Photoreactions of Sc₃N@C₈₀ with Disilirane, Silirane, and Digermirane: A Photochemical Method to Separate I_h and D_5h Isomers

Abstract

Under photoirradiation, Sc₃N@I_h-C₈₀ reacted readily with disilirane 1, silirane 4, and digermirane 7 to afford the corresponding 1:1 adducts, whereas Sc₃N@D_5h-C₈₀ was recovered without producing those adducts. Based on these results, we described a novel method for the exclusive separation of I_h and D_5h isomers of Sc₃N@C₈₀. The method includes three procedures: selective derivatization of Sc₃N@I_h-C₈₀ using 1, 4, and 7, facile HPLC separation of pristine Sc₃N@D_5h-C₈₀ and Sc₃N@I_h-C₈₀ derivatives, and thermolysis of Sc₃N@I_h-C₈₀ derivatives to collect pristine Sc₃N@I_h-C₈₀. In addition, laser flash photolysis experiments were conducted to elucidate the reaction mechanism. Decay of the transient absorption of ³Sc₃N@I_h-C₈₀^* was observed to be enhanced in the presence of 1, indicating the quenching process. When Sc₃N@D_5h-C₈₀ was used, the transient absorption was much less intensive. Therefore, the quenching of ³Sc₃N@D_5h-C₈₀^* by 1 could not be confirmed. Furthermore, we applied time-dependent density functional theory (TD-DFT) calculations of the photoexcited states of Sc₃N@C₈₀ to obtain insights into the reaction mechanism.

1. Introduction

Endohedral metallofullerenes (EMFs) have been investigated extensively because of their fascinating structures based on electron transfer from encapsulated metal species to carbon cages [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17]. Among EMFs, trimetallic nitride template endohedral metallofullerenes (TNT-EMFs) constitute a major EMF family for which extensive studies have been conducted to ascertain and apply their remarkable properties [2,9,10]. In fact, Sc₃N@I_h-C₈₀ has been well studied among TNT-EMFs because of its high production yield [9]. A few years after the discovery of Sc₃N@I_h-C₈₀, the D_5h isomer of Sc₃N@C₈₀ was isolated and characterized to demonstrate its higher reactivity than that of the I_h isomer [18,19]. For the synthesis of Sc₃N@C₈₀, however, separation of the I_h and D_5h isomers by high-performance liquid chromatography (HPLC) is not efficient because the retention times of these isomers are mutually similar using commercial HPLC columns.

To date, many chemical procedures without HPLC separation have been reported for separation of mixtures of fullerenes based on their chemical reactivity differences [20]. For example, Diels–Alder reactions using a cyclopentadiene-functionalized resin followed by retro-addition was used to facilitate the separation of the I_h and D_5h isomers of Sc₃N@C₈₀ and Lu₃N@C₈₀ [21]. Selective complexation procedures were developed using aminosilica and Lewis acids that precipitate with some EMFs [22,23]. These methods enabled separation of Sc₃N@I_h-C₈₀ in gram quantities. More recently, a selective oxidation procedure using a ferrocenium salt was applied to separate both isomers based on differences in their oxidation potentials. This procedure involves sequential column chromatographic separation of the unreactive Sc₃N@I_h-C₈₀ and the oxidized Sc₃N@D_5h-C₈₀, which were subsequently recovered by reduction [24]. Although Sc₃N@D_5h-C₈₀ and Sc₃N@C₆₈ were obtained as the same fraction in this method, it was subsequently reported that Sc₃N@D_5h-C₈₀ was separated from Sc₃N@C₆₈ based on the predominant reactivity of the latter in methano-derivatization using a tosyl hydrazone reagent [25].

