Non-Chromatographic Purification of Endohedral Metallofullerenes

The purification of endohedral metallofullerenes by high performance liquid chromatography is very time-consuming and expensive. A number of rapid and inexpensive non-chromatographic methods have thus been developed for large-scale purification of metallofullerenes. In this review, we summarize recent advances in non-chromatographic purification methods of metallofullerenes. Lewis acid-based complexation is one of the most efficient and powerful methods for separation of metallofullerenes from empty fullerenes. The first oxidation potential of metallofullerenes is a critical factor that affects the separation efficiency of the Lewis acid-based method. Supramolecular methods are effective for separation of fullerenes and metallofullerenes that are different in size and shape. Chemical/electrochemical reduction and exohedral functionalization are also utilized to separate and purify metallofullerenes on a large scale.

Metallofullerenes are attractive materials for a wide range of potential applications. In 2001, Gd-fullerenol (Gd@C 82 (OH) n ) was found to have a much higher proton relaxivity than that of the commercial magnetic resonance imaging (MRI) contrast agents [12]. Since then, much work has been done for the development of a metallofullerene-based new class of MRI contrast agents [13,14]. Gd-based metallofullerenes also have the potential for cancer therapy. Aggregated Gd@C 82 (OH) 22 nanoparticles were found to exhibit a high anticancer efficiency and a low toxicity at the same time [15]. It has been found that Gd@C 82 dissolved in water with the aid of poly(ethylene glycol)-block-poly (2-(N,N-diethylamino)ethyl methacrylate) can be used for neutron capture therapy [16]. Besides the applications in biomedicine, metallofullerenes also hold great promise in other fields such as organic photovoltaics [17].
Large-scale production and purification of metallofullerenes is a crucial step toward their practical applications. Metallofullerenes are usually produced by an arc-discharge method in the laboratory [2,3]. The yields of metallofullerenes are typically in the milligram scale or even much lower. Further studies on their applications are hampered by their low yields. In recent years, new techniques in the mass production of metallofullerenes have been developed [18]. Some specific metallofullerenes, e.g., Gd@C 82 , can be produced and purified in a gram scale. This progress will pave the way for future research on applications of metallofullerenes.
In most of the previous studies, metallofullerenes were purified by high performance liquid chromatography (HPLC) [2,3]. Solvent-extract of arc-produced soot normally contains a large amount of the so-called empty fullerenes and various types of endohedral fullerenes. Multi-stage HPLC separation on different kinds of chromatographic columns as stationary phases is required to separate/purify the metallofullerenes. In some cases, tedious HPLC recycle performance is necessary to separate metallofullerene fractions with very close retention times [3]. As for large-scale separation/purification of metallofullerenes, the HPLC method is no doubt unfavorable because it is very time-consuming and expensive. Therefore, rapid and inexpensive non-chromatographic methods for separation/purification of metallofullerenes are highly sought after. This review paper summarizes recent advances of non-chromatographic separation/purification of metallofullerenes, which are much more convenient and rapid than the conventional HPLC isolation. In principle, most of the non-chromatographic methods were developed on the basis of the different properties of fullerenes and metallofullerenes toward complexation, oxidation, reduction, or other chemical reactions. The development of non-chromatographic methods makes it much easier to obtain a large quantity of metallofullerenes than it was before.

Separation/Purification of Metallofullerenes with Lewis Acids
Fullerenes and endohedral fullerenes have π-electron-rich surfaces, so generally they are reactive toward strong Lewis acids. When fullerenes/endohedral fullerenes are mixed with Lewis acids in a solution, complexation and/or redox reaction may take place, leading to precipitation of fullerenes/endohedral fullerenes. The reactivity of fullerenes/endohedral fullerenes toward Lewis acids is correlated to their intrinsic electronic structures. In 1994, Olah and coworkers reported that C 70 forms a complex with AlCl 3 in carbon disulfide much faster than C 60 does [19]. This difference in the reaction rate allows the purification of C 60 on a large scale. Most of C 70 can be removed after a reaction period of several days. AlCl 3 was also used as a strong oxidant to oxidize insoluble metallofullerenes such as Gd@C 72 and Gd@C 74 into soluble cations [20,21]. In 2009, Stevenson and coworkers reported that metallic nitride and oxometallic endohedral fullerenes exhibit different reactivity toward Lewis acids such as AlCl 3 or FeCl 3 [22]. They found that the reactivity order was Sc 4 O 2 @I h -C 80 > Sc 3 N@C 78 > Sc 3 N@C 68 > Sc 3 N@D 5h -C 80 > Sc 3 N@I h -C 80 . The empty-cage fullerenes, such as C 60 and C 70 , are largely unreactive toward the Lewis acids. A separation scheme was designed for the separation of the metallofullerenes on the basis of selective complexation and precipitation.
