Functionalization of Endohedral Metallofullerenes with Reactive Silicon and Germanium Compounds †

Exohedral derivatization of endohedral metallofullerenes (EMFs) has been exploited as a useful method for characterizing the structural and chemical properties of EMFs, and for functionalizing them for potential applications. The introduction of heteroatoms, such as electropositive silicon atoms, to fullerene cages is a novel functionalization method that remarkably affects the electronic characteristics of fullerenes. This review comprehensively describes the results of the reactions of monometallofullerene, dimetallofullerene, and trimetallic nitride template EMFs with disilirane, silirane, silylene, and digermirane, which afforded the corresponding silylated and germylated fullerenes. Several examples emphasize that exohedral functionalization regulates the dynamic behaviors of the encapsulated metal atoms and clusters in the fullerene cages. The electronic effects of silyl and germyl groups are represented by comparing the redox properties of silylated and germylated EMFs with those of other EMFs derivatized with carbon-atom-based functional groups.


Introduction
Endohedral metallofullerenes (EMFs) are attracting great interest because of their unique structures and properties [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15]. Encapsulation of metal atoms and metal clusters inside EMFs results in electron transfer from the inner species to the carbon cages. Results of earlier studies show that EMFs have remarkable electronic properties such as oxidation and reduction potentials. The difference in redox potentials between EMFs and hollow fullerenes is a key factor characterizing the properties and reactivities of EMFs.
Exohedral derivatization of EMFs and empty fullerenes has been explored extensively as an effective method to modify and enhance the characteristics of fullerenes for various nanomaterials science and biomedicine applications [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15]. To date, many reactions have been applied for the exohedral derivatization of EMFs with respect to their use for functional materials. However, the derivatization reactions of EMFs remain sparse compared to those of empty fullerenes. In many cases, the derivatizing organic groups are bonded covalently to the carbon cages of EMFs through C-C bonds. Therefore, derivatization that exploits the heteroatom properties is expected to contribute to the development of EMFs as functional materials. In this context, derivatization of fullerenes using silicon and germanium atoms is an attractive method for modifying the electronic properties of fullerenes because of the electron-donating effects of silyl and germyl groups [16][17][18][19]. We found that the reactions of empty fullerenes with reactive silicon and germanium compounds afforded the corresponding fullerene derivatives [20][21][22][23][24][25][26][27][28][29][30][31][32]. These results prompted us to investigate properties of EMFs by examining their reactivities toward silicon and germanium compounds. The combination of fullerenes and silicon and germanium compounds provide a basis for the production of novel functional materials for electronic device, as mentioned in the section on potential applications. This review presents a summary of our results of silylation and germylation reactions as part of our continuing investigation of the chemical functionalization of EMFs [1,[4][5][6][7][8]11,12,15]. We demonstrate that the introduction of silicon and germanium atoms markedly perturbs the electronic properties of fullerenes because of the electron-donating effects of these atoms. More specifically, derivatization of EMFs using reactive compounds containing heavier group 14 elements, such as silylene (silicon analog of carbene), disilirane (disilacyclopropane), silirane (silacyclopropane), and digermirane (digermacyclopropane) will be outlined. For several silylated and germylated derivatives of EMFs, the electronic properties are discussed in comparison with reference compounds that possess the same addition patterns because the redox behaviors of derivatized EMFs depend on the regiochemistry of the addends on the carbon cages. We also describe how derivatization with organosilicon compounds influences the behaviors of encapsulated metal atoms. In this review, we will focus on the functionalization of the following EMFs: La@C 2v -C 82 , La 2 @I h -C 80 , Ce 2 @I h -C 80 , Ce 2 @D 3h -C 78 , Sc 3 N@I h -C 80 , and Lu 3 N@I h -C 80 (Figure 1).
Molecules 2017, 22,1179 2 of 17 C-C bonds. Therefore, derivatization that exploits the heteroatom properties is expected to contribute to the development of EMFs as functional materials. In this context, derivatization of fullerenes using silicon and germanium atoms is an attractive method for modifying the electronic properties of fullerenes because of the electron-donating effects of silyl and germyl groups [16][17][18][19]. We found that the reactions of empty fullerenes with reactive silicon and germanium compounds afforded the corresponding fullerene derivatives [20][21][22][23][24][25][26][27][28][29][30][31][32]. These results prompted us to investigate properties of EMFs by examining their reactivities toward silicon and germanium compounds. The combination of fullerenes and silicon and germanium compounds provide a basis for the production of novel functional materials for electronic device, as mentioned in the section on potential applications. This review presents a summary of our results of silylation and germylation reactions as part of our continuing investigation of the chemical functionalization of EMFs [1,[4][5][6][7][8]11,12,15]. We demonstrate that the introduction of silicon and germanium atoms markedly perturbs the electronic properties of fullerenes because of the electron-donating effects of these atoms. More specifically, derivatization of EMFs using reactive compounds containing heavier group 14 elements, such as silylene (silicon analog of carbene), disilirane (disilacyclopropane), silirane (silacyclopropane), and digermirane (digermacyclopropane) will be outlined. For several silylated and germylated derivatives of EMFs, the electronic properties are discussed in comparison with reference compounds that possess the same addition patterns because the redox behaviors of derivatized EMFs depend on the regiochemistry of the addends on the carbon cages. We also describe how derivatization with organosilicon compounds influences the behaviors of encapsulated metal atoms. In this review, we will focus on the functionalization of the following EMFs: La@C2v-C82, La2@Ih-C80, Ce2@Ih-C80, Ce2@D3h-C78, Sc3N@Ih-C80, and Lu3N@Ih-C80 ( Figure 1).

Comparative Studies of Reactions of EMFs with Disilirane
In 1995, the first chemical derivatization of EMFs was achieved by the photochemical reaction of La@C2v-C82 and disilirane 1 [33]. It is particularly interesting that La@C2v-C82 reacted both thermally and photochemically with 1 to afford the corresponding adduct, although the empty fullerenes C60 and C70 did not react with 1 under thermal reactions in the absence of photoirradiation [21,23]. This result can be reasonably explained by the higher electron-accepting ability of La@C2v-C82 than that of empty fullerenes. Since this discovery, the reactivities of various EMFs have been investigated using 1 as a chemical probe (Scheme 1). Digermirane 3 reacted with La@C2v-C82 showing higher thermal reactivity than 1, which results from the enhanced electron-donating property of 3 [34].

