On the Structural and Vibrational Properties of Solid Endohedral Metallofullerene Li@C₆₀

Abstract

The endohedral lithium fulleride, Li⁺@C₆₀^•−, is a potential precursor for new families of molecular superconducting and electronic materials beyond those accessible to date from C₆₀ itself. Solid Li@C₆₀ comprises (Li@C₆₀)₂ dimers, isostructural and isoelectronic with the (C₅₉N)₂ units found in solid azafullerene. Here, we investigate the structural and vibrational properties of Li@C₆₀ samples synthesized by electrolytic reduction routes. The resulting materials are of high quality, with crystallinity far superior to that of their antecedents isolated by chemical reduction. They permit facile, unambiguous identification of both the reduced state of the fulleride units and the interball C-C bonds responsible for dimerization. However, severe orientational disorder conceals any crystal symmetry lowering due to the presence of dimers. Diffraction reveals the adoption of a hexagonal crystal structure (space group P6₃/mmc) at both low temperatures and high pressures, typically associated with close-packing of spherical monomer units. Such a situation is reminiscent of the structural behavior of the high-pressure Phase I of solid dihydrogen, H₂.

1. Introduction

Following the serendipitous discovery of the fullerene family of molecular carbon nearly 40 years ago, research on the chemical, physical, and materials properties of its members and those of their derivatives has proceeded unabated. In brief, there are four distinct ways of manipulating the molecular and crystalline forms of the fullerenes. Firstly, direct chemical functionalization of the π-electron molecular framework leads to a rich organic chemistry, affording numerous derivatives of both C₆₀ and its higher fullerene antecedents that have applications in diverse fields from molecular electronics to solar cells [1].

Secondly, reductive intercalation of metal ions into the voids of the crystalline structures of fullerenes—most prominently, that of C₆₀—leads to sensitively modifying the electronic structure and has resulted in a plethora of fulleride salts with M_xC₆₀ stoichiometry and x ranging from a low value of 1 (e.g., in RbC₆₀) to a high value of 12 (e.g., in Li₁₂C₆₀). In particular, exohedrally modified fulleride materials with A₃C₆₀ (A = alkali metal) stoichiometry have provided examples of high-T_c molecular superconductors with transition temperatures as high as 38 K and with the zero-resistance performance surviving to record high magnetic fields (exceeding 90 T). Of significance is also the unconventional, highly correlated nature of superconductivity, which is proximate to an antiferromagnetic Mott insulating state out of which it emerges upon lattice contraction via the application of chemical and/or physical pressure [2,3,4,5].

Thirdly, the electronic structure can be tuned by on-ball substitution of heteroatoms such as B or N. The most prominent derivative has been the azafullerene radical, C₅₉N^•, which dimerizes to afford the non-magnetic (C₅₉N)₂ moiety [6]. Further manipulation of the properties of such heterofullerenes can be achieved by exohedral metal intercalation, resulting in salts such as K₆C₅₉N, in which electron doping is achieved synergistically both by on- and in-between-the-balls donors [7].

Finally, a large family of fullerene derivatives can be accessed by endohedral encapsulation of atomic or molecular units to afford A@C_n moieties. For a long while, this field of fullerene research was dominated by metal encapsulation of higher fullerenes [8,9], with bulk availability of endohedral C₆₀ units remaining elusive. This situation has changed dramatically following two significant discoveries: (a) the development of the “molecular surgery” synthetic technique [10], which has led to numerous endohedral C₆₀ fullerenes incorporating noble gas atoms (e.g., He, Ne, Ar, Kr) or small molecules (e.g., H₂, HF, CO, H₂O, CH₄) [11,12] and (b) the successful synthesis of fullerene salts, (Li⁺@C₆₀)X⁻ (X⁻ = PF₆⁻, SbCl₆⁻), which comprise Li⁺@C₆₀ endohedral molecular ions and X⁻ spacer units, which serve a charge-balancing role [13,14,15]. Remarkably, both chemical and electrochemical reduction of the Li⁺@C₆₀ cation lead to the formation of the neutral radical anion, Li⁺@C₆₀^•− [16,17], which, in analogy with the on-ball-substituted azafullerene C₅₉N^• radical, dimerizes to afford (Li@C₆₀)₂, isostructural and isoelectronic with the (C₅₉N)₂ molecule [18,19].