Recently, our group and coworkers reported a TiCl 4 -based non-chromatographic purification method [23,24], as schematically described in Figure 1, which is much more effective and easy-to-handle than any other Lewis acids. We found that metallofullerenes and empty fullerenes behave very differently toward TiCl 4 complexation. When a certain amount of TiCl 4 is added to an organic solution of fullerene/metallofullerene mixture, the metallofullerenes form a complex with TiCl 4 immediately and then precipitate from the solution. The empty fullerenes, on the contrary, are almost unreactive toward TiCl 4 and remain in the solution. The metallofullerenes and empty fullerenes can be separated from each other very easily by filtration. Then, the metallofullerenes are recovered completely by hydrolysis of the metallofullerene-TiCl 4 complex. One of the advantages of this TiCl 4 -based method is that the purification process is very fast. Typically, only several minutes is needed to precipitate the metallofullerenes completely. The TiCl4-based method is applicable to the separation/purification of all of the metallofullerenes that we have examined so far. Monometallofullerenes M@C2n based on both trivalent metal ions (e.g., Y 3+ , La 3+ , Ce 3+ , Er 3+ , Gd 3+ , Lu 3+ ) and divalent metal ions (e.g., Sm 2+ , Eu 2+ , Tm 2+ , Yb 2+ ) can be separated from empty fullerenes with up to >99% purity in a single step through complexation with TiCl4. Additionally, dimetallofullerenes M2@C2n, and clusterfullerenes, e.g., M2C2@C2n and M3N@C2n, can also be purified with TiCl4 treatment. Some typical results of separation/purification with TiCl4 are described in Figure 2.  The TiCl 4 -based method is applicable to the separation/purification of all of the metallofullerenes that we have examined so far. Monometallofullerenes M@C 2n based on both trivalent metal ions (e.g., Y 3+ , La 3+ , Ce 3+ , Er 3+ , Gd 3+ , Lu 3+ ) and divalent metal ions (e.g., Sm 2+ , Eu 2+ , Tm 2+ , Yb 2+ ) can be separated from empty fullerenes with up to >99% purity in a single step through complexation with TiCl 4 . Additionally, dimetallofullerenes M 2 @C 2n , and clusterfullerenes, e.g., M 2 C 2 @C 2n and M 3 N@C 2n , can also be purified with TiCl 4 treatment. Some typical results of separation/purification with TiCl 4 are described in Figure 2. The TiCl4-based method is applicable to the separation/purification of all of the metallofullerenes that we have examined so far. Monometallofullerenes M@C2n based on both trivalent metal ions (e.g., Y 3+ , La 3+ , Ce 3+ , Er 3+ , Gd 3+ , Lu 3+ ) and divalent metal ions (e.g., Sm 2+ , Eu 2+ , Tm 2+ , Yb 2+ ) can be separated from empty fullerenes with up to >99% purity in a single step through complexation with TiCl4. Additionally, dimetallofullerenes M2@C2n, and clusterfullerenes, e.g., M2C2@C2n and M3N@C2n, can also be purified with TiCl4 treatment. Some typical results of separation/purification with TiCl4 are described in Figure 2.   We have examined the effects of solvent and reaction time on the separation efficiency of the La@C 82 metallofullerene ( Figure 3). When the reaction time is within 10 min, a separation efficiency of >99% is achieved for solvents of carbon disulfide, toluene, o-xylene, or 1,2,4-trichlorobenzene (TCB). However, the separation efficiency decreases gradually if the reaction time is longer than 100 min, which is presumably caused by an irreversible evolution of the La@C 82 -TiCl 4 complex under ambient conditions. We have examined the effects of solvent and reaction time on the separation efficiency of the La@C82 metallofullerene ( Figure 3). When the reaction time is within 10 min, a separation efficiency of >99% is achieved for solvents of carbon disulfide, toluene, o-xylene, or 1,2,4-trichlorobenzene (TCB). However, the separation efficiency decreases gradually if the reaction time is longer than 100 min, which is presumably caused by an irreversible evolution of the La@C82-TiCl4 complex under ambient conditions. Importantly, in addition to pristine metallofullerenes, TiCl4 is equally effective at separating/purifying metallofullerene derivatives. Recently, we reported the solvent-extraction and isolation of missing small-bandgap metallofullerenes through exterior functionalization with CF3 groups [25,26]. The trifluoromethylated derivatives of small-bandgap metallofullerenes, Y@C2n(CF3)m (2n = 60, 70, 72, or 74, m = 1 or 3), exhibit high reactivity toward TiCl4 ( Figure 4) [25], which are similar to the case of pristine metallofullerenes. The removal of the empty fullerenes (e.g., C60, C70) with TiCl4 is a critical step for final purification of M@C2n(CF3)m. Because the derivatives have very similar retention times with C60 and C70 on chromatographic columns, it would be very difficult to isolate the derivatives by HPLC if the sample contains large amounts of C60 and C70.  