig. 5). At the same time, a large and rapid change in volume at P ∼ 1 GPa has been observed.⁷¹ The XRD pattern confirms that the nearest-neighbor distance (about 9.1–9.2 Å) in the orthorhombic phase polymer decreases by almost 10% compared with 10.02 Å in pristine C₆₀, owing to the formation of covalent bonds.^39,55 It is interesting to note that two different orthorhombic phase C₆₀ polymers can be formed, corresponding to “high” (above ∼2–3 GPa) and “low” P regions. These structures are recognized by the different orientations of the chains around their axes. The chain links are parallel to the plane in the orthorhombic (O) structure,³³ while they are alternately tilted by angles +μ and −μ from chain to chain in another orthorhombic (O′) phase.^39,72 As a result, the O and O′ phases demonstrate a pseudo-tetragonal Immm space group and orthorhombic Pmnn space group, respectively. It is hard to estimate the angle μ, but lattice energy minimization predicts an optimal value of 29°.⁶⁸

At higher T (>700 K) and P (1.5–9 GPa), four or six four-membered rings per C₆₀ molecule participate in the construction of the 2D C₆₀ polymer. The formation of the polymer layers by cross-coupling of chains appears in two possible unit cells (tetragonal and trigonal). After that, the layers stack in a close-packed arrangement to form the corresponding tetragonal or rhombohedral structures. When [2 + 2] cycloaddition occurs in the (001) plane of the original fcc phase, a 2D tetragonal phase is formed [shown in Fig. 3(e)].² For example, a single crystal of the “tetragonal” polymer was obtained when C₆₀ was treated at 2.5 GPa and 500 °C.⁴³ A single-crystal study identified the tetragonal phase to be a pseudo-tetragonal packing of translationally identical adjacent 2D layers with an orthorhombic space group Immm.⁴³ In this case, the chain orientations are almost parallel. In another case, the chain links are alternately parallel and perpendicular to the stacking axis and exhibit a P4₂/mmc space group. Interestingly, Wågberg et al.⁷⁴ reported that the structural transition from an orthorhombic to a tetragonal structure occurs via an intermediate dimer state as an orthorhombic phase to a dimer and then from this dimer to a tetragonal phase. In addition, the second process is slower than the first.

When treated at higher P (500–800 °C at 5 GPa), C₆₀ polymerized via [2 + 2] cycloaddition in the (111) plane of the original fcc phase into a rhombohedral structure with a hexagonal lattice.^2,32 It has been reported that during decompression from 6 GPa to 2.2 GPa at 873 K, the rhombohedral polymer could transform to a tetragonal phase.⁴⁰ A trigonal symmetry instead of a hexagonal one was confirmed to be the most stable mode by an XRD experiment.⁷⁵ The hexagonal layers stack in a close-packed arrangement of the ABCABC type with space group R3-m.^33,76 All the former four-membered rings rotate 60°, forming another structure with ACBACB stacking type. Single-crystal XRD studies confirm this structure,⁴² and a density functional theory (DFT) calculation demonstrates slightly more stability.⁷⁷

Under much higher P (>13 GPa) at elevated T, 3D polymers are obtained. For example, one superhard phase of a 3D polymer quenched from 13 GPa and 820 K features [3 + 3] cycloaddition bonding between the [2 + 2] cycloaddition-bonded 2D tetragonal layers.^47,48 Single-crystal XRD results demonstrate that the 3D C₆₀ polymer is in a body-centered orthorhombic space group Immm structure (Fig. 6).⁷⁸ The resulting sp³–sp² hybridized system is expected to show metallic conductivity. In contrast, the 2D rhombohedral phase transforms into a disordered phase at around 15 GPa,⁷⁹ and the 2D tetragonal phase transforms into a 3D metastable polymer over 20 GPa at room temperature.⁸⁰In situ Raman spectra strongly support the above conclusion by revealing a transformation from a 2D to a 3D phase.^81–84 Also, the in situ X-ray measurements reveal a first-order irreversible transition of the tetragonal polymeric phase at around 24 GPa.⁸²