For our ongoing study of fullerene chemistry, disilirane 1 has been used as a versatile derivatizing reagent [15,16]. In general, 1 reacts with EMFs efficiently under visible light to afford the corresponding silylated EMFs. Furthermore, EMFs that exhibit less negative first reduction potentials are reactive toward 1 under both thermal and photochemical conditions [15,16]. These results led us to examine the reactivity of Sc₃N@D_5h-C₈₀ for comparison with that of the corresponding I_h isomer. This report describes differences in the photochemical reactivities of the I_h and D_5h isomers of Sc₃N@C₈₀ toward 1 as well as silirane 4 [26,27] and digermirane 7 [28,29]. Very interestingly, Sc₃N@D_5h-C₈₀ was found to be photochemically inert toward 1–3, whereas Sc₃N@I_h-C₈₀ undergoes facile addition reactions under identical conditions. This result indicates a novel photochemical method for the exclusive separation of the I_h and D_5h isomers of Sc₃N@C₈₀ as follows: (i) selective derivatization of Sc₃N@I_h-C₈₀ in mixtures of the I_h and D_5h isomers, (ii) facile HPLC separation of Sc₃N@D_5h-C₈₀ from the derivatized Sc₃N@I_h-C₈₀, and (iii) recovery of pristine Sc₃N@I_h-C₈₀ by thermolysis of its derivative.

In addition, the laser flash photolysis of Sc₃N@C₈₀ was conducted to elucidate differences in the reactivities of I_h and D_5h isomers. To date, few examples of comparative studies of the photoreactions of fullerene isomers with different cage symmetries have been reported. We reported earlier that C_2v-C₇₈ undergoes photoreaction with 1 to afford the corresponding silylated C₇₈, whereas D₃-C₇₈ was inert under the same conditions, indicating a procedure for the separation of C_2v-C₇₈ and D₃-C₇₈ [30]. More recently, the photodynamics of three isomers of Sc₂C₂@C₈₂ have been reported as depending on the different fullerene cage symmetries although the intermolecular reactions of the photoexcited Sc₂C₂@C₈₂ with organic molecules have not been examined yet [31]. For this study, we investigate the mechanistic origins for the difference in the reactivities of the Sc₃N@C₈₀ based on cage symmetries using spectroscopic and theoretical studies.

2. Results and Discussion

2.1. Separation of Sc₃N@I_h-C₈₀ and Sc₃N@D_5h-C₈₀ Using Photochemical Functionalization

As described above, earlier reports have shown that Sc₃N@I_h-C₈₀ reacts with 1 under photolytic conditions to give the corresponding 1,2-adduct 2 and 1,4-adduct 3 [32]. To examine the reactivity of Sc₃N@D_5h-C₈₀, a toluene solution of Sc₃N@D_5h-C₈₀ and 1 was irradiated for 20 h using two 500 W halogen lamps (cut off < 400 nm) under an argon atmosphere (Scheme 1). However, Sc₃N@D_5h-C₈₀ was found to be inert toward 1, as shown in the HPLC profiles of the photoreaction (Figure S1 in Supplementary Materials).

This result led us to apply this photoreaction to the separation of the I_h and D_5h isomers of Sc₃N@C₈₀. When a mixture of the I_h and D_5h isomers and 1 in toluene was irradiated for 40 h, the I_h isomer was consumed with the concomitant formation of 2 and 3, whereas the D_5h isomer remained intact (Figure 1). The pristine Sc₃N@D_5h-C₈₀ and the mixture of 2 and 3 were separated easily from the reaction mixture by preparative HPLC without recycling procedures. Finally, thermal desilylation of the mixture of 2 and 3 was performed at 160–170 °C in o-dichlorobenzene (ODCB) for 20 h. Subsequent HPLC separation afforded pristine Sc₃N@I_h-C₈₀ (Figure S2). These procedures established a straightforward method to separate the I_h and D_5h isomers of Sc₃N@C₈₀.

Such a separation method employing silirane 4 as an alternative derivatizing reagent was also examined based on our earlier reported result obtained from the photochemical addition of 4 to Sc₃N@I_h-C₈₀ (Scheme 2) [26,27]. In fact, it was confirmed that 4 did not undergo an addition reaction with Sc₃N@D_5h-C₈₀ under the photolytic condition used for Sc₃N@I_h-C₈₀ (Figure S3). As expected, 4 also worked well as a selective carbosilylating reagent for Sc₃N@I_h-C₈₀ without reaction with Sc₃N@D_5h-C₈₀. Consequently, photoirradiation of a mixture of the I_h and D_5h isomers and 4 in toluene for 60 h followed by HPLC separation gave pristine Sc₃N@D_5h-C₈₀ and a mixture of 5a, 5b, and 6 [27], as shown in Figure 2. However, photochemical reactivity of 4 was somewhat lower than that of 1 considering the reaction time necessary to consume Sc₃N@I_h-C₈₀. In addition, thermal decomposition of the mixture of 5a, 5b, and 6 was performed at 160–170 °C in ODCB for 40 h to give Sc₃N@I_h-C₈₀ along with a recovered mixture of 5a, 5b, and 6 (Figure S4). This result indicates that thermal extrusion reactions of the addends in 5a, 5b, and 6 were less efficient than those of 2 and 3, which might reflect the relative stabilities of these adducts.