Importantly, in addition to pristine metallofullerenes, TiCl 4 is equally effective at separating/purifying metallofullerene derivatives. Recently, we reported the solvent-extraction and isolation of missing small-bandgap metallofullerenes through exterior functionalization with CF 3 groups [25,26]. The trifluoromethylated derivatives of small-bandgap metallofullerenes, Y@C 2n (CF 3 ) m (2n = 60, 70, 72, or 74, m = 1 or 3), exhibit high reactivity toward TiCl 4 ( Figure 4) [25], which are similar to the case of pristine metallofullerenes. The removal of the empty fullerenes (e.g., C 60 , C 70 ) with TiCl 4 is a critical step for final purification of M@C 2n (CF 3 ) m . Because the derivatives have very similar retention times with C 60 and C 70 on chromatographic columns, it would be very difficult to isolate the derivatives by HPLC if the sample contains large amounts of C 60 and C 70 . We have examined the effects of solvent and reaction time on the separation efficiency of the La@C82 metallofullerene ( Figure 3). When the reaction time is within 10 min, a separation efficiency of >99% is achieved for solvents of carbon disulfide, toluene, o-xylene, or 1,2,4-trichlorobenzene (TCB). However, the separation efficiency decreases gradually if the reaction time is longer than 100 min, which is presumably caused by an irreversible evolution of the La@C82-TiCl4 complex under ambient conditions. Importantly, in addition to pristine metallofullerenes, TiCl4 is equally effective at separating/purifying metallofullerene derivatives. Recently, we reported the solvent-extraction and isolation of missing small-bandgap metallofullerenes through exterior functionalization with CF3 groups [25,26]. The trifluoromethylated derivatives of small-bandgap metallofullerenes, Y@C2n(CF3)m (2n = 60, 70, 72, or 74, m = 1 or 3), exhibit high reactivity toward TiCl4 ( Figure 4) [25], which are similar to the case of pristine metallofullerenes. The removal of the empty fullerenes (e.g., C60, C70) with TiCl4 is a critical step for final purification of M@C2n(CF3)m. Because the derivatives have very similar retention times with C60 and C70 on chromatographic columns, it would be very difficult to isolate the derivatives by HPLC if the sample contains large amounts of C60 and C70.  To gain a better understanding on the complexation of metallofullerenes with TiCl 4 , we investigated the origin and mechanism of the separation/purification process. We found that the first oxidation potential is a crucial physical parameter that determines the reactivity of fullerenes and metallofullerenes toward TiCl 4 complexation [24]. The reactivity of fullerenes and metallofullerenes with first oxidation potentials ranging from −0.07 to 1.26 V versus Fc/Fc + (Fc = ferrocene) has been examined. Interestingly, metallofullerenes with first oxidation potentials of −0.07-0.62 V are reactive toward TiCl 4 ; on the other hand, the empty fullerenes with first oxidation potentials of 0.72-1.26 V are almost nonreactive. Thus, the threshold value in the first oxidation potential for reaction with TiCl 4 is determined to be 0.62-0.72 V ( Figure 5). To the best of our knowledge, all of the reported metallofullerenes have first oxidation potentials below or close to this threshold value. Thus, the TiCl 4 -based purification method can be applied to virtually all of the reported metallofullerenes. For example, a nitride clusterfullerene Dy 3 N@I h -C 80 , which has a first oxidation potential of 0.70 V, can be purified efficiently with TiCl 4 in our laboratory. To gain a better understanding on the complexation of metallofullerenes with TiCl4, we investigated the origin and mechanism of the separation/purification process. We found that the first oxidation potential is a crucial physical parameter that determines the reactivity of fullerenes and metallofullerenes toward TiCl4 complexation [24]. The reactivity of fullerenes and metallofullerenes with first oxidation potentials ranging from −0.07 to 1.26 V versus Fc/Fc + (Fc = ferrocene) has been examined. Interestingly, metallofullerenes with first oxidation potentials of −0.07-0.62 V are reactive toward TiCl4; on the other hand, the empty fullerenes with first oxidation potentials of 0.72-1.26 V are almost nonreactive. Thus, the threshold value in the first oxidation potential for reaction with TiCl4 is determined to be 0.62-0.72 V ( Figure 5). To the best of our knowledge, all of the reported metallofullerenes have first oxidation potentials below or close to this threshold value. Thus, the TiCl4-based purification method can be applied to virtually all of the reported metallofullerenes. For example, a nitride clusterfullerene Dy3N@Ih-C80, which has a first oxidation potential of 0.70 V, can be purified efficiently with TiCl4 in our laboratory. Figure 5. Plot of first oxidation potential (vs. ferrocene (Fc/Fc +) ) of fullerenes and metallofullerenes and the threshold (dashed line) in the first oxidation potential for reaction with TiCl4. Blue and red colors denote the reactive and nonreactive fullerenes/metallofullerenes, respectively. Reproduced with permission from [24]. Copyright 2012 American Chemical Society.