Although several studies have demonstrated the existence of 3D C₆₀ polymers, there is still no consensus on the preparation conditions, owing to experimental difficulties and the structural complexity of the products. Generally, under 8–9 GPa, a cubic structure is observed up to about 600–700 K. Disorder increases, and covalent linking extends in all directions with rising T.^5,68 Over 9.5 GPa, a body-centered structure presents.⁷⁸ When T rises above 900–1000 K, the C₆₀ cages start to collapse, and this, combined with the covalent bonding distortion of the cage, makes the structure become amorphous or even graphitic at higher T.⁴ Superhard materials, even cubic diamond crystallites, have also been found.⁸⁵

In conclusion, several tentative phase diagrams of C₆₀ at HPHT have been reported.^2,4,55,76 It should be noted that the polymeric phase formation and purity depend strongly on T, P, and treatment time. Typical conditions for the formation of C₆₀ polymers are as follows: for the 1D orthorhombic phase, 1–1.2 GPa, 550–585 K for 5 h;^39,55 for the 2D tetragonal phase, 2–2.5 GPa, 700–873 K for 0.5–4 h;^42,43,55 and for the 2D rhombohedral phase, 5–6 GPa, 773–873 K for 0.5–1 h.^44,55 Another problem is establishing the stability of the polymeric phases. A quenched intermediate transition or experimental artifacts all possibly exist, but only a few in situ studies have been carried out.

It is important to highlight that heating-then-pressing from rapidly rotating monomers leads to disordered states, because chains form in all directions. Compression benefits the formation of well-ordered 1D or 2D polymers. Nevertheless, pressing-then-heating favors the optimal [2 + 2] cycloaddition route to dimerization, but orientational tuning at high T promotes linear or planar bonding.⁶⁸ However, different paths still give distinct products. For instance, the heating-then-pressing path tends to form rhombohedral polymers.⁸²

When we consider the polymeric phase transition shown in Fig. 1, the orthorhombic-to-tetragonal transformation involves proper rotations of the chains followed by interchain cross-linking in the {100}_c planes. Furthermore, the transformation from an orthorhombic to a rhombohedral phase demands that the chains rotate to a fixed angle, while the tetragonal-to-rhombohedral transition requires bond rearrangement (Fig. 5).⁷³

The physical properties of the low-dimensional phases have already been well studied.^2,4 The essential difference between the C₆₀ monomer and the polymers is the covalent intermolecular linkage in the latter. ¹³C nuclear magnetic resonance (NMR) studies of orthorhombic,^86,87 tetragonal,⁸⁸ and rhombohedral^86,89 C₆₀ clearly demonstrate that these structural differences are based on the distinguishable frequencies of sp² and sp³ hybridized carbon. The relative numbers of sp² and sp³ hybridized atoms can be used to identify the intra- and intermolecular bonds in the polymeric phase. In particular, the sp² peak can be decomposed into lines corresponding to the number of unique carbon atom environments (Fig. 7).^88,89

The reduced symmetry of icosahedral C₆₀ and the formation of intermolecular bonds in the polymeric phase lead to splitting of degenerate molecular orbitals and a narrowing of the HOMO–LUMO gap. The electronic structures of 1D and 2D C₆₀ polymers have been studied theoretically. The 1D linear polymer is a semiconductor with a finite band gap of 1.148 eV and almost pure s-type intermolecular bonding.⁹⁰ The 2D tetragonal and rhombohedral C₆₀ polymers are also semiconductors. With different intermolecular covalent bonds, the tetragonal polymeric and rhombohedral polymeric phases exhibit band gap values of 1.2 eV and 1.0 eV, respectively.³⁴

C₆₀ is reportedly stable up to 950 °C under low-pressure inert gas conditions.⁹¹ High-temperature synchrotron radiation XRD results show that C₆₀ is stable below temperatures of 900 °C at 80 MPa.⁹² However, all the polymers are less stable than the monomer with free C₆₀ molecules. For example, dimers are stable to about 420 K,^93,94 the orthorhombic phase breaks down near 500 K, and the tetragonal and rhombohedral 2D polymers are stable to 550 K.^94,95 Thus, the stability sequence is dimer < 1D polymer < 2D polymer < monomer.³⁴ It should be noted that the thermal “stability limit” is dependent on time scale. The activation energy E_a reflects the stability of the polymers; it is 1.75 eV and 1.9 eV for the dimers and orthorhombic C₆₀ polymeric phases, respectively.^34,94