Comparative Studies of Reactions of EMFs with Disilirane
In 1995, the first chemical derivatization of EMFs was achieved by the photochemical reaction of La@C 2v -C 82 and disilirane 1 [33]. It is particularly interesting that La@C 2v -C 82 reacted both thermally and photochemically with 1 to afford the corresponding adduct, although the empty fullerenes C 60 and C 70 did not react with 1 under thermal reactions in the absence of photoirradiation [21,23]. This result can be reasonably explained by the higher electron-accepting ability of La@C 2v -C 82 than that of empty fullerenes. Since this discovery, the reactivities of various EMFs have been investigated using 1 as a chemical probe (Scheme 1). Digermirane 3 reacted with La@C 2v -C 82 showing higher thermal reactivity than 1, which results from the enhanced electron-donating property of 3 [34]. Redox potentials of selected EMFs and empty fullerenes are shown in Figure 2. Values of the first reduction (denoted as E red 1) potentials are regarded as an important index for the reactivities toward 1 because fullerenes are believed to function as electron acceptors toward 1, which is a good electron donor [15]. All the fullerenes shown in Figure 2 reacted with 1 by photoirradiation to afford the corresponding adducts. However, the EMFs and empty fullerenes with lower E red 1 potentials did not undergo addition reactions under thermal conditions at 80 °C [15]. For example, La2@Ih-C80 reacted with 1 both photochemically and thermally [35], although Sc3N@Ih-C80 was found to be reactive only under photolytic conditions [36,37], although the first oxidation (denoted as E°x1) potentials of these EMFs are similar. Therefore, the higher thermal reactivity of La2@Ih-C80 toward 1 might be attributed to its good electron acceptor properties. It is noteworthy that the reactivities of EMFs were tunable by chemical oxidation or reduction [38]. Oxidation of La@C2v-C82 by (p-BrC6H4)3N + SbCl6 − produced the corresponding cation, [La@C2v-C82] + SbCl6 − , which reacted with 1 even at room temperature. In contrast, the addition of NaSCH3 to La@C2v-C82 afforded the corresponding anionic species, Na + [La@C2v-C82] − , which was found to be inert to 1. These results demonstrate that the electrophilicity of EMFs plays an important role in controlling their reactivities toward 1. La@C 2v -C 82 [33,39] Ce@C 2v -C 82 [89] Pr@C 2v -C 82 [90] Gd@C 2v -C 82 [91] Y@C 2v -C 82 [ Redox potentials of selected EMFs and empty fullerenes are shown in Figure 2. Values of the first reduction (denoted as E red 1 ) potentials are regarded as an important index for the reactivities toward 1 because fullerenes are believed to function as electron acceptors toward 1, which is a good electron donor [15]. All the fullerenes shown in Figure 2 reacted with 1 by photoirradiation to afford the corresponding adducts. However, the EMFs and empty fullerenes with lower E red 1 potentials did not undergo addition reactions under thermal conditions at 80 • C [15]. For example, La 2 @I h -C 80 reacted with 1 both photochemically and thermally [35], although Sc 3 N@I h -C 80 was found to be reactive only under photolytic conditions [36,37], although the first oxidation (denoted as E •x 1 ) potentials of these EMFs are similar. Therefore, the higher thermal reactivity of La 2 @I h -C 80 toward 1 might be attributed to its good electron acceptor properties. Redox potentials of selected EMFs and empty fullerenes are shown in Figure 2. Values of the first reduction (denoted as E red 1) potentials are regarded as an important index for the reactivities toward 1 because fullerenes are believed to function as electron acceptors toward 1, which is a good electron donor [15]. All the fullerenes shown in Figure 2 reacted with 1 by photoirradiation to afford the corresponding adducts. However, the EMFs and empty fullerenes with lower E red 1 potentials did not undergo addition reactions under thermal conditions at 80 °C [15]. For example, La2@Ih-C80 reacted with 1 both photochemically and thermally [35], although Sc3N@Ih-C80 was found to be reactive only under photolytic conditions [36,37], although the first oxidation (denoted as E°x1) potentials of these EMFs are similar. Therefore, the higher thermal reactivity of La2@Ih-C80 toward 1 might be attributed to its good electron acceptor properties. It is noteworthy that the reactivities of EMFs were tunable by chemical oxidation or reduction [38]. Oxidation of La@C2v-C82 by (p-BrC6H4)3N + SbCl6 − produced the corresponding cation, [La@C2v-C82] + SbCl6 − , which reacted with 1 even at room temperature. In contrast, the addition of NaSCH3 to La@C2v-C82 afforded the corresponding anionic species, Na + [La@C2v-C82] − , which was found to be inert to 1. These results demonstrate that the electrophilicity of EMFs plays an important role in controlling their reactivities toward 1. La@C 2v -C 82 [33,39] Ce@C 2v -C 82 [89] Pr@C 2v -C 82 [90] Gd@C 2v -C 82 [91] Y@C 2v -C 82 [  It is noteworthy that the reactivities of EMFs were tunable by chemical oxidation or reduction [38]. Oxidation of La@C 2v -C 82 by (p-BrC 6 H 4 ) 3 N + SbCl 6 − produced the corresponding cation, [La@C 2v -C 82 ] + SbCl 6 − , which reacted with 1 even at room temperature. In contrast, the addition of NaSCH 3 to La@C 2v -C 82 afforded the corresponding anionic species, Na + [La@C 2v -C 82 ] − , which was found to be inert to 1. These results demonstrate that the electrophilicity of EMFs plays an important role in controlling their reactivities toward 1. Figure 2 shows that La@C 2v -C 82 is highly reactive toward 1 because of its good electron-accepting ability. The molecular symmetry of La@C 2v -C 82 , which possesses 24 non-equivalent carbon atoms in the C 2v -C 82 cage, is low. Therefore, addition reactions of La@C 2v -C 82 might produce multiple structural isomers. In fact, thermal reactions of La@C 2v -C 82 with 1 afforded at least six isomers of the mono-adducts based on analysis of the EPR spectrum of the reaction mixture [33]. A similar result was obtained for the thermal addition of digermirane 3 [34], indicating that the addition reactions of 1 and 3 to La@C 2v -C 82 were not regioselective.

Reactions of La@C 2v -C 82 with Disilirane 1
The major isomer of the adducts derived from the reaction of La@C 2v -C 82 with 1 was isolated using HPLC separation [39]. The EPR spectrum of this isomer showed only one set of an octet hyperfine that was smaller than that of pristine La@C 2v -C 82 . The addition of 1 also caused a change in the absorption spectrum, suggesting that the π-conjugation of the carbon cage was altered. Although the structures of the disilirane adducts of La@C 2v -C 82 have not been determined, their redox properties were obtained from electrochemical measurements. The results show that the E •x 1 and E red 1 potentials of the disilirane adducts of La@C 2v -C 82 were shifted cathodically, respectively, by 140 mV and 80 mV. These cathodic shifts of the redox potentials of silylated derivatives of EMFs are reasonably explained by the electron-donating effects of the silyl group, including σ-π conjugation between the C-Si σ-orbitals and the π-orbitals of the fullerene cages [16][17][18][19]. The reaction of Y@C 2v -C 82 with 1 afforded two isomeric products as the 1:1 adducts, for which the cathodic shifts of redox potentials were observed compared to those of pristine Y@C 2v -C 82 [39].