The reaction mechanism between metallofullerenes and TiCl4 was investigated by monitoring the absorption spectrum. It was found that the absorption features of metallofullerene disappear and a new band from metallofullerene cation arises immediately after the addition of TiCl4 [24]. This dramatic change indicates that electron transfer occurs from metallofullerene to TiCl4. Element analysis reveals that the molar ratio of metallofullerene to TiCl4 is about 1:18-19 in the metallofullerene-TiCl4 complex [23]. The metallofullerene might be surrounded by the TiCl4 molecules. Further work is needed to determine the precise structure of the complex. There are several reports on the interaction between TiCl4 and empty cage fullerenes [27][28][29]. For example, TiCl4 has been used to chlorinate C70 fullerene [27]. The structures of TiCl4-solvated empty fullerenes, such as C60•3TiCl4 and C70•2TiCl4, have been crystallographically characterized [28,29]. It is possible to disclose the structure of the metallofullerene-TiCl4 complex by crystallographic study in the future.
Besides TiCl4, there are some other metal halides (e.g., AlCl3, FeCl3, and CuCl2) that can be used to separate and purify metallofullerenes through selective complexation [22,[30][31][32]. The metal halides differ from each other with respect to their complexation abilities toward metallofullerenes. As compared with TiCl4, CuCl2 is a much weaker Lewis acid. It has been found that the precipitation threshold in the first oxidation potential for CuCl2 is about 0.19 V (vs. Fc/Fc + ) [30], which is much Figure 5. Plot of first oxidation potential (vs. ferrocene (Fc/Fc +) ) of fullerenes and metallofullerenes and the threshold (dashed line) in the first oxidation potential for reaction with TiCl 4 . Blue and red colors denote the reactive and nonreactive fullerenes/metallofullerenes, respectively. Reproduced with permission from [24]. Copyright 2012 American Chemical Society.
The reaction mechanism between metallofullerenes and TiCl 4 was investigated by monitoring the absorption spectrum. It was found that the absorption features of metallofullerene disappear and a new band from metallofullerene cation arises immediately after the addition of TiCl 4 [24]. This dramatic change indicates that electron transfer occurs from metallofullerene to TiCl 4 . Element analysis reveals that the molar ratio of metallofullerene to TiCl 4 is about 1:18-19 in the metallofullerene-TiCl 4 complex [23]. The metallofullerene might be surrounded by the TiCl 4 molecules. Further work is needed to determine the precise structure of the complex. There are several reports on the interaction between TiCl 4 and empty cage fullerenes [27][28][29]. For example, TiCl 4 has been used to chlorinate C 70 fullerene [27]. The structures of TiCl 4 -solvated empty fullerenes, such as C 60 •3TiCl 4 and C 70 •2TiCl 4 , have been crystallographically characterized [28,29]. It is possible to disclose the structure of the metallofullerene-TiCl 4 complex by crystallographic study in the future.