An earlier report described that digermirane 7 is more reactive than its silicon analog 8 toward Lu₃N@I_h-C₈₀ under visible irradiation because of the excellent electron-donor property of 7 [29]. This remarkable result led us to evaluate 7 as a third candidate for use as a selective derivatizing reagent for Sc₃N@I_h-C₈₀. First, the photoreaction of 7 with Sc₃N@I_h-C₈₀ was performed in a manner similar to that used for 1 and 4. During photolysis, HPLC analysis indicated that a product peak developed intensively as the peak of Sc₃N@I_h-C₈₀ decreased (Figure S5). After consumption of Sc₃N@I_h-C₈₀, preparative HPLC separation of the reaction mixture afforded the 1,4-adduct 9 as the first example of germylated Sc₃N@I_h-C₈₀ derivative. The structure of 9 was established by X-ray crystallographic analysis as described below.

However, as expected, Sc₃N@D_5h-C₈₀ did not react with 7 under identical conditions for prolonged photoirradiation (Figure S6). This result led us to apply this germylation reaction to the separation of the I_h and D_5h isomers of Sc₃N@C₈₀ (Scheme 3). When a mixture of the I_h and D_5h isomers and 7 in toluene was irradiated for 20 h, the I_h isomer was consumed with the formation of 9, whereas the D_5h isomer remained intact (Figure 3). The adduct 9 and the pristine D_5h isomer were separated easily using preparative HPLC. Finally, degermylation of 9 was accomplished by thermolysis at 130 °C in ODCB for 15 h (Figure S7). This thermolysis is apparently more efficient even at lower temperatures than in the cases of 2, 3, 5a, 5b, and 6, probably because of the lower bond energies of C–Ge bonds (242 kcal/mol) than those of C–C and C–Si bonds (348 kcal/mol and 301 kcal/mol, respectively) [33].

2.2. Characterization of Germylated Sc₃N@I_h-C₈₀ 9

Structural analysis of 9 was conducted based on our earlier studies of the related derivatives 2, 10, 11, and 12 (Figure 4) [29,32]. The visible-near-IR (vis-NIR) spectrum of 9 closely resembles that of 2 (Figure 5). In addition, the NMR spectral features of 9 are similar to those of 11 and 12. The existence of two isomeric molecules is inferred because the ¹H NMR spectrum shows two Ge-CH₂-Ge methylene groups, respectively, as singlets at 2.33 and 2.57 ppm (Figure S8). In the ¹³C NMR spectrum, two methylene carbon signals that are attributable to Ge-CH₂-Ge were observed at 23.04 and 26.81 ppm. The ¹³C NMR spectrum also shows two sets of four methyl groups, two sp³ carbon signals of the I_h-C₈₀ cages, and a total of 102 sp² carbon signals that are attributable to the I_h-C₈₀ cages and the Dep ring carbons (Figure S9). These results indicate the existence of a mixture of conformational isomers of 1,4-adducts with C₂ symmetries, as observed for 2, 10, 11, and 12. To examine the conformational exchange in 9, variable temperature (VT) ¹H NMR experiments were performed between 303 and 363 K (Figure S10). As expected, the signals coalesced as the temperatures increased to show broad signals at 363 K. The spectrum at 303 K was reproduced when the NMR probe temperature was decreased to room temperature. The matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrum showed that no molecular ion peak expected for 9⁻ was observed, although 1,1,4,4-tetraphenyl-1,3-butadiene (TPB), 9-nitroanthracene (9-NA), and 2,5-dihydroxybenzoic acid (DHB) were used as matrices. This result is attributable to the low stability of the molecular ion 9⁻, as reported for the MALDI-TOF spectrum of 12 [29].