Reactions of La 2 @I h -C 80 and Related EMFs with Disiliranes 1 and 2
Due to the symmetrical structure of the I h -C 80 cage, the number of isomers is limited in the addition reactions of La 2 @I h -C 80 . Only two non-equivalent carbon atoms exist in the I h -C 80 cage: one is shared by one five-membered ring and two six-membered rings (denoted as the [5,6,6]-ring junction); the other is shared by three six-membered rings (denoted as the [6,6,6]-ring junction). The I h -C 80 cage also has two non-equivalent C-C bonds: one is shared by one five-membered ring and one six-membered ring (denoted as the [5,6]-ring junction); the other is shared by two six-membered rings (denoted as the [6,6]-ring junction).
Heating a toluene solution of La 2 @I h -C 80 and disilirane 2 at 80 • C afforded the corresponding 1:1 adduct 5 as the sole product [40] (Scheme 2). The NMR and X-ray crystallographic analyses established the structures of 5 as the 1,4-adduct, indicating that the silyl groups are bonded to the [5,6,6]-junctions of the I h -C 80 cage. In the crystal structure, the two encapsulated metal atoms were located close to the two six-membered rings on the equator of the cage. The reaction of La 2 @I h -C 80 and 1 produced a similar adduct 4. The corresponding adduct of Ce 2 @I h -C 80 6 was also obtained by a similar thermal reaction with 1, showing the same 1,4-addition pattern [41] (Scheme 2).  Figure 2 shows that La@C2v-C82 is highly reactive toward 1 because of its good electron-accepting ability. The molecular symmetry of La@C2v-C82, which possesses 24 non-equivalent carbon atoms in the C2v-C82 cage, is low. Therefore, addition reactions of La@C2v-C82 might produce multiple structural isomers. In fact, thermal reactions of La@C2v-C82 with 1 afforded at least six isomers of the mono-adducts based on analysis of the EPR spectrum of the reaction mixture [33]. A similar result was obtained for the thermal addition of digermirane 3 [34], indicating that the addition reactions of 1 and 3 to La@C2v-C82 were not regioselective.

Reactions of La@C2v-C82 with Disilirane 1
The major isomer of the adducts derived from the reaction of La@C2v-C82 with 1 was isolated using HPLC separation [39]. The EPR spectrum of this isomer showed only one set of an octet hyperfine that was smaller than that of pristine La@C2v-C82. The addition of 1 also caused a change in the absorption spectrum, suggesting that the π-conjugation of the carbon cage was altered. Although the structures of the disilirane adducts of La@C2v-C82 have not been determined, their redox properties were obtained from electrochemical measurements. The results show that the E°x1 and E red 1 potentials of the disilirane adducts of La@C2v-C82 were shifted cathodically, respectively, by 140 mV and 80 mV. These cathodic shifts of the redox potentials of silylated derivatives of EMFs are reasonably explained by the electron-donating effects of the silyl group, including σ-π conjugation between the C-Si σ-orbitals and the π-orbitals of the fullerene cages [16][17][18][19]. The reaction of Y@C2v-C82 with 1 afforded two isomeric products as the 1:1 adducts, for which the cathodic shifts of redox potentials were observed compared to those of pristine Y@C2v-C82 [39].

Reactions of La2@Ih-C80 and Related EMFs with Disiliranes 1 and 2
Due to the symmetrical structure of the Ih-C80 cage, the number of isomers is limited in the addition reactions of La2@Ih-C80. Only two non-equivalent carbon atoms exist in the Ih-C80 cage: one is shared by one five-membered ring and two six-membered rings (denoted as the [5,6,6]-ring junction); the other is shared by three six-membered rings (denoted as the [6,6,6]-ring junction). The Ih-C80 cage also has two non-equivalent C-C bonds: one is shared by one five-membered ring and one six-membered ring (denoted as the [5,6]-ring junction); the other is shared by two six-membered rings (denoted as the [6,6]-ring junction).
Heating a toluene solution of La2@Ih-C80 and disilirane 2 at 80 °C afforded the corresponding 1:1 adduct 5 as the sole product [40] (Scheme 2). The NMR and X-ray crystallographic analyses established the structures of 5 as the 1,4-adduct, indicating that the silyl groups are bonded to the [5,6,6]-junctions of the Ih-C80 cage. In the crystal structure, the two encapsulated metal atoms were located close to the two six-membered rings on the equator of the cage. The reaction of La2@Ih-C80 and 1 produced a similar adduct 4. The corresponding adduct of Ce2@Ih-C80 6 was also obtained by a similar thermal reaction with 1, showing the same 1,4-addition pattern [41] (Scheme 2). In this review, the parent EMFs are shown respectively in brackets. Scheme 3 shows that Ce2@D3h-C78 also reacts readily with disilirane 1 to produce the corresponding 1,4-adduct 7 [42]. Based on X-ray crystallographic studies, the Ce atoms were found to be located at two positions facing the six-membered rings on the C3 axis of the D3h-C78 cage. In the In this review, the parent EMFs are shown respectively in brackets. Scheme 3 shows that Ce 2 @D 3h -C 78 also reacts readily with disilirane 1 to produce the corresponding 1,4-adduct 7 [42]. Based on X-ray crystallographic studies, the Ce atoms were found to be located at two positions facing the six-membered rings on the C 3 axis of the D 3h -C 78 cage. In the paramagnetic 13 C-NMR spectral analysis, the observed carbon signals were more temperature-dependent than those of pristine Ce 2 @D 3h -C 78 . Therefore, the two Ce atoms in 7 likely interact with the C 78 cage more firmly than those in pristine Ce 2 @D 3h -C 78 [42].

Reactions of La2@Ih-C80 with Silirane 8
Scheme 4 shows that La2@Ih-C80 also reacts with silirane 8 under thermal conditions to afford the corresponding 1,2-carbosilylated derivatives 9a and 9b [43]. Structural analyses, including X-ray crystallographic measurements, determined 9a and 9b to be a pair of diastereomers of the [5,6]-adducts. Furthermore, electrochemical measurements revealed that carbosilylation slightly altered the electronic properties of La2@Ih-C80. Results of X-ray crystallographic analysis showed that the crystal structure of 9a has eight sites for the La atoms because of disorder [43]. Theoretical calculations of adducts of La2@Ih-C80 and 8 indicated that changing the La atom positions resulted in the differences in relative energies within the range of 0.85-5.89 kcal mol −1 . These small energy differences suggest the movability of the La atoms inside the cage [43]. The La atoms sites are arranged along the ten contiguously fused six-membered rings that constitute part of the Ih-C80 cage. Figure 3 presents the redox potentials of pristine and derivatized dimetallofullerenes. Remarkable cathodic shifts of both the E°x1 and E red 1 potentials were observed for the silylated derivatives 4, 5, 9a and 9b relative to that of pristine La2@Ih-C80. Results also confirmed that the redox potentials shifted cathodically as the number of silicon atoms introduced into La2@Ih-C80 increased as shown by those of 4, 5, 9a, 9b and 10 [40,43,44]. The electronic effects of the addends are represented more exactly by comparing the [5,6]-carbosilylated adducts 9a, 9b, the [5,6]-pyrrolidino adduct 10 [44], and the [5,6]-tetracyanoethylene oxide (TCNEO) adduct 11 [45], which have the same addition patterns. As expected from the effects of the electron-donating silyl group and electron-withdrawing TCNEO moieties, the redox potentials of 9a/9b and 11 shifted, respectively, to more negative and more positive potentials than those of reference compound 10. Similar cathodic shifts of the redox potentials were observed for 6 and 7, although the shift of E°x1 potential of 7 was somewhat smaller than expected for a disilirane adduct. Scheme 3. Synthesis of a disilirane adduct of Ce 2 @D 3h -C 78 .