Besides TiCl 4 , there are some other metal halides (e.g., AlCl 3 , FeCl 3 , and CuCl 2 ) that can be used to separate and purify metallofullerenes through selective complexation [22,[30][31][32]. The metal halides differ from each other with respect to their complexation abilities toward metallofullerenes. As compared with TiCl 4 , CuCl 2 is a much weaker Lewis acid. It has been found that the precipitation threshold in the first oxidation potential for CuCl 2 is about 0.19 V (vs. Fc/Fc + ) [30], which is much lower than that for TiCl 4 (0.62-0.72 V). Therefore, CuCl 2 can be used to selectively precipitate the most reactive metallofullerenes. For example, metallofullerenes Sc 3 C 2 @I h -C 80 and Sc 4 O 2 @I h -C 80 , with a first oxidation potential of −0.03 and 0.00 V, respectively, can be separated rapidly from a more inert metallofullerene Sc 3 N@I h -C 80 by using CuCl 2 ; Sc 3 N@I h -C 80 can be subsequently purified through a second step of complexation with a stronger Lewis acid (e.g., FeCl 3 ) [30]. For a mixture of metallofullerenes, the purification process can be greatly simplified if a number of Lewis acids are used consecutively to separate the metallofullerenes into distinct classes in terms of their reactivity. Stevenson and coworkers investigated the reactivity of a series of Lewis acids and reported the order of increasing reactivity as follows: CaCl 2 < ZnCl 2 < NiCl 2 < MgCl 2 < MnCl 2 < CuCl 2 < WCl 4 WCl 6 < ZrCl 4 < AlCl 3 < FeCl 3 [31]. They have isolated Gd 3 N@D 2 (35)-C 88 and CeLu 2 N@I h -C 80 by using the Lewis acid method [32,33]. The precipitation thresholds in the first oxidation potential of metallofullerenes for different Lewis acids are summarized in Table 1. Because the metallofullerene isomers usually have different oxidation potentials, it is possible to separate them on the basis of selective chemical oxidation. In 2005, Echegoyen and coworkers reported that I h and D 5h isomers of Sc 3 N@C 80 can be separated by selective oxidation with tris(pbromophenyl)aminium hexachloroantimonate (TPBAH) [34]. Because of its lower first oxidation potential, the D 5h isomer of Sc 3 N@C 80 is oxidized preferentially by TPBAH. The resulted cation of Sc 3 N@D 5h -C 80 and neutral Sc 3 N@I h -C 80 are then separated with a silica column. This method was recently modified by using acetylferrocenium tetrakis(pentafluorophenyl)borate ([Fe(COCH 3 C 5 H 4 )Cp][TFAB]) instead of TPBAH as the oxidant [35]. The oxidation ability of [Fe(COCH 3 C 5 H 4 )Cp][TFAB] is weaker than that of TPBAH. In addition to Sc 3 N@D 5h -C 80 and Sc 3 N@I h -C 80 , Sc 3 N@C 68 and Sc 3 N@C 78 are also purified or fractionated by using [Fe(COCH 3
For the achievement of size-selective extraction, we have used cycloparapheneylenes (CPPs), which are cyclic compounds composed of solely para-substituted benzenes (Figure 6a) [44][45][46]. Some groups including the present laboratory have already reported that the pi-conjugated system is effective for the metallofullerene-selective complexation [40][41][42][43]. In order to realize size-selectivity, cyclic paraphenylene cavities are suitable candidates for the encapsulation of fullerenes [47][48][49]. Notably, [11]CPP has an ideal inner space with a diameter of 1.51 nm [50] for the inclusion of M@C 82 families (0.85 nm in diameter) considering the interlayer distance of convex-concave pi-pi interactions such as fullerene nano-peapods (ca. 0.33 nm) [51]. A theoretical calculation also suggested that C 82 could be fit almost completely into the cavity of [11]CPP (Figure 6b) [52]. After the establishment of the first selective synthesis of [11]CPP [53], we determined the binding abilities of [11]CPP to a series of C82-based metallofullerenes by fluorescence quenching experiments [52]. The fluorescence intensity of [11]CPP gradually decreased with the addition of metallofullerenes (Figure 6c). The binding constant (Ka) of [11]CPP with Gd@C82 is 1.8 × 10 6 M −1 , which is one of the highest values for the complexes of higher metallofullerenes [40][41][42]. With other metallofullerenes such as Tm@C82, Lu2@C82, and Sc3N@C80, subequal binding constants were recorded. In contrast, the strong fluorescence quenching of larger [12]CPP [54] (1.66 nm in diameter) was not observed with the addition of C82-based metallofullerenes.