Fortunately, the 1,4-addition structure of 9 was determined firmly using the following X-ray crystallographic analysis. Black block crystals of 9 suitable for X-ray diffraction were obtained using the liquid–liquid bilayer diffusion method with CS₂ and hexane at 0 °C. The crystal structure of 9 shows two disordered positions in the I_h-C₈₀ cage with occupancies of 0.72 and 0.28, whereas the digermirane addend is ordered (Figure S11). This result suggests that the crystal structure of 9 includes a pair of diasteromers, which is consistent with the NMR observations. Disorder also exists in the orientations of the Sc₃N clusters. They involve six locations of Sc atoms with a common N atom position. These Sc atom sites fall into two Sc₃N sets with occupancies of 0.68 and 0.32. Figure 6 presents orientation of the cage and the Sc₃N cluster in 9 with major occupancies.

We have already reported that the redox properties of silylated and germylated EMFs are altered considerably compared to the corresponding pristine fullerenes because of electron-donating effects of the silyl and germyl groups [16]. The redox property of 9 was verified using cyclic voltammetry (CV) and differential pulse voltammetry (DPV), as shown in Figure 7. The first oxidation (E^ox₁) and first reduction (E^red₁) potentials of 9 are shifted cathodically by 510 and 250 mV, respectively, relative to those of Sc₃N@I_h-C₈₀ as presented in

Table 1. Redox potentials (V) ^a and calculated HOMO/LUMO levels (eV) of 2, 9, and Sc₃N@I_h-C₈₀.

Compound	Eox1	Ered1	HOMO	LUMO
9	+0.06 b	−1.51	−4.82	−2.85
2	+0.08 b,c	−1.45 c	−4.81	−2.84
Sc3N@Ih-C80	+0.57 d	−1.26 d	−5.50	−3.30

^a Values obtained by DPV are in volts relative to the ferrocene/ferrocenium couple (Fc/Fc⁺). ^b Irreversible. ^c Data from Ref. [32]. ^d Data from Ref. [21].

. In addition, both E^ox₁ and E^red₁ potentials of 9 are slightly more negative than those of the silylated derivative 2 [32]. Furthermore, the density functional theory calculations of 9 were conducted at the B3LYP/6-31G*~SDD level to obtain a basis for its electronic structure [34,35,36,37]. The optimized structure of 9 was calculated using an initial structure resembling that of the X-ray structure, as shown in Figure 6. The highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) levels of 9 are higher than those of pristine Sc₃N@I_h-C₈₀ by 0.68 and 0.45 eV, respectively (Table 1). These changes of the HOMO and LUMO levels are qualitatively consistent with those of the redox potentials of 2 and Sc₃N@I_h-C₈₀.

2.3. Transient Absorption Spectroscopy of Photoreactions of Sc₃N@C₈₀

Laser flash photolysis experiments were conducted to shed light on the differences in the reactivities of the I_h and D_5h isomers. Transient absorption spectra were observed in visible (420–760 nm) and NIR (810–1028 nm, 1055–1400 nm) regions using laser excitation at 387 nm. The gaps in the probed spectral range stem from the detection limits of the visible and NIR detectors, gap between 760 and 810 nm, and from the fundamental wavelength of the white light laser source, gap between 1028 and 1055 nm. In toluene, the absorption band of the triplet excited state of Sc₃N@I_h-C₈₀ (³Sc₃N@I_h-C₈₀^*) was observed at λ_max 520 nm, as shown in Figure 8 and Figure S12. A report of an earlier study described how the singlet excited state of Sc₃N@I_h-C₈₀ (¹Sc₃N@I_h-C₈₀^*) undergoes facile intersystem crossing (ISC) to give ³Sc₃N@I_h-C₈₀^* with an absorption band around 500 nm [38]. Upon addition of 200 times equimolar amounts of 7, the decay of ³Sc₃N@I_h-C₈₀^* was accelerated considerably, as shown in comparisons of decay plots both at 500 nm (Figure 9a,b) and at 1104 nm (Figure 9c,d).