Structural and electronic characterization of the (Li@C₆₀)₂ solid obtained by chemical reduction confirmed its dimeric structure in the solid state [17]. However, the as-obtained material lacked good crystallinity, resulting in moderate quality diffraction and spectroscopic data. Attempts to sublime the material led to the escape of the metal ion from the endofullerene space, precluding any improvement in crystallinity, in contrast to its azafullerene antecedent [19]. Here, we report on the structural and vibrational properties of the (Li@C₆₀)₂ solid obtained by electrochemical reduction—this is a highly crystalline material, which permits detailed study of the structural and electronic properties. While vibrational spectroscopy readily identifies the vibrational signatures of interball C-C bond formation consistent with dimerization, the material is shown to adopt a highly symmetric hexagonal-close-packed crystal structure—typically associated with monomer sphere packing—as a result of disorder both at ambient pressure down to 3.4 K and at ambient temperature up to 13.1 GPa.

2. Results and Discussion

2.1. Vibrational Properties

The Raman spectrum of (Li@C₆₀)₂ obtained by electrochemical reduction of the (Li⁺@C₆₀)(PF₆^–) precursor, along with those of monomeric C₆₀, (Li@C₆₀)(PF₆), and dimeric (C₅₉N)₂ solids, is illustrated in Figure 1—the corresponding molecular units are generated by VESTA [20]. The frequencies of the observed Raman peaks are summarized in Table S1. In the case of the monomeric C₆₀ and (Li@C₆₀)(PF₆) fullerene materials, the 785 nm laser line was used for excitation to prevent photoinduced polymerization upon illumination with visible or ultraviolet light [21]. On the other hand, owing to the stability of the dimeric (Li@C₆₀)₂ and (C₅₉N)₂ fullerenes against photoinduced polymerization, the 633 nm line of a He-Ne laser was used for their excitation to take advantage of the improved throughput of the Raman spectrometer in the corresponding spectral region. The Raman spectra of C₆₀ solid and (Li@C₆₀)(PF₆) are nearly identical to each other with small frequency shifts (≤3 cm⁻¹) between the corresponding peaks and are compatible with the monomeric nature of the C₆₀ units. Both spectra are dominated by the ten Raman-active intramolecular C₆₀ modes, attributed to the two totally symmetric A_g(1) and A_g(2) modes and the eight fivefold-degenerate H_g(1)-H_g(8) of the C₆₀ molecule in icosahedral I_h symmetry [22]. The weak crystal field in solid C₆₀ causes only the splitting of the radial H_g(1) mode into two components (at 266 and 272 cm⁻¹), while the appearance of a weak peak at 1458 cm⁻¹ can be attributed to the presence of traces of polymerized C₆₀ (A_g(2) mode of polymeric chains [23]).

The Raman spectrum of dimeric (C₅₉N)₂ solid is far richer in structure compared to the monomeric systems, as the cage distortion and the concomitant lowering of the molecular symmetry upon dimerization cause the appearance of additional Raman peaks. These originate from the splitting of the degenerate H_g modes and the activation and splitting of inactive vibrational modes of the C₆₀ molecule in I_h symmetry [24,25]. The plethora of Raman split components appearing in the spectrum of (C₅₉N)₂ is in fair agreement with those earlier reported in the literature and assigned to intramolecular C₆₀ modes in I_h symmetry [24]. The same assignment of the observed Raman peaks in the present work was also adopted in Table S1. As inferred from Figure 1 and Table S1, the similarity of the Raman spectrum profile and the frequencies of the constituent Raman peaks of (Li@C₆₀)₂ obtained by electrochemical reduction with those of dimeric (C₅₉N)₂ is evident, establishing its dimeric nature unambiguously. Moreover, comparison of the overall Raman spectrum profile of the present (Li@C₆₀)₂ sample with that of the neutral Li@C₆₀ obtained by chemical reduction of the Li⁺@C₆₀ cation by decamethylferrocene (Figure S1) clearly illustrates the stronger intensities and narrower lineshapes of the split Raman components in (Li@C₆₀)₂ obtained by electrochemical reduction—this is in addition to the similarities regarding the appearance of the Raman peaks associated with the dimerization of the C₆₀ cages. The present results are compatible with the superior crystallinity of the present material, in accordance with its powder XRD data (vide infra), contrary to the situation encountered in (Li@C₆₀)₂ obtained by chemical reduction, where a small fraction of monomers is also present [17].