The separation protocol of Gd@C82 is described below. To a toluene solution of the fullerene mixture from raw soot that included Gd@C82, [11]CPP was added. After the solution was concentrated under reduced pressure, a brown solid precipitated out. Low solubility of the complex between Gd@C82 and [11]CPP realizes this non-chromatographic extraction separation technology. A positive-ion laser desorption mass spectra (LD-MS) of the crude extract, the filtrate solution, and the separated precipitate are shown in Figure 7. The depression of the Gd@C82 peak in the filtrate and a strong peak of Gd@C82 with the fragmentation of [11]CPP from the precipitate were observed. These results indicate that most of Gd@C82 are collected as the precipitate. It should be emphasized that higher metallofullerenes such as Gd@C94 are not extracted by the complexation with [11]CPP due to size mismatch between the fullerene molecule and the cavity of [11]CPP. Table 2 summarizes a comparison of the extraction efficiency of Gd@C82 (yield, amount of solvent used, and time required) using the complexation method and the multistage HPLC method. The HPLC extraction of Gd@C82 was carried out for the same amount (10 mg) of crude fullerene mixture. The amounts of each Gd@C82 obtained were determined from the reported absorption coefficient. Although the HPLC method is somewhat superior in yield (72 µg vs. 56 µg), the present complexation method is much superior to the multistage HPLC method in terms of the amount of solvent used (50 mL vs. 5 L) and the time required for extraction (30 min vs. 5 h). After the establishment of the first selective synthesis of [11]CPP [53], we determined the binding abilities of [11]CPP to a series of C 82 -based metallofullerenes by fluorescence quenching experiments [52]. The fluorescence intensity of [11]CPP gradually decreased with the addition of metallofullerenes (Figure 6c). The binding constant (K a ) of [11]CPP with Gd@C 82 is 1.8 × 10 6 M −1 , which is one of the highest values for the complexes of higher metallofullerenes [40][41][42]. With other metallofullerenes such as Tm@C 82 , Lu 2 @C 82 , and Sc 3 N@C 80 , subequal binding constants were recorded. In contrast, the strong fluorescence quenching of larger [12]CPP [54] (1.66 nm in diameter) was not observed with the addition of C 82 -based metallofullerenes.
The separation protocol of Gd@C 82 is described below. To a toluene solution of the fullerene mixture from raw soot that included Gd@C 82 , [11]CPP was added. After the solution was concentrated under reduced pressure, a brown solid precipitated out. Low solubility of the complex between Gd@C 82 and [11]CPP realizes this non-chromatographic extraction separation technology. A positive-ion laser desorption mass spectra (LD-MS) of the crude extract, the filtrate solution, and the separated precipitate are shown in Figure 7. The depression of the Gd@C 82 peak in the filtrate and a strong peak of Gd@C 82 with the fragmentation of [11]CPP from the precipitate were observed. These results indicate that most of Gd@C 82 are collected as the precipitate. It should be emphasized that higher metallofullerenes such as Gd@C 94 are not extracted by the complexation with [11]CPP due to size mismatch between the fullerene molecule and the cavity of [11]CPP. Table 2 summarizes a comparison of the extraction efficiency of Gd@C 82 (yield, amount of solvent used, and time required) using the complexation method and the multistage HPLC method. The HPLC extraction of Gd@C 82 was carried out for the same amount (10 mg) of crude fullerene mixture. The amounts of each Gd@C 82 obtained were determined from the reported absorption coefficient. Although the HPLC method is somewhat superior in yield (72 µg vs. 56 µg), the present complexation method is much superior to the multistage HPLC method in terms of the amount of solvent used (50 mL vs. 5 L) and the time required for extraction (30 min vs. 5 h).
After our report of the size-selective extraction of metallofullerenes with CPPs, the group of Yamago revealed the partial electron transfer from [11]CPP to La@C 82 [55]. This polar complex formation might support the selective recognition of metallofullerene and trigger a change to poor solubility in non-polar solvents like toluene.  After our report of the size-selective extraction of metallofullerenes with CPPs, the group of Yamago revealed the partial electron transfer from [11]CPP to La@C82 [55]. This polar complex formation might support the selective recognition of metallofullerene and trigger a change to poor solubility in non-polar solvents like toluene.

Separation/Purification of Metallofullerenes through Chemical/Electrochemical Reduction
The electrochemical properties of empty fullerenes and metallofullerenes have been extensively studied during the past two decades [3,56]. The reported redox potentials provide useful information for separation/purification of metallofullerenes based on selective reduction/oxidation. Due to their higher first reduction potentials, the metallofullerenes are much easier to reduce as compared with the empty fullerenes. For example, La@C82 and La2@C80 in TCB extract of La soot can be selectively reduced into anions electrochemically [57]. The anions can easily be separated from neutral fullerenes as they have different solubilities in organic solvents. Subsequent oxidation of the anions with dichloroacetic acid leads to the recovery of the neutral metallofullerenes. In addition to the classical metallofullerene M@C82, there are a large number of the so-called missing small-bandgap metallofullerenes such as M@C60 and M@C70 [2]. The small-bandgap metallofullerenes are insoluble in organic solvents because they tend to form polymers or oligomers in raw soot. Alford and coworkers reported that the small-bandgap metallofullerenes become soluble and stable in organic solvents after electrochemical reduction [58], thus enabling further separation/purification of these unconventional metallofullerenes.