However, when Sc₃N@D_5h-C₈₀ was photolyzed under the same conditions, the observed transient absorption peak was much weaker than in the case of Sc₃N@I_h-C₈₀, as shown in Figure 10 and Figure S14. Singlet and triplet excited states of Sc₃N@D_5h-C₈₀ (¹Sc₃N@D_5h-C₈₀^* and ³Sc₃N@D_5h-C₈₀^*, respectively) have not been hitherto characterized by spectroscopic studies. Therefore, several mechanistic possibilities should be examined for the weak transient absorption in Figure 10. For example, the photoexcitation of Sc₃N@D_5h-C₈₀ might not be as efficient as that of Sc₃N@I_h-C₈₀, although the former has a lower but comparable molar extinction coefficient at λ = 387 nm (excitation wavelength in laser flash photolysis) compared to that of the latter, as shown in Figure S16. Alternatively, if it is assumed that the molar extinction coefficient and the lifetime of ³Sc₃N@D_5h-C₈₀^* is not significantly different from those of ³Sc₃N@I_h-C₈₀^*, then the weak absorption observed in Figure 10 suggests the low concentration of ³Sc₃N@D_5h-C₈₀^* under photolytic conditions. One possible explanation for this point is that the ISC process from ¹Sc₃N@D_5h-C₈₀^* to ³Sc₃N@D_5h-C₈₀^* might be less effective than in the case of Sc₃N@I_h-C₈₀. In the presence of 200 times equimolar amounts of 7, no appreciable difference in intensity of the transient absorption at 500 nm was observed because intense absorption of the triplet excited state of toluene hindered the evaluation (Figure 11a,b) [39]. Although slight differences of intensity at 1215 nm were noted upon addition of 7 (Figure 11c,d), they are mostly attributable to the poor signal-to-noise ratio caused by the low intensity in the NIR region. As such, a quantitative analysis of the photoactivity of Sc₃N@D_5h-C₈₀ and, in turn, a comparison with Sc₃N@I_h-C₈₀ is rendered impossible.

2.4. Theoretical Calculations of Photoreactions of Sc₃N@C₈₀

To gain insight into the photochemical processes of Sc₃N@C₈₀, we applied time-dependent density functional theory (TD-DFT) [40] calculations on the ten lowest excited states including the singlet states (S_n; n = 1, 2, …) and triplet states (T_n; n = 1, 2, …), respectively, for the I_h and D_5h isomers of Sc₃N@C₈₀. The corresponding electronic excitation energies are schematized in Figure 12. Major orbital transition configurations are presented in Tables S1 and S2.

According to Kasha’s rule [41], when the energy gaps between two singlet states (S_n and S_m) are small, internal conversion (IC) processes from energetically higher singlet states to lower singlet states are enhanced more effectively than the ISC processes. In the case of Sc₃N@I_h-C₈₀, there are the triplet excited states T₂ and T₃, with energies closely approximating those of the lowest singlet excited state S₁. The energy differences between S₁-T₂ and S₁-T₃ are, respectively, 0.0387 and 0.0381 eV (Figure 12a). In the case of Sc₃N@D_5h-C₈₀, T₁ is energetically lower than S₁, which is lower than other triplet states T_n (n > 1). S₁ and T₁ are different from each other in energy by 0.1239 eV (Figure 12b). However, these energy differences do not explain the efficiencies of the ISC processes in Sc₃N@C₈₀. Alternatively, the efficiency of the ISC processes might depend on the spin-orbit coupling (SOC) interaction between the singlet and triplet excited states. The behaviors of encapsulated metal clusters inside the carbon cages might therefore affect enhancement of the SOC interaction. Further understanding of the photoreactivities of Sc₃N@C₈₀ must await investigations of those photoexcited states of the corresponding I_h and D_5h isomers.

It was proposed in an earlier report of the relevant literature that the photoreaction of C₆₀ and 1 proceeds via the electron donor–acceptor interaction between ³C₆₀* and 1 based on a quenching experiment of ³C₆₀* by 1 [42]. The Rehm–Weller equation [43,44] for estimating the free energy change ΔG of electron transfer (ET) between 1 and ³C₆₀* in non-polar solvents such as toluene exhibited a positive value. Therefore, results suggest that the ET between 1 and ³C₆₀* is not efficient in non-polar solvents, but that process is regarded as possible because the ΔG value is not so large [42]. Additionally, it has been reported that photoirradiation of C₆₀ and siliranes that possess benzylsilane structures, which are good electron donors, afforded the corresponding adducts [45]. In contrast, when siliranes without benzylsilane structures were used as substrates, the photoaddition reaction proceeded very slowly.