Reactions of La 2 @I h -C 80 with Silirane 8
Scheme 4 shows that La 2 @I h -C 80 also reacts with silirane 8 under thermal conditions to afford the corresponding 1,2-carbosilylated derivatives 9a and 9b [43]. Structural analyses, including X-ray crystallographic measurements, determined 9a and 9b to be a pair of diastereomers of the [5,6]-adducts. Furthermore, electrochemical measurements revealed that carbosilylation slightly altered the electronic properties of La 2 @I h -C 80 . Results of X-ray crystallographic analysis showed that the crystal structure of 9a has eight sites for the La atoms because of disorder [43]. Theoretical calculations of adducts of La 2 @I h -C 80 and 8 indicated that changing the La atom positions resulted in the differences in relative energies within the range of 0.85-5.89 kcal mol −1 . These small energy differences suggest the movability of the La atoms inside the cage [43]. The La atoms sites are arranged along the ten contiguously fused six-membered rings that constitute part of the I h -C 80 cage. Figure 3 presents the redox potentials of pristine and derivatized dimetallofullerenes. Remarkable cathodic shifts of both the E •x 1 and E red 1 potentials were observed for the silylated derivatives 4, 5, 9a and 9b relative to that of pristine La 2 @I h -C 80 . Results also confirmed that the redox potentials shifted cathodically as the number of silicon atoms introduced into La 2 @I h -C 80 increased as shown by those of 4, 5, 9a, 9b and 10 [40,43,44]. The electronic effects of the addends are represented more exactly by comparing the [5,6]-carbosilylated adducts 9a, 9b, the [5,6]-pyrrolidino adduct 10 [44], and the [5,6]-tetracyanoethylene oxide (TCNEO) adduct 11 [45], which have the same addition patterns. As expected from the effects of the electron-donating silyl group and electron-withdrawing TCNEO moieties, the redox potentials of 9a/9b and 11 shifted, respectively, to more negative and more positive potentials than those of reference compound 10. Similar cathodic shifts of the redox potentials were observed for 6 and 7, although the shift of E •x 1 potential of 7 was somewhat smaller than expected for a disilirane adduct.

Reactions of Sc3N@Ih-C80 with Disilirane 1
Details of the structures and physical properties of silylated derivatives of Sc3N@Ih-C80, the most abundant trimetallic nitride template (TNT) fullerene, have been reported [37]. Irradiation of disilirane 1 and Sc3N@Ih-C80 in a mixed solvent of toluene/1,3,5-trichlorobenzene (3/1) produced the corresponding adducts, which were subsequently purified using HPLC to afford a mixture of two adducts 12a and 12b in a ratio of approximately 3:2 (Scheme 5). Although 12a and 12b could not be isolated, the NMR spectra of the mixture suggested that 12a and 12b consisted, respectively, of a 1,2-adduct at the [5,6]-junction, and a 1,4-adduct at the [5,6,6]-junctions on the Ih-C80 cage. It is particularly interesting that when a mixture of 12a and 12b was heated in o-dichlorobenzene (ODCB) at 353 K, facile isomerization of 12a to 12b occurred, which suggests that 12b is thermodynamically more stable than 12a. The structure of 12b was confirmed by X-ray crystallographic analysis to be the 1,4-adduct. Results also indicate the location of the encapsulated Sc3N cluster, which is oriented so that the Y-shaped Sc3N cluster straddles the addition site. Furthermore, theoretical calculations suggest that the free circular motion of the Sc3N cluster is restricted in 12b. Similar addition patterns were reported for the 1,4-adducts, Sc3N@Ih-C80(CF3)2 [46] and Sc3N@Ih-C80(CH2C6H5)2 [47], which were synthesized, respectively, by radical addition reactions of CF3 and CH2C6H5 radicals. [35,40] 4 [40] 5 [40] 9a [43] 9b [43] 10

Reactions of Sc3N@Ih-C80 with Disilirane 1
Details of the structures and physical properties of silylated derivatives of Sc3N@Ih-C80, the most abundant trimetallic nitride template (TNT) fullerene, have been reported [37]. Irradiation of disilirane 1 and Sc3N@Ih-C80 in a mixed solvent of toluene/1,3,5-trichlorobenzene (3/1) produced the corresponding adducts, which were subsequently purified using HPLC to afford a mixture of two adducts 12a and 12b in a ratio of approximately 3:2 (Scheme 5). Although 12a and 12b could not be isolated, the NMR spectra of the mixture suggested that 12a and 12b consisted, respectively, of a 1,2-adduct at the [5,6]-junction, and a 1,4-adduct at the [5,6,6]-junctions on the Ih-C80 cage. It is particularly interesting that when a mixture of 12a and 12b was heated in o-dichlorobenzene (ODCB) at 353 K, facile isomerization of 12a to 12b occurred, which suggests that 12b is thermodynamically more stable than 12a. The structure of 12b was confirmed by X-ray crystallographic analysis to be the 1,4-adduct. Results also indicate the location of the encapsulated Sc3N cluster, which is oriented so that the Y-shaped Sc3N cluster straddles the addition site. Furthermore, theoretical calculations suggest that the free circular motion of the Sc3N cluster is restricted in 12b. Similar addition patterns were reported for the 1,4-adducts, Sc3N@Ih-C80(CF3)2 [46] and Sc3N@Ih-C80(CH2C6H5)2 [47], which were synthesized, respectively, by radical addition reactions of CF3 and CH2C6H5 radicals. [35,40] 4 [40] 5 [40] 9a [43] 9b [43] 10