The degree of polymerization, the polymeric structure, and the ratio of sp² to sp³ C dominate the conductivity of polymerized C₆₀. In contrast to C₆₀, an n-type semiconductor, the 1D plasma-polymerized C₆₀ polymer acts as a p-type semiconductor in field effect transistor (FET) experiments owing to variable-range hopping (VRH).⁹⁶ With the increasing degree of polymerization, the band gap reduces and the electrical conductivity increases.⁹⁵ The conductivity of 2D C₆₀ polymers is anisotropic. In the covalent bond linking layer, it shows a metallic-like behavior. In the weak van der Waals interacting direction of the interlayers, it exhibits semiconductor-like behavior.^97,98 This phenomenon is ascribed to the weak localization of the carriers, and the metallic behavior could be attributed to a self-doping effect. Structural calculations show that lattice defects may induce metallic behavior.⁹⁹ Electrical resistivity measurements show that amorphous polymeric C₆₀ phases are semimetallic, with VRH and semiconducting behavior.¹⁰⁰ Recent first-principles DFT calculations have demonstrated that the structures of 3D C₆₀ polymers in which each molecule is in one of the two standard orientations are analogous to ordered binary-alloy-type structures.¹⁰¹ Furthermore, all the investigated structures in this case are predicted to be metallic, which is supported by an experimental result where a polymeric 56/56 2 + 2 cycloaddition bond is analogous to the Ising antiferromagnetic interaction in a 3D C₆₀ polymer.¹⁰²

Thermal conductivity theoretically increases with increasing P. In addition, it increases with the formation of covalent bonds between near neighboring C₆₀’s because the resulting appearance of new phonon modes contributes to heat transport. Polymeric phase disorder will compensate for the thermal conductivity effect.¹⁰³

C₇₀ molecules undergo quasi-free rotation in a close-packed fcc lattice above room temperature. However, only the five bonds radiating out from each polar pentagon have significant reactivity, which leads to topological constraints on the formation of C₇₀ polymers. The anisotropic C₇₀ molecule exhibits at least three phases: free rotation at high T, uniaxial rotation at intermediate T, and frozen rotation or ratcheting at low T.^3,104

Under compression, the rotational motion is reduced to uniaxial rotation around the molecular axis. Subsequently, the C₇₀ molecules tend to arrange in a line, and a rhombohedral structure is obtained (Fig. 8).^106,107 Regarding symmetry, the C₇₀ monomer with an fcc lattice cannot form a long-range ordered polymeric structure. In contrast, C₇₀ crystallized in the hcp structure can polymerize into an ordered zig-zag chain structure with an orthorhombic lattice.¹⁰⁸

At T between 400 and 500 K and P around 1 GPa, a dimer-rich phase is observed, while the dimer ratio reduces at higher temperatures. Notably, C₇₀ exhibits lower reactivity than C₆₀,¹⁰⁵ so the dimerization of C₇₀ is much slower under high P.

Several polymeric phases have been observed in C₇₀ at higher P and T. Incompatibility between the lattice symmetry and the geometry of the reactive bonds means that an ordered polymeric phase of C₇₀ is hard to obtain at P in the range of 0.82 GPa at T between 350 and 600 K. However, Blank et al.^109,110 and Marques et al.^111,112 reported that C₇₀ with an fcc lattice could transform to an ordered structure under sufficiently high P, which was thought to occur only in hcp C₇₀. It should be noted that the low reactivity and topochemical constraints on the polymerization lead to the slower formation of C₇₀ oligomers and polymers_. Above 700 K, a “disordered cross-linked layered” structure and a 3D polymerized tetragonal structure have been reported.¹⁰⁹

The interaction between fullerenes and solvents leads to the formation of solvated fullerene.¹¹³ C₆₀ and C₇₀ can form solvated crystals with almost all common solvents.¹² Evaporating fullerene solutions is the general preparation method to obtain solvated fullerenes. In 1991, Fleming et al.¹¹⁴ reported solvated C₆₀ and C₇₀ for the first time, inspiring subsequent research on the synthesis and characterization of this series of materials.^115–122 Several conclusions have been drawn based on this research:(a)

Solvates are typical van der Waals complexes with negative excess volumes. Bonding between fullerenes and a solvent will enhance the stability of the solvate.¹²⁰