Based on these results, the ΔG values for the ET processes from 1 to the excited triplet states of Sc₃N@I_h-C₈₀ and Sc₃N@D_5h-C₈₀ were calculated using the oxidation potential (E_ox) of 1 (+0.27 V vs. Fc/Fc⁺) [42] and the first reduction potentials (E_red¹) of Sc₃N@I_h-C₈₀ (−1.26 V vs. Fc/Fc⁺) [21] and Sc₃N@D_5h-C₈₀ (−1.33 V vs. Fc/Fc⁺) [21], respectively. Assuming that the IC process from higher excited triplet states T_n (n >1) to the lowest state T₁ occurs rapidly, the energies of the T₁ states were evaluated using TD-DFT calculations as 1.82 eV for Sc₃N@I_h-C₈₀ and 1.49 eV for Sc₃N@D_5h-C₈₀. As a result, the ΔG values were estimated as +10.36 kcal/mol for Sc₃N@I_h-C₈₀ and +19.50 kcal/mol for Sc₃N@D_5h-C₈₀ [43,46,47]. These values are positive, as in the case of C₆₀, but the value of the I_h isomer is small, whereas that of the D_5h isomer is nearly twice as large as that of the I_h isomer. These results suggest that the photoinduced electron transfer process of ³Sc₃N@D_5h-C₈₀^* should be less efficient than that of ³Sc₃N@I_h-C₈₀^* even if they take place. Based on these estimations, the poor electron acceptor property of ³Sc₃N@D_5h-C₈₀^* might decrease its photochemical reactivity toward 1, 4, and 7.

3. Experimental Section

Separation of Sc₃N@C₈₀ using 1: A mixture of the I_h and D_5h isomers of Sc₃N@C₈₀ (2.7 mg) and 1 (58 mg) in toluene (20 mL) was degassed using freeze–pump–thaw cycles under reduced pressure in a Pyrex tube (ϕ20 mm). The solution was irradiated for 40 h with two 500 W halogen lamps using an aqueous sodium nitrite filter solution (cutoff < 400 nm) under an argon atmosphere. Preparative HPLC separation with a Buckyprep-M column of the reaction mixture afforded pristine Sc₃N@D_5h-C₈₀ (0.7 mg) and a mixture of 2 and 3 (2.2 mg).

Thermal desilylation of 2 and 3: A solution of 2 and 3 (2.1 mg) in ODCB (5 mL) was heated at 160–170 °C under an argon atmosphere in a Schlenk tube in the dark for 20 h. After removal of ODCB in vacuo, Sc₃N@I_h-C₈₀ (0.8 mg) was obtained by preparative HPLC separation with a Buckyprep-M column.

Selective carbosilylation of Sc₃N@C₈₀ using 4: A mixture of I_h and D_5h isomers of Sc₃N@C₈₀ (2.6 mg) and 4 (61 mg) in toluene (20 mL) was irradiated as described above for 60 h. Preparative HPLC separation with a Buckyprep-M column of the reaction mixture afforded pristine Sc₃N@D_5h-C₈₀ (0.5 mg) and a mixture of 5a, 5b, and 6 (2.5 mg).

Thermal decarbosilylation of 5a, 5b, and 6: A solution of 5a, 5b, and 6 (2.1 mg) in ODCB (5 mL) was heated at 160–170 °C under an argon atmosphere in a Schlenk tube in the dark for 40 h. After removal of ODCB in vacuo, Sc₃N@I_h-C₈₀ (0.9 mg) was obtained along with a recovered mixture of 5a, 5b, and 6 (0.7 mg) by preparative HPLC separation with a Buckyprep-M column.