Reactions of Sc 3 N@I h -C 80 with Disilirane 1
Details of the structures and physical properties of silylated derivatives of Sc 3 N@I h -C 80 , the most abundant trimetallic nitride template (TNT) fullerene, have been reported [37]. Irradiation of disilirane 1 and Sc 3 N@I h -C 80 in a mixed solvent of toluene/1,3,5-trichlorobenzene (3/1) produced the corresponding adducts, which were subsequently purified using HPLC to afford a mixture of two adducts 12a and 12b in a ratio of approximately 3:2 (Scheme 5). Although 12a and 12b could not be isolated, the NMR spectra of the mixture suggested that 12a and 12b consisted, respectively, of a 1,2-adduct at the [5,6]-junction, and a 1,4-adduct at the [5,6,6]-junctions on the I h -C 80 cage. It is particularly interesting that when a mixture of 12a and 12b was heated in o-dichlorobenzene (ODCB) at 353 K, facile isomerization of 12a to 12b occurred, which suggests that 12b is thermodynamically more stable than 12a. The structure of 12b was confirmed by X-ray crystallographic analysis to be the 1,4-adduct. Results also indicate the location of the encapsulated Sc 3 N cluster, which is oriented so that the Y-shaped Sc 3 N cluster straddles the addition site. Furthermore, theoretical calculations suggest that the free circular motion of the Sc 3 N cluster is restricted in 12b. Similar addition patterns were reported for the 1,4-adducts, Sc 3 N@I h -C 80 (CF 3 ) 2 [46] and Sc 3 N@I h -C 80 (CH 2 C 6 H 5 ) 2 [47], which were synthesized, respectively, by radical addition reactions of CF 3 and CH 2 C 6 H 5 radicals. The E°x1 and E red 1 potentials of 12b shifted remarkably to negative values by 540 and 230 mV, respectively, compared to those of pristine Sc3N@Ih-C80, reflecting the addition of two silyl groups ( Figure 4). These shifts of the redox potentials correspond to changes in the lowest unoccupied molecular orbital (LUMO) and highest occupied molecular orbital (HOMO) levels according to theoretical calculations [37]. The HOMO and LUMO levels of 12b are higher than those of pristine Sc3N@Ih-C80. In addition, the HOMO-LUMO energy gap of 12b is smaller than that of Sc3N@Ih-C80. For comparison, the E°x1 and E red 1 potentials of Sc3N@Ih-C80(CF3)2 were reported as +0.47 V and −1.16 V versus the ferrocene/ferrocenium couple (Fc/Fc + ), respectively, in agreement with the electron-withdrawing effects of the CF3 groups [46].

Reactions of Lu3N@Ih-C80 with Silylene
In recent years, among the TNT EMFs, Lu3N@Ih-C80 has emerged as a suitable acceptor for organic solar cells because the LUMO level of Lu3N@Ih-C80 is higher than that of C60 and other TNT EMFs, M3N@Ih-C80 (M = Sc, Y, etc.) [48,49]. Indeed, methano-bridged derivatives of Lu3N@Ih-C80 have been synthesized for application in organic photovoltaic (OPV) devices, which show increases in open circuit voltage and higher efficiencies. For the application of EMFs to OPV devices, the silylation reaction is attractive to adjust the LUMO levels of the EMFs. These results prompted us to investigate the synthesis and properties of organosilicon derivatives of Lu3N@Ih-C80.

Sc 3 N@I
h -C 80 [21] 12b [23] 29a [57] 29b [57]  The E •x 1 and E red 1 potentials of 12b shifted remarkably to negative values by 540 and 230 mV, respectively, compared to those of pristine Sc 3 N@I h -C 80 , reflecting the addition of two silyl groups ( Figure 4). These shifts of the redox potentials correspond to changes in the lowest unoccupied molecular orbital (LUMO) and highest occupied molecular orbital (HOMO) levels according to theoretical calculations [37]. The HOMO and LUMO levels of 12b are higher than those of pristine Sc 3 N@I h -C 80 . In addition, the HOMO-LUMO energy gap of 12b is smaller than that of Sc 3 N@I h -C 80 . For comparison, the E •x 1 and E red 1 potentials of Sc 3 N@I h -C 80 (CF 3 ) 2 were reported as +0.47 V and −1.16 V versus the ferrocene/ferrocenium couple (Fc/Fc + ), respectively, in agreement with the electron-withdrawing effects of the CF 3 groups [46]. The E°x1 and E red 1 potentials of 12b shifted remarkably to negative values by 540 and 230 mV, respectively, compared to those of pristine Sc3N@Ih-C80, reflecting the addition of two silyl groups ( Figure 4). These shifts of the redox potentials correspond to changes in the lowest unoccupied molecular orbital (LUMO) and highest occupied molecular orbital (HOMO) levels according to theoretical calculations [37]. The HOMO and LUMO levels of 12b are higher than those of pristine Sc3N@Ih-C80. In addition, the HOMO-LUMO energy gap of 12b is smaller than that of Sc3N@Ih-C80. For comparison, the E°x1 and E red 1 potentials of Sc3N@Ih-C80(CF3)2 were reported as +0.47 V and −1.16 V versus the ferrocene/ferrocenium couple (Fc/Fc + ), respectively, in agreement with the electron-withdrawing effects of the CF3 groups [46].

Reactions of Lu3N@Ih-C80 with Silylene
In recent years, among the TNT EMFs, Lu3N@Ih-C80 has emerged as a suitable acceptor for organic solar cells because the LUMO level of Lu3N@Ih-C80 is higher than that of C60 and other TNT EMFs, M3N@Ih-C80 (M = Sc, Y, etc.) [48,49]. Indeed, methano-bridged derivatives of Lu3N@Ih-C80 have been synthesized for application in organic photovoltaic (OPV) devices, which show increases in open circuit voltage and higher efficiencies. For the application of EMFs to OPV devices, the silylation reaction is attractive to adjust the LUMO levels of the EMFs. These results prompted us to investigate the synthesis and properties of organosilicon derivatives of Lu3N@Ih-C80.

