Diamond Rings or Dumbbells: Controlling the Structure of Poly(ethylene glycol)−Fullerene [60] Adducts by Varying Linking Chain Length

Fullerene [60] (C60) and poly(ethylene glycol) (PEG) adducts were prepared via 1,3-dipolar cycloaddition of bis-azido-terminated linear PEG (N3-PEG-N3) and C60. While long chain length PEGs (Mn = 1000 and 3350 Da) yielded the cycl ic diamond ring structures (⊂ ⊃ C PEG1000 60 and ⊂ ⊃ C PEG3350 60 ), short-chained PEG (Mn = 200 Da) yielded a telechelic dumbbell structure (C60-PEG200-C60). Both dumbbell and diamond ring structures were characterized by MALDI-TOF-MS, H NMR, FTIR, UV−vis, and elemental analysis (EA). The results highlight the structural dependence of the PEG-C60 adducts on the chain length of the azido-terminated polymer, a factor previously neglected. ■ INTRODUCTION Fullerene [60] (C60) has been a focus of scientific attention since its discovery in the mid-1980s, attracting particular interest in its rich chemistry with electron-rich reactive species including carbenes, nitrenes, and azides. The strong electronaccepting character of fullerenes and the versatility of their grafting chemistry have motivated their incorporation in diverse applications, including organic photovoltaics, as antioxidants, and as MRI agents. In many contexts, especially bulk heterojunction solar cells, fullerene’s highly hydrophobic nature has a deterministic effect on phase behavior in polymer blends. However, the low solubility of C60 in common solvents limits direct use in many circumstances. To enhance the compatibility and processability, C60 has been attached covalently to a wide range of polymers, in many different architectures, including “pearl necklace” main chain, side chain pendant, and telechelic configurations; the properties of various polymer blends incorporating C60-polymer adducts have been tested. Both monoand double-fullereneterminated PEGs have received particular attention as simple model systems due to their linearity and ease of preparation. PEG is an appealing backbone for C60 end-capped adducts due to its wide solubility, structural inertness, and the commercial availability of a variety of chain lengths and terminations. Goh and co-workers have promoted the synthesis of dumbbells using the 1,3-dipolar cycloaddition reaction of N3PEG-N3 (Mn for PEG = 2500, 5000, 10 000 and 20 000 g mol−1) with C60 and subsequently prepared blends with poly(acrylic acid), poly(methacrylic acid), and poly(p-vinylphenol). These blends showed an increased glass transition temperature (Tg) compared to their parent polymers, suggesting that addition of PEG-C60 alters the phase behavior of corresponding polymer blends. The synthetic protocol for C60 end-capping followed three steps: (i) the conversion of the terminal PEG hydroxyl to chloride with thionyl chloride, (ii) the formation of N3-PEG-N3 using sodium azide, and (iii) the 1,3-dipolar cycloaddition reaction of PEG-N3 with C60 to yield dumbbells. Although a well-defined synthetic protocol has been reported, the existing characterization appears ambiguous since it mainly relies on UV−vis analysis of the resulting fullerene−polymer adducts. Other core characterization data, particularly NMR and mass spectrometry characterization, have not been reported, although they are vital in determining the product geometry. In addition to dumbbells, a range of monofullerene polymer adducts, including some based on PEG, have been prepared by using end-protected polymers to limit grafting to single fullerenes only. This paper explores the established “PEG dumbbell” reaction using different molecular weight PEG chains to control both monoand bis-fullerene end-capped poly(ethylene glycol) (PEG) adducts. The distinct nature of these products was established by fully characterizing both starting materials and products using matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS), nuclear magnetic resonance spectroscopy (H NMR), Fourier transform infrared spectroscopy (FTIR), and ultraviolet visible spectroscopy (UV−vis). ■ RESULTS AND DISCUSSION The C60-PEG200-C60 dumbbell was synthesized via the 1,3dipolar cycloaddition reaction of N3-PEG-N3 with C60, Received: June 10, 2014 Revised: July 9, 2014 Published: July 21, 2014 Article pubs.acs.org/Macromolecules © 2014 American Chemical Society 4870 dx.doi.org/10.1021/ma501200e | Macromolecules 2014, 47, 4870−4875 Terms of Use CC-BY


■ INTRODUCTION
Fullerene [60] (C 60 ) has been a focus of scientific attention since its discovery in the mid-1980s, 1 attracting particular interest in its rich chemistry with electron-rich reactive species including carbenes, nitrenes, and azides. 2 The strong electronaccepting character of fullerenes and the versatility of their grafting chemistry have motivated their incorporation in diverse applications, including organic photovoltaics, 3,4 as antioxidants, 5 and as MRI agents. 6 In many contexts, especially bulk heterojunction solar cells, fullerene's highly hydrophobic nature has a deterministic effect on phase behavior in polymer blends. 7,8 However, the low solubility of C 60 in common solvents limits direct use in many circumstances. To enhance the compatibility and processability, C 60 has been attached covalently to a wide range of polymers, in many different architectures, including "pearl necklace" main chain, side chain pendant, and telechelic configurations; 9 the properties of various polymer blends incorporating C 60 -polymer adducts have been tested. 10−12 Both mono-and double-fullereneterminated PEGs have received particular attention as simple model systems due to their linearity and ease of preparation. 13−15 PEG is an appealing backbone for C 60 end-capped adducts due to its wide solubility, structural inertness, and the commercial availability of a variety of chain lengths and terminations.