Photoreaction of Sc₃N@I_h-C₈₀ with 7: A solution of Sc₃N@I_h-C₈₀ (2.0 mg) and 7 (12.6 mg) in toluene (15 mL) was irradiated for 5 h as described above. Preparative HPLC separation with a Buckyprep-M column of the reaction mixture afforded 9 (2.1 mg). Spectral data for 9: The following NMR data are described based on the existence of two conformational isomers with C₂ symmetries. ¹H NMR (500 MHz, CS₂/CDCl₃ (1:1), 298 K) δ 7.33–7.28 (m, 6H), 7.22 (t, J = 7.5 Hz, 2H), 7.18 (d, J = 7.5 Hz, 2H), 7.13 (d, J = 7.5 Hz, 2H), 7.10 (t, J = 7.5 Hz, 2H), 7.02–6.96 (m, 8H), 6.91 (d, J = 7.5 Hz, 2H), 3.75 (dq, J = 7.5 Hz, 15 Hz, 2H), 3.68–3.43 (m, 8H), 3.31 (dq, J = 7.5 Hz, 15 Hz, 2H), 3.09 (dq, J = 7.5 Hz, 15 Hz, 2H), 2.98–2.74 (m, 8H), 2.63 (dq, J = 7.5 Hz, 15 Hz, 2H), 2.57 (s, 2H), 2.51–2.36 (m, 8H), 2.33 (s, 2H), 1.70 (t, J = 7.5 Hz, 6H), 1.43 (t, J = 7.5 Hz, 6H), 0.82 (t, J = 7.5 Hz, 6H), 0.76 (t, J = 7.5 Hz, 6H), 0.66–0.63 (m, 18H), 0.58 (t, J = 7.5 Hz, 6H); ¹³C NMR (125 MHz, CS₂/CDCl₃ (1:1), 298 K) δ 178.97, 176.15, 153.01, 152.92, 152.67, 152.58, 151.85, 151.68, 150.91, 150.87, 149.81, 149.67, 149.06, 148.45(2set), 147.72, 147.62, 147.42, 147.27, 147.22, 147.05, 147.00, 146.85, 146.67, 146.56, 146.52, 146.45, 146.38, 145.91, 146.88, 145.55, 145.52, 145.44, 145.36, 145.23, 144.97, 144.78, 144.58, 144.32, 143.16, 142.75, 142.37, 142.24, 141.95, 141.75, 141.15, 141.06, 140.88, 140.74, 140.64, 140.54, 140.45, 140.37, 139.21, 138.67, 138.53, 138.47, 137.51, 137.48, 136.04, 136.02, 135.91, 135.60, 135.43, 135.41, 135.29, 135.25, 134.95, 134.88, 134.82, 134.58, 134.44, 134.31, 134.22, 133.54, 133.17, 133.11, 132.51, 132.08, 132.00, 129.94, 129.59, 129.42, 129.32, 127.76, 127.67, 127.64, 126.88, 126.46, 124.03, 123.94, 116.40, 115.54, 59.09, 33.35, 32.87, 32.43, 32.11, 30.25, 29.92, 29.26, 29.04, 26.81, 23.04, 15.14, 15.06, 14.98, 14.78, 13.73; vis-NIR (CS₂) λ_max 926 nm; MALDI-TOF MS (TPB) m/z 1109 (Sc₃N@C₈₀⁻).

Selective germylation of Sc₃N@C₈₀ using 7: A mixture of I_h and D_5h isomers of Sc₃N@C₈₀ (2.4 mg) and 7 (65 mg) in toluene (20 mL) was irradiated as described above for 20 h. Preparative HPLC separation with a Buckyprep-M column of the reaction mixture afforded pristine Sc₃N@D_5h-C₈₀ (0.7 mg) and 9 (2.4 mg).

Thermal degermylation of 9: A solution of 9 (2.4 mg) in ODCB (5 mL) was heated at 130 °C under an argon atmosphere in a Schlenk tube in the dark for 15 h. After the removal of ODCB in vacuo, Sc₃N@I_h-C₈₀ (1.1 mg) was obtained by preparative HPLC separation using a Buckyprep-M column.