Reactions of Lu 3 N@I h -C 80 with Silylene
In recent years, among the TNT EMFs, Lu 3 N@I h -C 80 has emerged as a suitable acceptor for organic solar cells because the LUMO level of Lu 3 N@I h -C 80 is higher than that of C 60 and other TNT EMFs, M 3 N@I h -C 80 (M = Sc, Y, etc.) [48,49]. Indeed, methano-bridged derivatives of Lu 3 N@I h -C 80 have been synthesized for application in organic photovoltaic (OPV) devices, which show increases in open circuit voltage and higher efficiencies. For the application of EMFs to OPV devices, the silylation reaction is attractive to adjust the LUMO levels of the EMFs. These results prompted us to investigate the synthesis and properties of organosilicon derivatives of Lu 3 N@I h -C 80 .
Previously, the addition of silylene to C 60 was achieved by photochemical reactions using trisilane compounds 13 [50] as precursors of silylenes (Scheme 6). However, this reaction requires short-wavelength ultraviolet irradiation using low-pressure mercury lamps. When a solution of Lu 3 N@I h -C 80 and 13 was photolyzed via ultraviolet irradiation, no silylene adduct was obtained, whereas Lu 3 N@I h -C 80 was consumed. This result indicates that ultraviolet irradiation is unsuitable for silylene addition reactions of EMFs. Therefore, we synthesized bicyclic silirane 14 as an alternative silylene precursor, which extrudes 15 under thermal reaction conditions [51]. An ODCB solution of 14 and Lu 3 N@I h -C 80 was heated in the dark to afford two silylene mono-adducts 16a and 16b (Scheme 6). Based on NMR analysis, the structures of 16a and 16b were determined respectively to be [5,6]-and [6,6]-adducts. Furthermore, the fulleroid structure of 16a was established via X-ray crystallographic analysis. Adducts 16a and 16b were sensitive to ambient light. They decomposed gradually to afford pristine Lu 3 N@I h -C 80 . In addition, the isomerization of 16b to 16a was also observed upon exposure to ambient light. In contrast, in the addition of silylene to Sc 3 N@I h -C 80  Previously, the addition of silylene to C60 was achieved by photochemical reactions using trisilane compounds 13 [50] as precursors of silylenes (Scheme 6). However, this reaction requires short-wavelength ultraviolet irradiation using low-pressure mercury lamps. When a solution of Lu3N@Ih-C80 and 13 was photolyzed via ultraviolet irradiation, no silylene adduct was obtained, whereas Lu3N@Ih-C80 was consumed. This result indicates that ultraviolet irradiation is unsuitable for silylene addition reactions of EMFs. Therefore, we synthesized bicyclic silirane 14 as an alternative silylene precursor, which extrudes 15 under thermal reaction conditions [51]. An ODCB solution of 14 and Lu3N@Ih-C80 was heated in the dark to afford two silylene mono-adducts 16a and 16b (Scheme 6). Based on NMR analysis, the structures of 16a and 16b were determined respectively to be [5,6]-and [6,6]-adducts. Furthermore, the fulleroid structure of 16a was established via X-ray crystallographic analysis. Adducts 16a and 16b were sensitive to ambient light. They decomposed gradually to afford pristine Lu3N@Ih-C80. In addition, the isomerization of 16b to 16a was also observed upon exposure to ambient light. In contrast, in the addition of silylene to Sc3N@Ih-C80 using 14 under the same conditions, the reaction was not efficient. It afforded a poor product yield, preventing further structural elucidation. Electrochemical measurements yielded the voltammograms of 16a and 16b, which showed irreversible oxidation and reduction processes. The formation of pristine Lu3N@Ih-C80 was observed from electrochemical measurements [51]. The E°x1 potentials of 16a and 16b shifted toward more negative values by 340 and 180 mV, respectively, compared with those of Lu3N@Ih-C80 ( Figure 5). However, the cathodic shifts of the E red 1 potentials of 16a and 16b were observed to be 40 and 130 mV, respectively, relative to that of Lu3N@Ih-C80. As observed for 16a and 16b, the redox behaviors of derivatized EMFs depend on the addition pattern on the carbon cages. The redox potentials of 16a and 16b were also compared with those of the [6,6]-Bingel-Hirsch derivatives 17 [52], 18 [52], 19 [53], the [6,6]-carbene adduct 20 [54], and the [5,6]-carbene adduct 21 [54] (Figure 6). The E°x1 values of the 17-21 were shifted less cathodically than those of 16a and 16b, indicating that the addition of silylenes raises the HOMO levels of the products. In addition, the cathodic shifts of the E red 1 values of the [6,6] Electrochemical measurements yielded the voltammograms of 16a and 16b, which showed irreversible oxidation and reduction processes. The formation of pristine Lu 3 N@I h -C 80 was observed from electrochemical measurements [51]. The E •x 1 potentials of 16a and 16b shifted toward more negative values by 340 and 180 mV, respectively, compared with those of Lu 3 N@I h -C 80 ( Figure 5). However, the cathodic shifts of the E red 1 potentials of 16a and 16b were observed to be 40 and 130 mV, respectively, relative to that of Lu 3 N@I h -C 80 . As observed for 16a and 16b, the redox behaviors of derivatized EMFs depend on the addition pattern on the carbon cages. The redox potentials of 16a and 16b were also compared with those of the [6,6]-Bingel-Hirsch derivatives 17 [52], 18 [52], 19 [53], the [6,6]-carbene adduct 20 [54], and the [5,6]-carbene adduct 21 [54] (Figure 6). The E •x 1 values of the 17-21 were shifted less cathodically than those of 16a and 16b, indicating that the addition of silylenes raises the HOMO levels of the products. In addition, the cathodic shifts of the E

Reactions of Lu 3 N@I h -C 80 with Disilirane and Digermirane
The electronic effects of silylene on the redox properties of EMFs were found to be moderate, as observed for 16a and 16b. Therefore, the reactions between Lu 3 N@I h -C 80 and disiliranes 1 and 2 were investigated to evaluate the electronic effects of bis-silylation on pristine Lu 3 N@I h -C 80 [55]. Photoreaction of Lu 3 N@I h -C 80 and disilirane 1 (or 2) was conducted in a mixed solvent of ODCB/toluene using halogen-tungsten lamps. As shown in Scheme 7, the reaction of 1 afforded 22 and 23, whereas that of 2 afforded 24, 25, and 26, respectively, as 1:1 adducts of Lu 3 N@I h -C 80 and disiliranes. In addition, 22, 24, and 25 were found to be unstable; 22 readily isomerized to 23, whereas 24 and 25 isomerized to 26. Therefore, elucidation of the structures and physical properties of 22, 24, and 25 has been hitherto unsuccessful. The stable adducts 23 and 26 were determined to be the 1,4-adducts on the [5,6,6]-junctions of the I h -C 80 cage via single X-ray crystallographic analysis [56]. Furthermore, the photoreaction of Lu3N@Ih-C80 with digermirane 3 was investigated to compare the electronic properties of the germylated and silylated derivatives [56]. Due to the higher orbital energy levels of C-Ge bonds compared to those of C-Si bonds, it is expected that germylation will perturb the electronic properties of fullerenes effectively. It is particularly interesting that the photoreaction of Lu3N@Ih-C80 with 3 proceeded faster than with 2 to afford the corresponding 1,4-adduct 27 under the same reaction conditions (Scheme 7). The structure of 27 was also confirmed using X-ray crystallographic analysis. Results of our earlier study demonstrate that C60 functions as an electron acceptor in the photoreactions with 1 [25]. Therefore, a similar electron donor-acceptor interaction is presumed to occur in the case of Lu3N@Ih-C80. In fact, the high electron donating property of 3 is explained respectively based on the oxidation peak potentials of 2 (+0.54 V) and 3 (+0.45 V) relative to the Fc/Fc + couple. This result was also supported by comparison of the HOMO levels of 2 and 3 obtained using theoretical calculations. Furthermore, the calculated structures of 2 and 3 indicate that the Ge-Ge and Ge-C bonds of 3 are slightly longer, respectively, than the Si-Si and Si-C bonds of 2. As expected, the high reactivity of 3 can be attributed to its good electron donor properties and its low steric hindrance around the Ge-Ge bond [56].
From electrochemical measurements of 23, 26 and 27, both oxidation and reduction processes were found to be irreversible. Figure 5 shows that the E°x1 values of 23, 26, and 27 shifted to more negative potentials by 530 to 610 mV, compared with that of pristine Lu3N@Ih-C80. Cathodic shifts of E red 1 values were also observed for 23, 26, and 27 within the range of 120 to 220 mV. The E°x1 and E red 1 values of 27 showed slight cathodic and anodic shifts, respectively, compared to the corresponding redox potentials of 23 and 26. As a result, 27 was found to be oxidized and reduced more readily than 23 or 26. These cathodic shifts of E°x1 values of 23, 26 and 27 were confirmed to be remarkably larger than those of 16a and 16b, as expected from the number of silicon and germanium atoms introduced into the fullerene cages.