Goh and co-workers have promoted the synthesis of dumbbells using the 1,3-dipolar cycloaddition reaction of N 3 -PEG-N 3 (M n for PEG = 2500, 5000, 10 000 and 20 000 g mol −1 ) with C 60 and subsequently prepared blends with poly(acrylic acid), poly(methacrylic acid), and poly(p-vinylphenol). 12,16,17 These blends showed an increased glass transition temperature (T g ) compared to their parent polymers, suggesting that addition of PEG-C 60 alters the phase behavior of corresponding polymer blends. The synthetic protocol for C 60 end-capping followed three steps: (i) the conversion of the terminal PEG hydroxyl to chloride with thionyl chloride, (ii) the formation of N 3 -PEG-N 3 using sodium azide, and (iii) the 1,3-dipolar cycloaddition reaction of PEG-N 3 with C 60 to yield dumbbells. 12,16−18 Although a well-defined synthetic protocol has been reported, the existing characterization appears ambiguous since it mainly relies on UV−vis analysis of the resulting fullerene−polymer adducts. Other core characterization data, particularly NMR and mass spectrometry characterization, have not been reported, although they are vital in determining the product geometry. 12,16−18 In addition to dumbbells, a range of monofullerene polymer adducts, including some based on PEG, have been prepared by using end-protected polymers to limit grafting to single fullerenes only. 17,19 This paper explores the established "PEG dumbbell" reaction 17 using different molecular weight PEG chains to control both mono-and bis-fullerene end-capped poly(ethylene glycol) (PEG) adducts. The distinct nature of these products was established by fully characterizing both starting materials and products using matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS), nuclear magnetic resonance spectroscopy ( 1 H NMR), Fourier transform infrared spectroscopy (FTIR), and ultraviolet visible spectroscopy (UV−vis).

■ RESULTS AND DISCUSSION
The C 60 -PEG 200 -C 60 dumbbell was synthesized via the 1,3dipolar cycloaddition reaction of N 3 -PEG-N 3 with C 60 , following a modified literature procedure; 17 unexpectedly, longer PEG chains produced diamond ring structures ). In a typical reaction, bisazido-terminated PEGs dissolved in o-dichlorobenzene (o-DCB) were treated with 20 equiv of C 60 at 140°C for 24 h, followed by work-up with solvent extraction or column chromatography (see Experimental Section). Fewer equivalents of C 60 (2.2 equiv, as in the literature procedure) produced similar results; C 60 content was increased in an attempt to favor dumbbell formation. Three new PEG-C 60 adducts (1, 2, and 3) were obtained in 45, 59, and 86% yields, respectively (Scheme 1). The structural assignments were based on MALDI-TOF-MS data and 1 H NMR spectra.