The frequency of the A_g(2) pentagonal pinch mode of the C₆₀ molecule deserves further attention. In particular, the A_g(2) frequency in C₆₀-based materials is extremely sensitive to both doping and intermolecular covalent bonding [23,24,25,26,27,28]. In monomeric alkali fullerides, it downshifts quasi-linearly by 6–7 cm⁻¹ per electron transferred from the metal atoms to the C₆₀ units [26,27,28], due to the elongation of the intramolecular bonds induced by the charge transfer process and the resulting softening of the force constants [29,30]. Moreover, in neutral polymeric C₆₀-based fullerene materials—including dimerized systems—the A_g(2) mode also softens due to the intermolecular covalent bond formation and the concomitant decrease in the mean intramolecular bond strength. The observed softening depends strongly on the number of the sp³-like coordinated carbon atoms per molecular cage [31]. Accordingly, and as evident in Table S1 (numbers in bold), there is a softening of the A_g(2) peak in neutral (C₅₉N)₂ compared to that in C₆₀ solid by ~6 cm⁻¹, which is attributed to the formation of the intermolecular covalent bond due to dimerization [25]. In the case of (Li@C₆₀)₂, the softening of this peak is even larger, ~13 and ~10 cm⁻¹ relative to monomeric C₆₀ and (Li@C₆₀)(PF₆), respectively. This can be understood in terms of the combined effect of covalent intermolecular bonding and uptake of one electron by the C₆₀ cage in (Li@C₆₀)₂—here, electrons are more delocalized over the radical carbanion cage than in the case of (C₅₉N)₂, where the electron pairs are localized on the nitrogen atoms [17].

The dimeric nature of (Li@C₆₀)₂ obtained by electrochemical reduction is further supported by its IR transmission spectrum illustrated in Figure 2 in comparison with that of monomeric C₆₀ solid. More specifically, the IR spectrum of C₆₀ solid is dominated by the four IR active modes of the C₆₀ molecule in I_h symmetry {F_1u(1)-F_1u(4)}, appearing at 525, 576, 1182, and 1428 cm⁻¹, in excellent agreement with the literature [32]. The additional weaker spectral features are due to forbidden fundamentals, activated by the distortion of the molecular I_h symmetry in the C₆₀ crystal. As in the case of the corresponding Raman spectrum, in the IR spectrum of (Li@C₆₀)₂, several split components appear due to intercage bonding and symmetry lowering, similar to the case of (C₅₉N)₂ [33]. Despite similarities of the IR spectrum with that of monomeric C₅₉HN, the appearance of characteristic features in the frequency region 800–850 cm⁻¹ (marked accordingly in Figure 2) provides a fingerprint of the dimerization of the molecular cages [33].