Furthermore, the solvent molecule also affects the physical properties. Świetlik et al.¹²³ and Graja and Świetlik¹²⁷ reported that the rotation of C₆₀ is hindered by the intercalated solvent molecule, which is reflected in the vibrational spectrum. The influence on the rotation of C₆₀ depends dramatically on the intermolecular interaction and structural arrangement. On the other hand, the symmetry decrease due to interactions between fullerene and solvent molecules enhances the photoluminescent intensity. For instance, the intensity rises about two orders of magnitude in C₆₀∗m-xylene compared with pristine C₆₀ crystal,^12,14 and it is about 30 times higher in C₇₀∗mesitylene than in pristine C₇₀ powder.¹²⁸ The solvent effect on the photoluminescent behavior correlates with the interaction of the cages.¹²⁹

Under high P, the incorporated solvent molecule diversified the phase transformation and polymerization of C₆₀. There are four kinds of P-induced solvent effect:(a)

Through a spacer, such as C₈H₈ in the C₆₀∗C₈H₈ crystal, which effectively shields the P in the interball voids, raising the phase transition P compared with that of C₆₀.¹³⁰

It must also be mentioned that any given solvent molecule generally plays more than one role in the structure control and property tuning of solvated fullerene. For instance, ferrocene acts as a spacer as well as a charge reservoir to tune the polymerization process and promotes a layered structure of the intercalated ferrocene in the C₇₀(Fe(C₅H₅)₂)₂ crystal.¹³⁵ The m-xylene in a C₆₀ solvate plays an important role as a spacer and a bridge by forming covalent bonds with neighbors. It preserves the stability of the highly deformed or amorphous C₆₀ molecules to form a long-range ordered structure under high P.⁸³ Hence, the solvent effects on the structure and properties of solvated molecules are various and diverse under high P.

The P-induced polymeric products can be divided into two categories: copolymerized products of C₆₀ and solvent molecules, or C₆₀ self-condensed materials. C₆₀∗C₈H₈ and C₆₀S₁₆ are in the former category. An anomaly at 0.5 GPa in the P dependences of the molecular vibrational frequencies indicates an orientational ordering transition at room temperature, while an anomaly at 1.3 GPa is attributed to a fullerene–cubane interaction in C₆₀∗C₈H₈.¹³⁶ In C₆₀S₁₆, the peaks from the S₈ rings disappear, and new peaks are observed above 7.2 GPa, suggesting that the sulfur rings break and covalent C–S bonds form between the C₆₀ and solvent molecules.¹³⁷ In the second category, the C₆₀ solvates grown from different solvent molecules exhibit various polymeric products under high P. For instance, C₆₀∗TDAE [where TDAE is tetrakis(dimethylamino)ethylene] (Refs. 138 and 139) and C₆₀(Fe(C₅H₅)₂)₂ (Refs. 140 and 141) both exhibit a one-dimensional polymerized phase above ∼1 GPa and 5 GPa, respectively. C₆₀{Pt(dbdtc)₂} starts to form intercage covalent bonds at 0.5 GPa and a novel polymer, which is coordinated with nearest-neighbor C₆₀ with 10 and 6 sp³-like carbon atoms per molecule above 2.5 GPa.¹³² The solvent molecule in C₆₀∗m-xylene confines the formation of ordered amorphous carbon clusters (OACCs) under high P, as shown in Fig. 10.¹⁶ This study on a small piece of C₆₀ solvate indicates the diversity of the structure and properties of these solvates under high P, and additionally predicts potential applications in a wide range of areas.

Recently, a new way to understand the pressure-induced structure formation and transformation in fullerene solvates under high P has been proposed by studies in which their building blocks have been tailored and their boundary interactions have been tuned.^{83,84,142,143} Solvent molecules with hexagonal carbon rings, such as m-xylene, 1,2,4-trimethylbenzene (TMB), and m-dichlorobenzene (m-Cl), tend to stabilize highly compressed or collapsed fullerene clusters in fullerene solvates and form a long-range ordered structure under compression. In contrast, ferrocene, with its pentagonal carbon ring, generally causes a transformation into an amorphous component together with a carbon cage and leads to the formation of a disordered fullerene solvate structure under high P. Cubane, C₈H₈, differs from these aromatic solvents. It reacts with C₆₀ in solvates under high P, but not all of the C₈H₈ molecules react with the carbon cage.¹⁴² The intercalated C₈H₈ and the reacted C₈H₈ both protect the C₆₀ cages from early collapse under compression. Furthermore, unlike C₆₀∗m-xylene, in which the C₆₀ molecules overcome the confining effect of the solvent molecules and take part in polymerization between the cages,^16,84 the absence of intermolecular bonds between the C₆₀ molecules in C₈H₈/C₆₀ tunes the boundary interactions of the carbon clusters to retain an ordered arrangement up to 45 GPa.