The PEG 3350 -C 60 adduct was selected for the most detailed characterization, as its long PEG chain gave the best solubility. Poor solubility complicated some aspects of spectroscopic analysis (especially NMR, see below) in the case of the PEG 1000 and PEG 200 fullerene adducts. In addition, the PEG 3350 -C 60 adduct is well suited for comparison with the existing literature reports 12 of similar fullerene adducts prepared with PEG 2000 and PEG 5000 . The initial Cl-PEG-Cl and N 3 -PEG-N 3 derivatives were prepared following the procedure reported by Goh et al. 17 The electronegativity difference of nitrogen compared to oxygen allows PEG-azide to be identified by 1 H NMR, since the α-proton chemical shift for the N 3 -PEG-N 3 appears at 3.42 ppm (lit. 3.36 ppm) 20 compared to 3.75 ppm for OHterminated PEG (see Supporting Information Figures S1−S3 for 1 H NMR of HO-PEG 3350 -OH, Cl-PEG 3350 -Cl, and N 3 -PEG 3350 -N 3 , respectively). The transformation of chlorides into azides is confirmed by the presence of azido stretching (ν (asymmetric N 3 ) 2095 cm −1 ) in FTIR ( Figure 1). 17 Further evidence for successful PEG end-group transformation is provided by MALDI-TOF-MS (see Supporting Information Figures S6 and S7 for full MALDI-TOF spectra of Cl-PEG 3350 -Cl and N 3 -PEG 3350 -N 3 , respectively); these data are especially helpful for following the transformation from hydroxyl to chloride due to the similar electronegativity of oxygen and chlorine which makes 1 H NMR assignment less conclusive. Isolated N 3 -PEG-N 3 was subsequently reacted with C 60 to afford C 60 end-capped PEG adducts. 17 Initially, following the literature protocol, 17 THF extraction was used to isolate the PEG 3350 -C 60 adduct from the crude mixture (60 mL of THF per 1 g of C 60 ). However, weak absorption bands, characteristic 21 for pristine C 60 , at ca. 2000− 2500 cm −1 in the FTIR spectrum of the PEG 3350 -C 60 adduct (THF extracted) indicated the continued presence of nongrafted C 60 (Figure 1), presumably due to the modest solubility of pristine C 60 in THF (0.06 mg mL −1 ). 22 Instead, after THF extraction, the PEG 3350 −C 60 product was further extracted with water, exploiting the solubility imparted by its long hydrophilic chain. The FTIR spectrum of water extracted PEG 3350 -C 60 adduct ( Figure 1) confirmed removal of unreacted C 60 . The FTIR analysis of the PEG 3350 -C 60 adduct (water extracted) also confirmed successful azide addition by the complete loss of the Scheme 1. Reaction Scheme for PEG-C 60 Adducts, Dumbbell [1] and Diamond Rings [2, 3] Figure 1. Transmission FTIR spectra of PEG 3350 derivatives. The feature at 2095 cm −1 was assigned to azido stretching whereas the 1720 cm −1 peak was assigned to aziridine (C−N) stretching. characteristic azido absorption band at 2095 cm −1 and the appearance of a medium intensity band at 1720 cm −1 which can be attributed to the aziridine ring (C−N) stretch ( Figure 1). Free aziridine rings show a (C−N) stretch 23 at around 1650 cm −1 , but the strain induced by the fullerene shifts the peak to higher wavenumber. The UV−vis spectrum of pure C 60 shows characteristic UV−vis absorption maxima at 336 and 409 nm, while the PEG itself is featureless in the range measured ( Figure 2). The UV−vis spectrum of the PEG 3350 -C 60 adduct showed strong absorbance at around 324 nm, confirming successful incorporation of C 60 onto PEG. Note that the shift to shorter wavelength is more consistent with the formation of a double addition to the fullerene than a monoadduct. 24,28 Further evidence for the formation of the PEG 3350 -C 60 adduct is provided by 1 H NMR ( Figure 3). Compared to the parent polymer (HO-PEG 3350 -OH), the 1 H NMR spectrum of the PEG 3350 -C 60 adduct shows two new signals at 4.45 and 3.80 ppm, associated with the new bond formation. Because of the conjugated, though not superaromatic, π-system, protons closer to the C 60 surface are deshielded. 25 Hence, the most deshielded new signal, at 4.45 ppm, is assigned to α-protons and the other signal at 3.80 ppm to the β-protons ( Figure 3). Despite the new signals relating to the covalently functionalized C 60 , the overall structure of the adduct is not determined. The two possible geometries include either a bis-fullerene adduct with a single PEG chain connecting two fullerenes (dumbbell) or a monofullerene adduct with both ends of the PEG chain attached to the same fullerene (diamond ring). Unfortunately, 1 H NMR does not resolve the two geometries due to the lack of protons on fullerene. No good 13 C NMR data were obtained due to the long relaxation times involved. 