2.2. Structural Properties at Ambient Pressure

The synchrotron X-ray powder diffraction patterns of the (Li@C₆₀)₂ sample recorded at 300 and 3.4 K at ambient pressure are shown in Figure 3. An immediate observation is that they appear strikingly similar to those reported for as-prepared solid (C₅₉N)₂, in terms of both the position and the relative intensity of the diffraction peaks present [18,34]. As-prepared (C₅₉N)₂ crystallizes in a hexagonal lattice (space group P6₃/mmc) with lattice constants at ambient temperature, a = 9.97 Å and c = 16.18 Å [18]. In a remarkable analogy, the patterns of (Li@C₆₀)₂ prepared by electrochemical reduction are also consistent with the material adopting an isostructural hexagonal closed-packed (hcp) arrangement of the fullerene units with a somewhat smaller (c/a) anisotropy (space group P6₃/mmc, at 300 K: a = 10.04374(10) Å and c = 16.13451(19) Å; at 3.4 K: a = 9.9560(3) Å and c = 16.0317(5) Å). Excellent refinements of the diffraction profiles are achieved by the LeBail pattern–decomposition technique using the GSAS suite (at 300 K: R_wp = 1.88%, R_exp = 1.48%; at 3.4 K: R_wp = 2.23%, R_exp = 1.43%) (Figure 3).

On the other hand, the measured profiles of the present (Li@C₆₀)₂ sample contrast sharply with those of both (Li@C₆₀)₂ prepared by chemical reduction route [17] and sublimed (C₅₉N)₂ [19]. The former displays a powder pattern, which can be indexed on a pseudo-cubic unit cell with a lattice size of ~14.15 Å (interfullerene separation ~10.01 Å) [17], whereas the latter provides clear evidence of a dimerized molecular structure and a lowering of crystal symmetry to monoclinic [19].

The apparent hcp crystal structure (Figure 3a inset) of (Li@C₆₀)₂ prepared by electrochemical reduction is an archetypal packing arrangement of monomer units. Therefore, it is counterintuitive and contrasts sharply with the dimerization of the molecular units that has been unambiguously established by vibrational spectroscopy (vide supra). In the P6₃/mmc space group, the Li@C₆₀ balls reside in the 2d positions, (⅔, ⅓, ¼) and (⅓, ⅔, ¾). As a result, their center-to-center distance at ambient temperature is 9.94 Å, implicit of the absence of interball bond formation. Due to the translational symmetry of the hexagonal unit cell, there are six nearest neighbors at this distance of 9.94 Å, together with six additional ones in the ab pane at 10.04 Å. These are close to the C₆₀—C₆₀ distances in monomeric C₆₀ (~10 Å) and are longer than those (~9.34 Å) in typical dimerized C₆₀ structures like that of RbC₆₀ [35,36,37].

The inconsistency with the vibrational results may be resolved if we introduce orientational disorder of the dimer units residing on the hcp sites. To this effect, we recall a much-celebrated analogous situation—namely, that of the Phase I of solid dihydrogen, H₂ [38]. At room temperature, this phase is stable between 5 and 190 GPa and comprises freely rotating dihydrogen molecules, forming an hcp crystal structure (space group P6₃/mmc). We therefore attempted Rietveld refinements of the diffraction data by modeling the pronounced orientational disorder of the fullerene moieties in terms of spherical shell scatterers placed at the 2d sites of $\overline{6}$m2 symmetry in the hcp lattice. In the course of the refinements, we found it necessary to also introduce scattering density at the 2a sites of $\overline{3}$m. symmetry, presumably reflecting the presence of residual CH₂Cl₂ solvent molecules intercalated during the electrochemical synthetic protocol. Rietveld refinements now proceeded smoothly, resulting in a stoichiometry of (Li@C₆₀)₂⋅(CH₂Cl₂)_0.81(2) for the material. We note that an analogous situation was encountered in the Rietveld refinements of the diffraction data of as-prepared hcp-structured azafullerene solid, whose stoichiometry was established as (C₅₉N)₂⋅(CS₂)_1.0 [34]. The refinement results are shown in Figure 4 and Table S2.

Figure 5a,b shows the temperature dependence of the unit–cell volume and the lattice parameters a and c, as extracted from the LeBail analyses of the diffraction data. The smooth evolution in lattice size confirms the absence of any structural phase transition down to 3.4 K. The only apparent effect is a decrease in the lattice anisotropy, as measured by the (c/a) ratio, which monotonically increases with decreasing temperature, approaching a value just larger than 1.61 at 3.4 K (Figure 5c). This is still smaller than the ideal value of 1.633 in perfect hexagonal close-packed architectures, but the decreased anisotropy on cooling may reflect a partial freezing out of the orientational disorder of the fullerene units.

