Metal-containing fullerenes, and particularly trimetallic nitride template endohedral metallofullerenes (TNT-EMFs), elicit increasing attention not only for their fascinating structure and the possibility to stabilize metallic nitrides, but also for their outstanding electronic and optical properties. TNT-EMFs are noteworthy electron-acceptors. They also possess larger absorption coefficients than C₆₀ in the visible region, making them promising candidates for replacing the well-known phenyl-C₆₁-butyric acid methyl ester (PCBM) in bulk heterojunction (BHJ) solar cells. They could then be used as potential auxiliary materials for singlet-exciton dissociation at the donor-acceptor interfaces, providing charge transport pathways across the semiconducting layer. Charge-carrier transport and light-emission efficiencies are crucially related to the molecular organization and degree of ordering, i.e. in thin film morphology (nanostructuration). It is widely recognized that self-organization by the formation of low-dimensional liquid-crystalline (LC) phases is a key strategy to control ordering and structuring of organic semiconductors as it helps to reduce or even suppress defect formation. Chemical functionalization of TNT-EMFs appears therefore as an original and promising approach to reach self-organization and to obtain processable materials. If a wide number of LC fullerenes have been studied since the early 90s, none study concerns the TNT-EMFs family.
Our main goal is to design and characterize TNT-EMF-based liquid crystals (1) and compare both their mesomorphic and photophysical behaviours with their related C₆₀ counterpart (2) (Figure 1). In a first report [1] both fullerene entities (C₆₀ and Y₃N@C₈₀) are chemically linked to two oligo(phenyleneethynylene) arms (OPE for short) mandating both self-organizing and absorbing properties. Experimentally, both the [5,6] and [6,6] types of double bonds can be potential reactive sites of TNT fullerenes, although the regioselectivity is usually very high.

NMR is a powerful technique to distinguish between these two addition patterns, but the low solubility of most of the TNT-EMF derivatives usually prevents their NMR characterization and the use of ¹³C labelled samples or solvent mixtures with CS2 are required. The presence of twelve long alkyl chains however made possible the complete NMR characterization of the Y₃N@C₈₀ derivative 1 in pure chloroform and confirmed that the product is a [6,6]-bridged fulleroid.
Photophysical investigations were performed in the laboratory of Prof. Paola Ceroni, U Bologna. The absorption spectra of 1 and 2 in toluene solution show the contribution of the two constituent chromophores: the band at 326 nm is mainly due to the OPE units, while the absorption at λ>380 nm is characteristic of the fullerene core (Figure 2a). It is worth noting that the molar absorption coefficient of 1 in the visible region is much higher than that of 2 and extends up to 750 nm because of the endohedral fullerene core.

The emission spectra of 1 and 2 in de-aerated toluene solutions (Figure 2b) show two bands: the first one at ca. 365 nm, which is centered on the OPE moiety and is strongly quenched (>20 times) compared to model compound dOPE, and the second one in the 680-900 nm region, which can be attributed to the fullerene core (Table 1). The quenching of the OPE fluorescence can be attributed to a 100% efficient energy transfer. Indeed, the same fullerene emission intensity was recorded upon excitation of two iso-absorbing solutions of 1 at 320 nm, where most of the light is absorbed by the pending OPEs, and at 405 nm, where only the fullerene absorbs light (Figure 1). The same result was observed for compound 2. Therefore, the two OPE units act as extremely efficient light-harvesting antennae for the sensitization of the fullerene emission.

The most striking difference between the two derivatives concerns the fullerene core emission: in the case of 2, a very weak fluorescence (Φ_em = 3 x 10-4) with a lifetime of 1.5 ns is observed, as expected for C₆₀ derivatives. In contrast, compound 1 exhibits outstanding luminescence properties in the near-IR region, even larger than the non-functionalized model Y₃N@C₈₀. As reported in Table 1, the emitting excited state of 1 is: (i) slightly lower in energy compared to the non-functionalized endohedral fullerene, (ii) quite highly emitting (0.08 in de-aerated solution), (iii) extremely long-lived (16 μs at 298 K, 20 times higher than Y₃N@C₈₀ and 13 ms at 77 K), and (iv) highly sensitive to the presence of dioxygen in fluid solution (vide infra). The same emission band is observed in the solid phase, both in the amorphous and LC mesophase, with a strong quenching by dioxygen.
Dioxygen quenches efficiently the emission of 1 also in solution with a rate constant kq = 8 x 108 M^-1 s^-1. This value is slightly lower than that of the pristine Y₃N@C₈₀ (kq = 2 x 109 M^-1 s^-1), consistent with an embedding of the fullerene core by the OPE units of 1. Quenching by dioxygen (Figure 3) leads to sensitization of 1O2 emission at 1270 nm with a quantum yield of 1.0 and 0.7 for 1 and Y₃N@C₈₀, respectively (Table 1). The fluorescence of compound 2 is not quenched by dioxygen because of the very short lifetime, which prevents dynamic quenching. However, 2 can sensitize 1O2 by its lowest lying triplet excited state (ESI) with efficiency close to 1, as expected for C₆₀ derivatives.
Quenching of the luminescence of 1 has been observed upon addition of ferrocene with kq = 6 x 109 M^-1 s^-1, a value lower than that of Y₃N@C₈₀ (kq = 1 x 10¹⁰ M^-1 s^-1) because of the embedding effect of the OPE units of 1. The quenching occurs by photoinduced electron-transfer from ferrocene to 1 since the ferrocene is easy to reduce and not to oxidize and it has no excited state lower than that of 1.

DFT calculations were performed in collaboration with Prof. Francesco Zerbetto, U Bologna. To further explore the nature of the long-lived emitting excited state, density functional theory (DFT) calculations were performed (Figure 3) on Y₃N@C₈₀ and compared to those obtained for Sc₃N@C₈₀, a closely related EMF with similar absorption spectrum, which does not emit.
The DFT optimized geometries at B3LYP/6-311G*/ECP level agreed with the structures proposed earlier. Both Sc₃N and Y₃N are planar. The amount of intramolecular charge transfer is larger for Y₃N@C₈₀ than Sc₃N@C₈₀, with the Y₃N doped-cage that receives 3.84(4.99) electrons while the corresponding value for the Sc₃N doped-cage is 2.98(3.41) with the B3LYP(HSE06) functional. The large values of charge-transfer suggest that molecule-cage Coulomb interactions play an important role in the structure and properties of these endohedral clusters. Further time-dependent DFT and ZINDO/S calculations showed that the lowest electronically excited singlet state S1 of both Sc₃N@C₈₀ and Y₃N@C₈₀ is mainly a HOMO-LUMO transition. The HSE06/6-311G*, CAM-B3LYP/6-311G* and ZINDO/S results locate S1 of Y₃N@C₈₀ at 2.02, 2.55, and 1.72 eV. For Sc₃N@C₈₀, the corresponding values are 1.72, 2.20, 1.74 eV. A crucial difference between Sc₃N@C₈₀ and Y₃N@C₈₀ emerges when the molecular orbitals involved in S1 are inspected visually. Figure 4 shows the HOMO’s and LUMO’s of the two endohedral clusters. In the non-emitting, short-lived S1 state of Sc₃N@C₈₀ the electron excitation is spread over the entire molecule: both the cage and the endohedrally confined Sc₃N participate in the excitation.