26 Overall, the 1 H NMR data, supported by FTIR and UV−vis, indicate successful covalent addition of N 3 -PEG 3350 -N 3 to C 60 . However, MALDI-TOF-MS was used to confirm that the diamond ring (⊂ ⊃ C PEG 60 ) geometry formed in preference to the dumbbell (C 60 -PEG-C 60 ) structure; a low laser intensity (90 mW) was used to minimize any fragmentation of the polymer chain or dissociation of terminal C 60 molecules. The mass spectrum of the PEG 3350 -C 60 adduct shows a Gaussian distribution of positively charged ions ranging from 2297 to 4720 m/z ( Figure 4). Taking an average of 65 ethylene oxide repeating units as a representative example for the chain (equivalent to the M n derived from MALDI, 27 Figure 4), the theoretically calculated masses ([M + Na] + ) for diamond ring ) and dumbbell (C 60 -PEG 3350 -C 60 ) can be calculated as 3660.7 and 4380.7 Da, respectively. It is clear from the mass spectrum that the reaction of N 3 -PEG 3350 -N 3 with excess C 60 does not yield a dumbbell structure (C 60 -PEG 3350 -C 60 ) or a double PEG chain structure (PEG 3350 -C 60 -PEG 3350 ); rather, the data indicate the formation of a diamond ring structure (single fullerene), at around 720 Da (C 60 = 720 Da) less than that of the theoretically calculated mass of C 60 -PEG 3350 -C 60 . Close inspection of positively charged ions  The MALDI and EA confirm the formation of a monofullerene-substituted PEG 3350 -C 60 adduct. Other monofullerene-substituted linear PEG 3350 chains bearing monoazide (C 60 -PEG 3350 -N 3 ), amine (C 60 -PEG 3350 -NH 2 ), or hydroxyl (C 60 -PEG 3350 -OH) end groups can also be ruled out as follows. While these alternatives vary only slightly in mass, a detailed comparison of the possible masses is only consistent with a diamond ring structure (see Supporting Information Table S1 for theoretically calculated monoisotopic masses for C 60 -PEG 3350 -N 3 , C 60 -PEG 3350 -NH 2 , and C 60 -PEG 3350 -OH). Furthermore, the experimental isotopic pattern for the 65 ethylene oxide repeating unit peak ([M + Na] + = 3660.7 Da) closely matches the theoretical calculated monoisotopic mass of the diamond ring ( Figure 5). The FTIR (Figure 1) excludes significant remaining azide content; a control experiment without fullerene showed that the pure N 3 -PEG 3350 -N 3 did not decompose to amine when heated under identical conditions, as monitored by FTIR (not shown). Furthermore, a ninhydrin test on the C 60 -PEG 3350 adduct found no primary amine.
In order to examine the effect of PEG chain length on the structure of fullerene-PEG adducts, two new bis-azidoterminated PEGs, N 3 -PEG 1000 -N 3 and N 3 -PEG 200 -N 3 were also synthesized under the same reaction conditions described above for N 3 -PEG 3350 -N 3 . In the case of N 3 -PEG 1000 -N 3 , the adduct was isolated using THF extraction (×3) since it was insoluble in water probably due to shorter PEG chain length compared to PEG 3350  The only difference between ⊂ ⊃ C PEG1000 60 and ⊂ ⊃ C PEG3350 60 is the length of the polymer backbone. However, the longer PEG chain offers a benefit in purification. As mentioned above, THF alone was not sufficient to remove all unreacted fullerene. Having a relatively long PEG chain in the fullerene/PEG adducts (PEG 3350 ) makes the adduct water-soluble, simplifying the complete removal of free fullerene since pristine C 60 is insoluble in water. Unlike the longer PEG-C 60 analogues (⊂ ⊃ C PEG1000 60 and ⊂ ⊃ C PEG3350 60 ), the PEG 200 -C 60 adduct was not easily isolated by extraction in water or THF. The combination of the short PEG 200 chain and bis-addition of fullerene (see below) makes the adduct insoluble in both water and THF; instead, unreacted fullerene was removed by column chromatography using toluene as eluent. In contrast to the PEG 3350 and PEG 1000 /fullerene products, the mass spectrum of the isolated PEG 200 -C 60 adduct showed no peaks related to diamond ring (⊂ ⊃ C PEG200 60 ) formation, around the theoretically expected mass range (975.1 m/z for four ethylene oxide repeating units). Instead, a Gaussian mass distribution of positively charged ions ranging from 1100 to 2300 m/z was observed, consistent with the formation of dumbbell structure (C 60 -PEG 200 -C 60 ) (see Supporting Information Figure S8 for full MALDI-TOF spectrum of C 60 -PEG 200 -C 60 ). The UV−vis spectrum ( Figure 2) for C 60 -PEG 200 -C 60 shows no shift in the primary fullerene absorbance, consistent with the addition of a single PEG chain to each fullerene, in this case, rather than the double PEG addition/diamond ring structure. 