2.3. Structural Properties at High Pressure

Synchrotron X-ray powder diffraction data of (Li@C₆₀)₂ were collected at ambient temperature and at selected pressures between 0.5 and 13.1 GPa. As the pressure increases, the diffraction profiles show gradual Bragg peak shifting towards higher 2θ angles consistent with unit cell contraction; this is accompanied by some peak broadening due to increasing inhomogeneities upon pressurization (Figure 6). No evidence of the occurrence of any structural phase transition is found in the pressure range of the present experiments. All diffraction patterns index with the same hexagonal unit cell (space group P6₃/mmc) found at all temperatures at ambient pressure. Representative LeBail refinements at base and highest pressures are included in Figure 6 (at 0.5 GPa: a = 9.89824(18) Å and c = 15.9333(3) Å; R_wp = 0.35%, R_exp = 4.60%; at 13.1 GPa: a = 9.2495(5) Å and c = 15.0437(11) Å; R_wp = 0.29%, R_exp = 4.94%). The evolution of the (Li@C₆₀)₂ unit cell dimensions upon pressurization was investigated by LeBail refinements of the diffraction profiles (Figure 7a,b). Some spurious weak peaks, which were present in the diffraction datasets when the sample was introduced into the diamond anvil cell (DAC)—but were absent for the same sample batch outside the pressure cell—were excluded from the refinements. Figure 7c displays the change in the lattice anisotropy of the hcp structure, as monitored by the evolution of the (c/a) ratio. Upon pressurization, c/a increases monotonically at the beginning but eventually saturates above ~10 GPa towards a value of ~1.626, close to the ideal value of 1.633 expected for perfect hcp stacking of solid spheres. Comparable behavior was also observed before for the isostructural dimeric (C₅₉N)₂ solid, in which the (c/a) ratio approached ~1.633 at ~6.5 GPa [18]. Such behavior implies that, at these elevated pressures, the intradimer and interdimer C-C distances in the compressed material have become comparable. Notably, this occurs in the same pressure range that the interdimer center-to-center separations in (Li@C₆₀)₂ have decreased to values on the order of ~9.2–9.3 Å. Such close interball contacts have provided the signature of single C-C bond formation in fullerene-derived polymers [35,36,37] before.

The pressure dependence of the unit cell volume of (Li@C₆₀)₂ was next investigated using the semi-empirical second-order Murnaghan equation-of-state (EoS):

$$P=\left(\frac{K_0}{{K_0}^{\prime }}\right)[{\left(\frac{V_0}{V}\right)}^{{K_0}^{\prime }}-1]$$

(1)

where K₀ is the atmospheric pressure isothermal bulk modulus, K₀′ is its pressure derivative (≡ dK₀/dP)_P=0, and V₀ is the unit–cell volume at zero pressure (approximated by the ambient pressure datum at 1 atm; here, V₀ = 1409.58 ų). The EoS fit to the V(P) data (see Equation (S1) in the Supplementary Materials) of (Li@C₆₀)₂ up to a pressure of 13.1 GPa results in values of K₀ = 10.8(3) GPa (leading to an initial volume compressibility, κ = −dlnV/dP = 0.093(3) GPa⁻¹) and of K₀′ = 11.4(2) (Figure 7a). The measured volume compressibility of (Li@C₆₀)₂ is higher than those of (C₅₉N)₂ (κ = 0.046(2) GPa⁻¹) and of the monomeric C₆₀ (κ = 0.055 GPa⁻¹) and (Li⁺@C₆₀)(PF₆) (κ = 0.0571(4) GPa⁻¹) [15,18,39] solids, implying more flexible nearest-neighbor bonds and somewhat less tight crystal packing. Modified EoS equations, in which K₀ and its pressure derivative, K₀′, were replaced by K_x and K_x′ (x = a or c) parameters, were used to fit the corresponding lattice axes data to extract initial axial compressibilities, β_x ( = 1/(3K_x) GPa⁻¹, x = a or c) (Figure 7b). As anticipated by the earlier discussion on the pressure dependence of the (c/a) ratio, we find that the compressibility along the c-axis (β_c = 0.0086(2) GPa⁻¹) is smaller than that along the a-axis (β_a = 0.0112(3) GPa⁻¹). Such a pressure response was also observed for (C₅₉N)₂ and was then interpreted as a consequence of the intradimer bonds being oriented along the (⅓, ⅔, ½) direction and therefore inclined at a larger angle to the basal plane than to the c-axis [18].