X-ray crystallography of 9: Black block crystals suitable for X-ray diffraction were obtained using the liquid–liquid bilayer diffusion method with solutions of 9 in CS₂ using hexane as a poor solvent at 0 °C. Single-crystal X-ray diffraction data of 9 were collected on a Saturn70 CCD diffractometer (Rigaku Corp.) equipped with a nitrogen-gas flow low-temperature apparatus providing a constant temperature at 120 K. Crystal data for Sc₃N@I_h-C₈₀(Dep₂Ge)₂CH₂(9)·1.5(CS₂): C_122.5H₅₄Ge₂Sc₃NS₃: M_r = 1915.90, black block, 0.25 × 0.13 × 0.07 mm, λ = 0.71069 Å, monoclinic, space group P2₁/n (no. 14), a = 19.5525(17), b = 20.9254(16), c = 19.6564(16) Å, β =111.0114(5)°, T = 120 K, V= 7506(11) ų, Z = 4, 168,721 reflections measured, 16,520 unique (R_int = 0.0572), which were used for all calculations, 2θ_max = 54.20; min/max transmission=0.782/0.941 (numerical absorption correction applied); the structure was solved using a direct method using SIR2014 [48] and was refined with SHELXL [49]. The final wR(F₂) was 0.0927 (all data), conventional R₁ = 0.0424 computed for 16,316 reflections with I >2σ (I) using 1937 parameters with 876 restraints. Crystallographic computations were performed with Yadokari-XG 2009 [50]. CCDC2127212 (9) contains the supplementary crystallographic data for this paper, and is obtainable free of charge from the Cambridge Crystallographic Data Centre.

Computational Methods: The computations were performed with the density functional theory (DFT) approach, namely using Becke’s three parameter functional [34] combined with the non-local Lee–Yang–Parr correlation functional [35] (B3LYP). The basis set applied to H, C, N, and Si atoms is the standard 6-31G* basis [36] whereas Sc and Ge atoms are treated in the SDD basis set [37] with the SDD effective core potential (the combined basis set is coded B3LYP/6-31G*~SDD). The geometry optimizations were performed with the analytically constructed energy gradients. In the optimized B3LYP/6-31G*~SDD geometries, the electronic excitation energies were evaluated with the time-dependent (TD) DFT response-theory method [40], again at the B3LYP/6-31G*~SDD level. The computations were performed using the Gaussian 09 program package [51] (See Supplementary Materials).

Transient Absorption Spectroscopy: The excitation was performed with an amplified CPA-2110 titanium:sapphire laser (1 kHz; 150 fs pulse width; 400 nJ laser energy) from Clark-MXR Inc. EOS SYSTEM (0–10 μs) from Ultrafast Systems, working with 1 kHz pump laser at 387 nm wavelength. Probing was performed with a 2 kHz continuous white light fiber laser. Data evaluation has been conducted by means of multiwavelength and global analysis using the GloTarAn package [52].

4. Conclusions

Photoreactions of Sc₃N@I_h-C₈₀ and 1, 4, and 7 afforded the corresponding 1:1 adducts, whereas Sc₃N@D_5h-C₈₀ was found to be inert under identical photolytic conditions. The derivatives of Sc₃N@I_h-C₈₀ and pristine Sc₃N@D_5h-C₈₀ were separated easily by HPLC without recycling processes. In addition, pristine Sc₃N@I_h-C₈₀ was recovered by thermolytic decomposition of the corresponding photoadducts. These procedures provide a novel method for the exclusive separation of Sc₃N@I_h-C₈₀ and Sc₃N@D_5h-C₈₀. In laser flash photolysis experiments, the decay of transient absorption for ³Sc₃N@I_h-C₈₀^* was accelerated in the presence of 7. In contrast, for Sc₃N@D_5h-C₈₀, the transient absorption was too weak to offer a basis for interaction of the corresponding triplet excited states with 7. In turn, a meaningful conclusion regarding the reactivity differences between Sc₃N@I_h-C₈₀ and Sc₃N@D_5h-C₈₀ was hampered. It is, however, expected that the electron-transfer processes between the ³Sc₃N@D_5h-C₈₀^* and 1 are much less likely to occur than those of Sc₃N@I_h-C₈₀, judging from the corresponding values of changes of free energies ΔG. Therefore, the photochemical inertness of Sc₃N@D_5h-C₈₀ toward 1, 4, and 7 might be partly attributed to the lower electron-acceptor ability of ³Sc₃N@D_5h-C₈₀^*. Further investigations of the dependence of the photochemical reactivities of EMFs on their structures including the carbon cage symmetries and the encapsulated metals are in progress.