24,28  It appears that the PEG chain length determines the molecular structure of the resulting PEG-C 60 adducts. Simple space filling models of C 60 -PEG-C 60 and ⊂ ⊃ C PEG 60 based on four ethylene oxide repeating units demonstrate that the PEG 200 chain is insufficiently long enough to wrap around the fullerene for trans-1 addition (the conformation requiring longest PEG chain, Figure 6). Other regioisomers may be unfavorable due to the relatively bulky PEG substituents which may hinder the addition at an adjacent position. In this case, with diamond ring formation suppressed, the dumbbell is the favored conformation. Previous literature suggests that the use of large excess of monoprotected linear polymer favors the addition of two or more polymer chains to each fullerene (10:1 equiv of grafting polymer to C 60 ). 28 In the current experiment, excess C 60 (1:10 relative to azide chain ends) was used in an attempt to favor dumbbell formation. Nevertheless, the diamond ring structure was favored when the PEG chain was sufficiently long. Most examples of more than one polymer attached to fullerene in the literature are synthesized via multiple azomethine ylide addition 28 of independent substituents. A recent example demonstrated a trans-4 isomer of a fullerene bis-adduct in which the fullerene is connected to a metallocene bis-aldehyde to give a cyclic diad (diamond ring) via the Prato reaction. 29 The main focus of current paper is to demonstrate the geometry of PEG-C 60 adduct with various PEG chain lengths; therefore, no HPLC was performed, and the resultant product was believed to be mixture of regioisomers.

■ CONCLUSIONS
Both diamond ring and dumbbell structures were successfully synthesized using the 1,3-dipolar cycloaddition reaction of N 3 -PEG-N 3 with C 60 . The final structure of the polymer−fullerene adducts is determined by the polymer chain length. Previous studies had assumed a dumbbell structure would be formed for all bis-azido-terminated PEG reactions with fullerene, but without sufficient spectroscopic evidence to identify the final structure fully. The current study shows that it is necessary to perform detailed characterization, particularly using MALDI to establish the structure. In fact, diamond ring structures dominate, except where they are sterically frustrated by the shortness of the linking PEG chain. This finding may be generally applicable to polymer/fullerene functionalization via telechelic reactive groups. This new type of fullerene diamond ring structure is potentially relevant to polymer blend and organic bulk heterojunction photovoltaics, as a means to increase the fullerene loading or control the critical phase separation of the charge carrying components. 30 Fullerene aggregation remains a major challenge in the polymer/fullerene active layer in bulk heterojunctions (BHJ) photovoltaics (PV). 7 Controlling the resultant structure of the fullerene derivative may provide an alternative route to enhance current fullerenebased PV performance since dumbbell and diamond ring structures are believed to have different degrees of aggregation. If dumbells are specifically required, more rigid spacers such as polythiophene or polyphenylene might restrict the formation of the diamond ring structure, as well as providing specific electronic function.

■ EXPERIMENTAL SECTION
Materials and Methods. C 60 (purity >99%) was obtained from SES Research, Houston, TX. PEG with M n = 200, 1000, and 3350 Da were purchased from Sigma-Aldrich and azeotropically dried overnight using toluene. Azeotropic distillation was performed with Dean−Stark trap and reflux condenser under N 2 . As water is denser than toluene, residual water is trapped in the bottom of the Dean−Stark trap. Sodium azide (99%), tetrabutylammonium iodide (≥99.5%), toluene (HPLC grade), anhydrous dichlorobenzene (o-DCB), anhydrous dimethylformamide (DMF), and trans-2-[3-(4-tert-butylphenyl)-2methyl-2-propenylidene]malononitrile (DCTB) (≥98%) were obtained from Sigma-Aldrich. Thionyl chloride (≥99.5%) was obtained from Fluka. All other solvents were purchased from VWR and used without further purification. 1 H NMR spectra were acquired on a Bruker 400 MHz NMR spectrometer operating at 9.4 T. The recycle delays for all spectra were one second, with 16 scans per spectrum unless otherwise specified. Fourier transform infrared spectra were recorded on a PerkinElmer ATR Spectrum 100 spectrometer with a resolution of 4 cm −1 and 32 scans per spectrum. Elemental analysis was determined by Mr.