Reactions of Sc3N@Ih-C80 with Silirane 28
As described above, the addition reaction between Sc3N@Ih-C80 and silylene produced the corresponding adduct in poor yield. This result prompted us to develop an alternative silylation reaction of Sc3N@Ih-C80 using silirane 28 [57]. Photoirradiation of a toluene solution of Sc3N@Ih-C80 and 28 afforded three compounds 29a, 29b, and 29c (Scheme 8). Results of MALDI-TOF mass spectrometry, VIS-NIR and NMR spectroscopy suggest that 29a and 29b are a pair of diastereomers of [5,6]-adducts, whereas 29c might be a [6,6]-adduct. However, further characterization of 29c has not yet been achieved because of its low yield. It is noteworthy that the carbosilylation of La2@Ih-C80 Furthermore, the photoreaction of Lu 3 N@I h -C 80 with digermirane 3 was investigated to compare the electronic properties of the germylated and silylated derivatives [56]. Due to the higher orbital energy levels of C-Ge bonds compared to those of C-Si bonds, it is expected that germylation will perturb the electronic properties of fullerenes effectively. It is particularly interesting that the photoreaction of Lu 3 N@I h -C 80 with 3 proceeded faster than with 2 to afford the corresponding 1,4-adduct 27 under the same reaction conditions (Scheme 7). The structure of 27 was also confirmed using X-ray crystallographic analysis. Results of our earlier study demonstrate that C 60 functions as an electron acceptor in the photoreactions with 1 [25]. Therefore, a similar electron donor-acceptor interaction is presumed to occur in the case of Lu 3 N@I h -C 80 . In fact, the high electron donating property of 3 is explained respectively based on the oxidation peak potentials of 2 (+0.54 V) and 3 (+0.45 V) relative to the Fc/Fc + couple. This result was also supported by comparison of the HOMO levels of 2 and 3 obtained using theoretical calculations. Furthermore, the calculated structures of 2 and 3 indicate that the Ge-Ge and Ge-C bonds of 3 are slightly longer, respectively, than the Si-Si and Si-C bonds of 2. As expected, the high reactivity of 3 can be attributed to its good electron donor properties and its low steric hindrance around the Ge-Ge bond [56].
From electrochemical measurements of 23, 26 and 27, both oxidation and reduction processes were found to be irreversible. Figure 5 shows that the E •x 1 values of 23, 26, and 27 shifted to more negative potentials by 530 to 610 mV, compared with that of pristine Lu 3 N@I h -C 80 . Cathodic shifts of E red 1 values were also observed for 23, 26, and 27 within the range of 120 to 220 mV. The E •x 1 and E red 1 values of 27 showed slight cathodic and anodic shifts, respectively, compared to the corresponding redox potentials of 23 and 26. As a result, 27 was found to be oxidized and reduced more readily than 23 or 26. These cathodic shifts of E •x 1 values of 23, 26 and 27 were confirmed to be remarkably larger than those of 16a and 16b, as expected from the number of silicon and germanium atoms introduced into the fullerene cages.

Reactions of Sc 3 N@I h -C 80 with Silirane 28
As described above, the addition reaction between Sc 3 N@I h -C 80 and silylene produced the corresponding adduct in poor yield. This result prompted us to develop an alternative silylation reaction of Sc 3 N@I h -C 80 using silirane 28 [57]. Photoirradiation of a toluene solution of Sc 3 N@I h -C 80 and 28 afforded three compounds 29a, 29b, and 29c (Scheme 8). Results of MALDI-TOF mass spectrometry, VIS-NIR and NMR spectroscopy suggest that 29a and 29b are a pair of diastereomers of [5,6]-adducts, whereas 29c might be a [6,6]-adduct. However, further characterization of 29c has not yet been achieved because of its low yield. It is noteworthy that the carbosilylation of La 2 @I h -C 80 using 8 was conducted under thermal conditions. However, the reaction of La 2 @I h -C 80 using 8 did not proceed efficiently under photolytic conditions [43].
The E°x1 values of 29a and 29b shifted toward more negative potentials by 380 and 370 mV, respectively, relative to that of Sc3N@Ih-C80 ( Figure 4). However, the E red 1 values of 29a and 29b were nearly equal to that of Sc3N@Ih-C80. These properties were verified based on the HOMO and LUMO energies of 29a and 29b obtained using theoretical calculations. To confirm the electronic effects of carbosilylation, these values were evaluated by comparing with those of the [5,6]-pyrrolidino adducts 30a [58], 30b [58], and 31 [59,60], and the [5,6]-benzyne adduct 32 [61] (Figure 7). Results show that both the E°x1 and E red 1 potentials of 29a and 29b were shifted slightly toward more negative values than those of 30a, 30b, 31 and 32. By coupling the redox properties of 9a and 9b, it appears that carbosilylation of EMFs produces slight cathodic shifts of the redox potentials compared to those produced by derivatization with carbon-based addends.

Dynamic Behaviors of the Encapsulated Atoms
The behavior of encapsulated metal atoms is an interesting issue affecting chemical and physical properties of EMFs. The movements of encapsulated metal atoms are affected by both the electrostatic metal-cage and the metal-metal interactions. In the case of monometallic EMF, metal atoms tend not to move around inside the cages. In dimetallofullerenes with highly-symmetric carbon cages, such as M2@Ih-C80, two atoms can rotate in every direction along the endohedral surface [62][63][64][65]. However, the positions and behaviors of metal atoms, as well as their electronic properties, were found to be altered by exohedral derivatization. For example, the encapsulated Ce atoms in 6 were observed to be fixed at two positions close to the six-membered ring on the equator of the Ih-C80 cage via X-ray crystallographic and paramagnetic 13 C-NMR spectral analysis [41]. In addition, variable temperature (VT) 139 La NMR and X-ray crystallographic analysis of 5 revealed that the La atoms circulate two-dimensionally along the equator of the C80 cage [40]. These results were consistent with earlier reported results of theoretical calculations [66]. The E •x 1 values of 29a and 29b shifted toward more negative potentials by 380 and 370 mV, respectively, relative to that of Sc 3 N@I h -C 80 ( Figure 4). However, the E red 1 values of 29a and 29b were nearly equal to that of Sc 3 N@I h -C 80 . These properties were verified based on the HOMO and LUMO energies of 29a and 29b obtained using theoretical calculations. To confirm the electronic effects of carbosilylation, these values were evaluated by comparing with those of the [5,6]-pyrrolidino adducts 30a [58], 30b [58], and 31 [59,60], and the [5,6]-benzyne adduct 32 [61] (Figure 7). Results show that both the E •x 1 and E red 1 potentials of 29a and 29b were shifted slightly toward more negative values than those of 30a, 30b, 31 and 32. By coupling the redox properties of 9a and 9b, it appears that carbosilylation of EMFs produces slight cathodic shifts of the redox potentials compared to those produced by derivatization with carbon-based addends. The E°x1 values of 29a and 29b shifted toward more negative potentials by 380 and 370 mV, respectively, relative to that of Sc3N@Ih-C80 ( Figure 4). However, the E red 1 values of 29a and 29b were nearly equal to that of Sc3N@Ih-C80. These properties were verified based on the HOMO and LUMO energies of 29a and 29b obtained using theoretical calculations. To confirm the electronic effects of carbosilylation, these values were evaluated by comparing with those of the [5,6]-pyrrolidino adducts 30a [58], 30b [58], and 31 [59,60], and the [5,6]-benzyne adduct 32 [61] (Figure 7). Results show that both the E°x1 and E red 1 potentials of 29a and 29b were shifted slightly toward more negative values than those of 30a, 30b, 31 and 32. By coupling the redox properties of 9a and 9b, it appears that carbosilylation of EMFs produces slight cathodic shifts of the redox potentials compared to those produced by derivatization with carbon-based addends.