Confinement of fullerene molecules by the intrinsic structure of a carbon nanotube (CNT) in HPHT synthesis is an important strategy for the polymerization of fullerenes.^144–150 C₆₀’s packed inside single-walled CNTs are considered “peapods”¹⁴⁵ and provide an opportunity to study original quasi-1D fullerene phases. Using high-resolution transmission electron microscopy, Smith and co-workers^145,146 discovered self-assembling C₆₀ fullerene chains inside CNTs. Molecular dynamics studies suggest that (9,9) and (10,10) nanotubes with diameter larger than 6.27 Å give the best fit for C₆₀.^147,148 The intercage distance in a CNT was reduced to 8.7 Å under compression to 25 GPa, and a one-dimensional polymer chain was obtained.^20,52 Larger CNTs have more space for the dimensional polymerization of C₆₀, as proven by the observed zigzag structure in C₆₀@CNT(15,15).

In contrast, the anisotropy of C₇₀ is reflected by two distinct orientations with regard to the nanotube axis. The C₇₀ long axis parallel to the nanotube long axis leads to a lying-down orientation, while perpendicular orientations form when the long axis is rotated 90° to the nanotube axis. The orientation of C₇₀ depends on the diameter of the nanotube. C₆₀ peapods fully polymerize at 1.5 GPa and 300 °C,¹⁴⁹ while C₇₀ peapods do not give a polymeric product even at higher P and T (1.5 GPa and 545 °C, or 2.5 GPa and 300 °C).¹⁰⁴ These results show that CNTs impose much stricter constraints on the reactivity of the ovoid C₇₀ than on that of the high-symmetry C₆₀.

Theoretical simulations have predicted several structures of fullerene peapods under high P.¹⁵¹ In 2017, a novel superhard sp³ carbon allotrope was obtained by cold compression of C₇₀ peapods.¹⁵⁰ This so-called V carbon is thermodynamically stable over a wide P range up to 100 GPa and can be recovered to ambient conditions once it is formed, providing a new strategy for synthesizing new superhard materials.

Polymerized alkali-metal-doped fullerenes have attracted much attention because of their hardness and stiffness, metallic behavior,^152,153 and especially superconductivity.¹⁵⁴ 1D and 2D polymeric alkali-metal-doped fullerenes can be obtained under HPHT. For example, AC₆₀ (A = K, Rb, Cs) polymerized to a 1D orthorhombic phase under high T.¹⁵⁵ At 5 GPa and 573 K, the 2D rhombohedral polymers (LiC₆₀)_HP and (NaC₆₀)_HP were synthesized starting from C₆₀ and AN₃ (A = Li, Na). Notably, large-radius alkali metals (Rb or Cs) and stoichiometric X₆C₆₀ stabilized the doped fullerene at much higher P compared with pristine C₆₀. Cs₆C₆₀ retains a monomeric structure until 45 GPa,¹⁵⁶ while Rb₆C₆₀ exhibits a reversible structural transition to a 2D polymeric phase at 35 GPa and 600 K.¹⁵⁷ A 3D polymeric fulleride can be synthesized by compressing 2D Na₄C₆₀.²¹ On the other hand, fullerene will polymerize in a lower P range when doped with smaller alkali metals such as Na or Li.^153,154 In addition, an intercalated alkali metal with a small radius could diffuse in the fullerene lattice such as in the case of the superionic conductivity in the Li₄C₆₀ polymer,^158,159 Na-ion diffusion in Na₂C₆₀ results in the formation of polymeric phases of C₆₀.¹⁶⁰

As pointed out at the start of this paper, pressure is a fundamental thermodynamic variable that can dramatically alter the atomic and electronic structure of materials, resulting in concomitant changes in their properties. Determining the evolution of the geometric and electronic structure of endohedral fullerenes under high P is particularly important because this information is necessary for understanding the electronic conduction mechanism, for discoveries of novel electronic properties, and for fabricating novel fulleride-based materials. On the other hand, the relatively low yield restricts the characterization of endohedral metallofullerenes by standard measurement techniques, while our limited understanding of cluster structure and properties is another challenge. However, in situ measurements at high P in a diamond anvil cell offer a complete workflow to circumvent these problems effectively.