Stephen Boyer at London Metropolitan University, Science Centre, Elemental-Analysis service. UV−vis spectra were recorded on PerkinElmer Lambda 950 spectrometer in toluene. Matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF) data were acquired on a Waters MICROMASS MALDI micro MX with a nitrogen laser at 333 nm in the positive reflectron mode with delayed extraction. Samples (1−2 mg) dissolved in toluene (10 μL) were added to the mixture of 10 μL of predissolved matrix (10 mg/mL DCTB in THF) and 3 μL of NaI (3 mg/mL in THF). Small aliquots of mixture were loaded on to a stainless steel target plate and allowed to air-dry at room temperature. The instrument was calibrated in the range of 500−5000 Da using a standard poly(ethylene glycol). The simulated isotopic peak patterns were constructed using online software MolE -Molecular Mass Calculator v2.02 (http://rna-mdb. cas.albany.edu/RNAmods/masspec/mol.htm).
Synthetic Procedures. Synthesis of Cl-PEG-Cl and N 3 -PEG-N 3 . Cl-PEG-Cl and N 3 -PEG-N 3 were synthesized following the literature procedure reported by Goh et al. 17 In brief, the bis-hydroxylterminated PEG was first converted to bis-chloro-terminated PEG (Cl-PEG-Cl) by reaction with thionyl chloride. The Cl-PEG-Cl was then reacted with sodium azide to afford bis-azido-terminated PEG (N 3 -PEG-N 3 ).
Synthesis of the C 60 -Terminated PEG 3350 Diamond Ring (⊂ ⊃ C PEG3350 60 ). Azido-terminated PEG 3350 (0.1 g, 1 equiv) was dissolved in 15 mL of o-dichlorobenzene in a two-necked 50 mL round-bottom flask fitted with a condenser and connected to vacuum line. C 60 (0.43 g, 20 equiv) was dissolved in 20 mL of odichlorobenzene and transferred into a pressure equalizing dropping funnel and fitted to the round-bottom flask. The C 60 solution was then added dropwise into the stirring PEG solution to ensure the mixture Figure 6. Space-filling models of C 60 -PEG 3350 -C 60 dumbbell and hypothetical PEG 200 -C 60 diamond ring. was mixed thoroughly. Once addition of azide was completed, the mixture was then heated to 140°C and stirred for 24 h. Solvent was evaporated in vacuo, and 25 mL of THF was added to the remaining solid. The THF solution was stirred at room temperature overnight and filtered under gravity to yield a dark brown solid. The solid was then dissolved in 20 mL of water to give a yellow solution and unreacted C 60 was filtered off under gravity. The water was evaporated in vacuo, and the solid was then freeze-dried in vacuo (liquid nitrogen, 0.11 mbar, overnight) to yield a dark brown solid ⊂ ⊃ C PEG3350 60 (0.08 g, 86%). Characterization data for ⊂ ⊃ C PEG3350 60 : R f = 0.26 toluene/ methanol (9/1 v/v). 1  Synthesis of the C 60 -Terminated PEG 1000 Diamond Ring (⊂ ⊃ C PEG1000

60
). The C 60 -terminated PEG 1000 was synthesized following the same procedure described above for ⊂ ⊃ C PEG3350 60 .
The resulting crude product after solvent removal was extracted by THF (3×) since it was not water-soluble. Removal of THF in vacuo Synthesis of the C 60 -Terminated PEG 200 Dumbbell (C 60 -PEG 200 -C 60 ). The C 60 -terminated PEG 200 was synthesized following the same procedure described above for ⊂ ⊃ C PEG3350 60 . The resulting crude product was purified by silica gel column chromatography with toluene/methanol (9/1 v/v) as eluent, giving C 60 -PEG 200 -C 60 (0.04 g, 45%) as a dark brown solid. Characterization data for C60-PEG 200 -C 60 .