3. Materials and Methods

The (Li⁺@C₆₀)(PF₆^–) sample used as a precursor in the present work was purchased from Idea International Corporation. Its electrochemical reduction to afford (Li@C₆₀)₂ was performed according to the experimental protocol described in the literature before [16].

Raman spectra of (Li@C₆₀)₂ were recorded in backscattering geometry using a LabRam HR (HORIBA, Kyoto, Japan) micro-Raman spectrometer, equipped with a Peltier cooled CCD detector. Laser lines at 633 and 785 nm were alternatively used for excitation, focused on the samples by means of a 100× objective at a power lower than 0.3 mW to avoid any laser-heating-induced effects.

Infrared (IR) transmission spectra of the samples under argon atmosphere were collected by means of a JASCO (Tokyo, Japan) FT-IR 6200 spectrometer.

Synchrotron X-ray powder diffraction measurements of (Li@C₆₀)₂ at various temperatures between 3.4 and 300 K at ambient pressure were performed with the high-resolution powder diffractometer on beamline BL44B2 at SPring-8, Hyogo, Japan using a helium-flow cryostat [40]. The powder sample was sealed in a 0.5 mm diameter thin-wall glass capillary under He gas. The wavelength was 0.800312(1) Å. LeBail and Rietveld analyses were performed with the GSAS and Fullprof suites of refinement programs [41,42,43].

Synchrotron X-ray powder diffraction measurements of (Li@C₆₀)₂ up to 13.1 GPa at ambient temperature were conducted at the BL10XU beamline, SPring-8, Hyogo, Japan. The powder X-ray diffraction data (monochromatized beam with λ = 0.41277(5) Å, beam size = 40 μm) were collected on an image plate detector (Rigaku (Tokyo, Japan) R-AXIS IV⁺⁺, 300 × 300 mm², and 0.10 mm pixel size) while oscillating the sample [44]. Masking of the Bragg reflections of the DAC and integration of the 2D diffraction images to afford one-dimensional diffraction profiles were processed with the IPanalyzer software [45]. LeBail analysis of the diffraction profiles was performed with the GSAS suite of refinement programs [41,42]. A DAC [46,47], equipped with single-crystal diamonds of 500 μm culet diameter and with an 80 μm-thick stainless-steel gasket, was used for pressure generation. Powder samples were introduced into 250 μm diameter holes made at the center of the gasket, together with a ruby sphere and helium gas (solidification pressure ∼12.1 GPa at room temperature, pressure uncertainty less than 0.5% at the highest pressure of the present experiment [48]) as pressure marker and pressure medium, respectively. The applied pressure was measured by the ruby–fluorescence method [49]. All measured pressure values are quoted to one decimal point.

4. Conclusions

In conclusion, we have presented a detailed structural and vibrational characterization of the dimerized endohedral metallofullerene, (Li@C₆₀)₂, in the solid state. The material used in the present work was obtained by electrochemical reduction of the trifluoromethanesulfonyl imide precursor salt, (Li⁺@C₆₀)(TFSI⁻) [16]. Its highly crystalline nature permits facile identification of spectroscopic features related to interfullerene C-C bonding. Remarkably, the crystal structure of the material remains hexagonal close-packed (space group P6₃/mmc) both at low temperatures and ambient pressure and at high pressures and ambient temperature, concealing the effects of dimerization and the anticipated lowering in symmetry. High orientational disorder (together with the presence of some residual dichloromethane solvent) can account for the apparent high symmetry. A comparable orientational disorder scenario has been used to rationalize the hcp Phase I of solid dihydrogen (H₂) at high pressures.