When an organic solvent such as carbon disulfide or o-xylene is used to extract metallofullerenes from raw soot, a large number of empty fullerenes are also dissolved in the solvent. To selectively extract metallofullerenes, some chemical reduction-based routes have been developed. Gu and coworkers reported the simultaneous reduction and extraction of metallofullerenes using Al-Ni alloy in a mixed solvent of toluene and tetrahydrofuran (THF) [59,60]. They found that the composition of the extracted metallofullerenes is dependent on the composition of the solvents used. Specifically, a mixture of Gd@C82/Gd2@C80 or pure Gd2@C80 can be extracted by tuning the ratio of toluene to THF.

Separation/Purification of Metallofullerenes through Chemical/Electrochemical Reduction
The electrochemical properties of empty fullerenes and metallofullerenes have been extensively studied during the past two decades [3,56]. The reported redox potentials provide useful information for separation/purification of metallofullerenes based on selective reduction/oxidation. Due to their higher first reduction potentials, the metallofullerenes are much easier to reduce as compared with the empty fullerenes. For example, La@C 82 and La 2 @C 80 in TCB extract of La soot can be selectively reduced into anions electrochemically [57]. The anions can easily be separated from neutral fullerenes as they have different solubilities in organic solvents. Subsequent oxidation of the anions with dichloroacetic acid leads to the recovery of the neutral metallofullerenes. In addition to the classical metallofullerene M@C 82 , there are a large number of the so-called missing small-bandgap metallofullerenes such as M@C 60 and M@C 70 [2]. The small-bandgap metallofullerenes are insoluble in organic solvents because they tend to form polymers or oligomers in raw soot Alford and coworkers reported that the small-bandgap metallofullerenes become soluble and stable in organic solvents after electrochemical reduction [58], thus enabling further separation/purification of these unconventional metallofullerenes.
When an organic solvent such as carbon disulfide or o-xylene is used to extract metallofullerenes from raw soot, a large number of empty fullerenes are also dissolved in the solvent. To selectively extract metallofullerenes, some chemical reduction-based routes have been developed. Gu and coworkers reported the simultaneous reduction and extraction of metallofullerenes using Al-Ni alloy in a mixed solvent of toluene and tetrahydrofuran (THF) [59,60]. They found that the composition of the extracted metallofullerenes is dependent on the composition of the solvents used. Specifically, a mixture of Gd@C 82 /Gd 2 @C 80 or pure Gd 2 @C 80 can be extracted by tuning the ratio of toluene to THF. Their work provides a simple method for selective extraction of M 2 @C 80 metallofullerene. Kodama and coworkers reported that Ce-metallofullerenes (Ce@C 82 , Ce 2 @C 78 , Ce 2 @C 80 ) can be extracted selectively using a mixed solvent of triethylamine and acetone [61]. They proposed that the reduction of Ce-metallofullerenes by triethylamine or the formation of metallofullerene-triethylamine donor-accepter complexes is responsible for the selective extraction of metallofullerenes.
Unlike non-polar solvents such as carbon disulfide and o-xylene, N,N-dimethylformamide (DMF) was found to be a special solvent that can dissolve metallofullerenes preferentially, excepting empty fullerenes [62][63][64][65][66]. Akasaka and coworkers proposed that the metallofullerenes dissolved in DMF are actually in the form of anions, not neutral metallofullerenes; these might be formed during the high-temperature extraction process [66]. As such, metallofullerenes with high first reduction potentials dominate the DMF extract.

Separation/Purification of Metallofullerenes through Exohedral Organic Reactions
Organic chemistry of fullerenes/metallofullerenes has been an active research area since their discovery. Exohedral functionalization of fullerenes and metallofullerenes provides new routes to realize efficient separation of fullerenes and metallofullerenes. Dorn and coworkers reported that metal nitride clusterfullerenes can be separated from empty fullerenes and classical metallofullerenes by using a column packed with cyclopentadienyl-functionalized resin (Figure 8a) [67]. Empty fullerenes and metallofullerenes are bound to the resin by the Diels-Alder reaction, whereas, metal nitride clusterfullerenes remain intact and pass through the column due to their inertness. The bound empty fullerenes and metallofullerenes can be recovered upon heating in the presence of maleic anhydride. Stevenson and coworkers reported an alternative way to separate metal nitride clusterfullerenes through selective reaction with cyclopentadienyl-or amino-functionalized silica, which they call the "Stir and Filter Approach" (SAFA) (Figure 8b) [68][69][70][71][72]. This approach can be used to remove Sc 3 N@D 5h -C 80 rapidly from its structural isomer, Sc 3 N@I h -C 80 [70]. Further studies demonstrate that the content of water and the substituents on the aromatic solvent molecules have significant effects on the separation efficiency of metal nitride clusterfullerenes when using amino-functionalized silica [72].