Cui et al.^23,24 have conducted high-P structural transformation studies on the mono-atom encapsulated fullerenes Sm@C₈₈ and solvated Sm@C₉₀. The high-P in situ IR spectra exhibit both red and blue shifts, indicating an anisotropic deformation of the carbon cage under compression. On the basis of these results, together with those of simulations, Cui et al. propose that the cage deforms in the vertical direction, changing from an ellipsoidal to an approximately spherical shape. With further compression, a peanut-like structure appears, before the cage finally collapses under higher P. At the same time, there is a significant P-induced reduction in the band gap because of the deformation of the carbon cage and the enhanced intermolecular interaction (Fig. 11). In this compression process, the Sm atom plays a role in restraining the compression of adjacent bonds. The structural evolution of solvated Sm@C₉₀ under high P is almost the same.²⁴ The solvent molecule plays a role protecting against cage collapse and postpones P-induced changes, contributing to the trapping of the metal atom in an OACC.

Polymeric fullerenes such as polymerized C₆₀ and C₇₀ exhibit unique properties such as high elastic constants and hardness.¹⁶¹ An HPHT-induced polymerization study of the endohedral fullerene La@C₈₂ revealed different polymeric phases such as dimers and 2D and 3D polymers.²² X-ray and electron diffraction results showed increasing density and disorder when the sample was treated at 9.5 GPa and 520 K. A fraction of dimer and oligomers mixed with mainly monomeric La@C₈₂ retained the hcp structure of the pristine sample. The hcp lattice constant decreased dramatically, and small amounts of body-centered tetragonal phase appeared when T was raised to 720 K.

HPHT studies on fullerene and its derivatives have been reviewed. The essential differences among the polymeric phases concern the orientational phenomena that they exhibit and the fullerene cage bonding modes. In view of these, various 1D, 2D, and 3D C₆₀ polymeric structures prepared under HPHT have been studied, utilizing FTIR and Raman spectroscopy and XRD. C₇₀, a nonspherical carbon cage with lower symmetry than C₆₀, shares the general features of the latter with regard to polymerization, but its polymeric structures exhibit significantly fewer phases. A deeper understanding of the polymerization processes and phase diagrams of the fullerenes will help promote the synthesis of new polymeric forms with interesting and useful structural, electrical, magnetic, and photoluminescence properties.

Based on a molecular engineering strategy, solvent molecules, CNTs, and metal atoms could be used to control the arrangement of carbon cages in the lattice. Under HPHT, the intercalated units dominate the evolution of the fullerene configuration, constrain polymerization behavior, and determine the structure and properties of the polymeric product. Further efforts are required so that the phase diagrams of these intercalated/doped fullerenes can be drawn. In addition, studies of intercalated/doped fullerenes at high P and very high T (>1000 °C) present exciting opportunities to discover unique carbon phases and even diamond-like polymeric structures.

Endohedral atoms/clusters greatly enrich the properties of fullerenes, and the resulting fullerene derivatives have properties ranging from semiconducting to metallic. In particular, the tunable electronic spin state inside the fullerene cage makes endohedral fullerenes candidates for use in quantum computing. The application of high P is a potentially useful method to control electronic spin and phonon behavior. However, to date, there have been few high-P studies of fullerenes that have considered this topic. Study of endohedral fullerenes at high P is a very recent field and much remains unexplored. What happens to the interplay of carbon cages and the endohedral species under HPHT conditions? How is charge transferred between the carbon cages and the encaged species when the density increases under compression? What unique properties due to the evolution of molecular and electronic structure will endohedral fullerenes exhibit under high P and indeed high T? The need to answer these questions is a strong motivation for further research in this field.