Dynamic Behaviors of the Encapsulated Atoms
The behavior of encapsulated metal atoms is an interesting issue affecting chemical and physical properties of EMFs. The movements of encapsulated metal atoms are affected by both the electrostatic metal-cage and the metal-metal interactions. In the case of monometallic EMF, metal atoms tend not to move around inside the cages. In dimetallofullerenes with highly-symmetric carbon cages, such as M2@Ih-C80, two atoms can rotate in every direction along the endohedral surface [62][63][64][65]. However, the positions and behaviors of metal atoms, as well as their electronic properties, were found to be altered by exohedral derivatization. For example, the encapsulated Ce atoms in 6 were observed to be fixed at two positions close to the six-membered ring on the equator of the Ih-C80 cage via X-ray crystallographic and paramagnetic 13 C-NMR spectral analysis [41]. In addition, variable temperature (VT) 139 La NMR and X-ray crystallographic analysis of 5 revealed that the La atoms circulate two-dimensionally along the equator of the C80 cage [40]. These results were consistent with earlier reported results of theoretical calculations [66].

Dynamic Behaviors of the Encapsulated Atoms
The behavior of encapsulated metal atoms is an interesting issue affecting chemical and physical properties of EMFs. The movements of encapsulated metal atoms are affected by both the electrostatic metal-cage and the metal-metal interactions. In the case of monometallic EMF, metal atoms tend not to move around inside the cages. In dimetallofullerenes with highly-symmetric carbon cages, such as M 2 @I h -C 80 , two atoms can rotate in every direction along the endohedral surface [62][63][64][65]. However, the positions and behaviors of metal atoms, as well as their electronic properties, were found to be altered by exohedral derivatization. For example, the encapsulated Ce atoms in 6 were observed to be fixed at two positions close to the six-membered ring on the equator of the I h -C 80 cage via X-ray crystallographic and paramagnetic 13 C-NMR spectral analysis [41]. In addition, variable temperature (VT) 139 La NMR and X-ray crystallographic analysis of 5 revealed that the La atoms circulate two-dimensionally along the equator of the C 80 cage [40]. These results were consistent with earlier reported results of theoretical calculations [66].
The dynamic behaviors of the encapsulated metal atoms in 9a and 9b were also studied using VT 139 La NMR spectroscopy and theoretical calculations [43]. The results showed that the line width of the 139 La signal is broadened at temperatures higher than 280 K, suggesting that the two La atoms move dynamically inside the carbon cage of 9a and 9b. Coupling of these results with the X-ray crystallographic analysis of 9a suggests that the La atoms are moving inside an annular belt formed from the six-membered rings [43].

Potential Applications
As shown in this review, various EMFs and empty fullerenes showed high reactivities toward disiliranes, which have activated Si-Si bonds. Moreover, polysilanes, i.e., polymers comprising of catenated Si-Si bonds as polymer backbones, have attracted considerable interest as novel functional materials because of their unique chemical and physical properties [67,68]. Based on the delocalization of σ-electrons along the backbones of polysilanes, they have been studied as future functional materials, such as conductive polymers [69][70][71][72][73][74][75][76][77][78], photoresists [79], and organic light emitting diodes (LEDs) [80]. Polysilanes are well-known for their high hole mobility, however, their photoinduced charge carrier generation is not efficient. Doping C 60 as an acceptor into polysilanes resulted in photoinduced electron transfer between them, which dramatically improved the charge carrier generation efficiency [70,71]. The electron transfer processes between photoexcited C 60 and polysilanes [81,82] or polygermanes [83] were confirmed by laser flash photolysis experiments.
Due to their photoconductive properties, polysilanes may be applied to novel organic semiconductor devices. Some researchers investigated the photovoltaic properties of bulk heterojunction solar cells using C 60 and [6,6]-phenyl C 61 -butyric acid methyl ester (PCBM) as electron acceptors and polysilanes as electron donors [84][85][86][87]. Furthermore, composite materials composed of polysilanes and EMFs may exhibit improved electronic properties as compared with those of empty fullerene-based materials. The organic solar cells using Lu 3 N@I h -C 80 as an acceptor material showed higher open circuit voltages and higher photovoltaic efficiencies than those of C 60 because of the higher LUMO level of Lu 3 N@I h -C 80 [48,49].
The MO levels of the silylated and germylated EMFs were remarkably altered by the electron donating silyl and germyl groups as compared with those of pristine EMFs. Therefore, silylation and germylation would be useful for adjusting the electronic properties of EMF derivatives in electronic devices. In addition, covalent bonding between fullerenes and polysilanes can be used to construct donor-acceptor conjugate structures. The first prototype polysilane that incorporated C 60 into the polymer backbone was synthesized by the photoreaction of a polysilane with C 60 [88][89][90][91]. Development of silylation and germylation reactions of fullerenes will provide useful methods for affording fullerene-polysilane/polygermane conjugates.

Summary
This review provided an overview of our studies of the synthesis and characterization of silylated and germylated derivatives of EMFs using active silicon compounds and germanium compounds. Among these, disilirane 1 has been used as a versatile substrate that is both photochemically and thermally reactive. A comparison of the redox potentials of EMFs revealed that the reactivities of EMFs toward 1 depend on the electron-accepting abilities of the EMFs. Moreover, NMR and X-ray crystallographic examinations revealed that silylation affects the dynamic behavior of the encapsulated metal atoms and clusters in the EMFs. For example, results show that the movement of the La atoms in La 2 @I h -C 80 was restricted to two-dimensional rotation near the equatorial plane of the carbon cage upon addition of 1. The electronic effects of silylation and germylation on EMFs were clarified by electrochemical examinations and theoretical calculations. In general, the redox potentials of silylated and germylated derivatives shifted to more negative values relative to those of pristine EMFs. In particular, the cathodic shifts of the oxidation potentials were found to be remarkably large, reflecting the electron-donating effects of the silyl and germyl groups. These results demonstrate that EMFs with novel physical properties and functionalities are obtainable through chemical modification by exploiting the properties of various heteroatom-based functional groups.