Their work provides a simple method for selective extraction of M2@C80 metallofullerene. Kodama and coworkers reported that Ce-metallofullerenes (Ce@C82, Ce2@C78, Ce2@C80) can be extracted selectively using a mixed solvent of triethylamine and acetone [61]. They proposed that the reduction of Ce-metallofullerenes by triethylamine or the formation of metallofullerene-triethylamine donoraccepter complexes is responsible for the selective extraction of metallofullerenes.
Unlike non-polar solvents such as carbon disulfide and o-xylene, N,N-dimethylformamide (DMF) was found to be a special solvent that can dissolve metallofullerenes preferentially, excepting empty fullerenes [62][63][64][65][66]. Akasaka and coworkers proposed that the metallofullerenes dissolved in DMF are actually in the form of anions, not neutral metallofullerenes; these might be formed during the high-temperature extraction process [66]. As such, metallofullerenes with high first reduction potentials dominate the DMF extract.

Separation/Purification of Metallofullerenes through Exohedral Organic Reactions
Organic chemistry of fullerenes/metallofullerenes has been an active research area since their discovery. Exohedral functionalization of fullerenes and metallofullerenes provides new routes to realize efficient separation of fullerenes and metallofullerenes. Dorn and coworkers reported that metal nitride clusterfullerenes can be separated from empty fullerenes and classical metallofullerenes by using a column packed with cyclopentadienyl-functionalized resin (Figure 8a) [67]. Empty fullerenes and metallofullerenes are bound to the resin by the Diels-Alder reaction, whereas, metal nitride clusterfullerenes remain intact and pass through the column due to their inertness. The bound empty fullerenes and metallofullerenes can be recovered upon heating in the presence of maleic anhydride. Stevenson and coworkers reported an alternative way to separate metal nitride clusterfullerenes through selective reaction with cyclopentadienyl-or amino-functionalized silica, which they call the "Stir and Filter Approach" (SAFA) (Figure 8b) [68][69][70][71][72]. This approach can be used to remove Sc3N@D5h-C80 rapidly from its structural isomer, Sc3N@Ih-C80 [70]. Further studies demonstrate that the content of water and the substituents on the aromatic solvent molecules have significant effects on the separation efficiency of metal nitride clusterfullerenes when using aminofunctionalized silica [72].  In addition to the support-based separation strategies, some support-free chemical methods have also been reported for purification of metal nitride clusterfullerenes. For example, molten 9-methylanthracene was employed to react with a solid mixture of empty fullerenes and metal nitride clusterfullerenes [73]. The empty fullerenes were converted to 9-methylanthracene adducts that have different solubilities in organic solvents as compared with intact metal nitride clusterfullerenes. After washing away the fullerene derivatives with ether, metal nitride clusterfullerenes was enriched to about 60% purity within one day. Another example of a non-chromatographic method is based on selective extraction with 2-aminoethanol [74]. Empty cage fullerenes are much more reactive than the endohedral metallofullerenes Sc 3 N@C 2n (n = 34, 39, 40) toward nucleophilic 2-aminoethanol. The empty cage fullerenes were selectively transferred to the 2-aminoethanol layer after reaction, and the endohedral metallofullerenes Sc 3 N@C 2n were enriched in the original solution. Wang and coworkers reported a purification method based on Prato-type 1,3-dipolar cycloaddition [75]. Both empty fullerenes and metal nitride clusterfullerenes undergo cycloaddition with N-ethylglycine and paraformaldehyde [76]; subsequent retro-cycloaddition using 3-chloroperoxybenzoic acid, however, only takes place for the derivatives of metal nitride clusterfullerenes. This difference realizes a facile separation of metal nitride clusterfullerenes from empty fullerenes.

Summary
In order to realize convenient and rapid separation/purification of metallofullerenes, a number of non-chromatographic methods have been developed. Lewis acid-based complexation is one of the most efficient and powerful methods for separation of metallofullerenes from empty fullerenes. For this Lewis acid-based method, the separation efficiency is basically determined by the first oxidation potentials of the metallofullerenes. Supramolecular methods are effective for separation of fullerenes and metallofullerenes that are different in size and shape. The metallofullerene M@C 82 can be isolated through selective complexation with [11]CPP. Chemical/electrochemical reduction and exohedral functionalization are also utilized to separate metallofullerenes in a large scale. It is much easier to obtain a large quantity of metallofullerenes than it was before by using the non-chromatographic methods. The development of non-chromatographic methods is a crucial step toward future industrial production of metallofullerenes.