Isotope-labelling experiments for determining the oxygen source. The successful oxidation of alkoxy indanofullerene 1a under the argon atmosphere, indicating that the oxygen source was not directly from the air. Then, the oxygen source for this one-step oxidation was determined by performing the reaction in the presence of 18O isotope-labelled water (H218O) within a sealed tube (Fig. 2). The control experiment was performed without the addition of H218O under the optimized conditions. Then, the molecular weight of the product 2a was measured by high-resolution mass spectrum (HRMS), which showed a mass-to-charge ratio (m/z) of 824.0260 corresponding to non-18O-labeled 2a (Fig. 2a). When the reaction was carried out in the presence of H218O, a peak at m/z 826.0342 was clearly observed by HRMS, indicating that the obtained ketone contained 18O in its carbonyl group (Fig. 2b). Therefore, the oxygen source for this oxidation reaction is H2O rather than the methoxy group or O2 from the air. Notably, although an excess of H218O was used, the mass peak of non-18O labelled ketone 2a can still be seen in the HRMS spectrum of the 18O-labeled ketone 2a-(18O). (See Supplementary Fig. 1, Tables 1 and 2 for details).
Kinetic studies. Reaction kinetics of this one-step direct oxidation reaction were carried out to further understand the reaction characteristics (Fig. 3). All reactions were performed under the same conditions except varying the temperature. The concentration changes over time were monitored by high-performance liquid chromatography (HPLC) (See Supplementary Figs. 2–5, Tables 3–6 for details). The change in concentration of reactant 1a and product 2a over time clearly indicated that this oxidation reaction reached equilibrium faster and gave higher yield when the reaction temperature increased (Figs. 3a and 3b). No by-products were formed during the transformation of 1a to 2a, demonstrating that this oxidation route has high selectivity and efficiency. The consumption ratio of 1a was plotted on a logarithmic scale to determine the reaction order. The natural logarithm of the 1a consumption ratio exhibited a strong linear time dependence, suggesting that this oxidation reaction exhibit the first-order characteristics (Fig. 3c). The rate constant (k) dramatically increased from 6.4 × 10-4 mol-1 L-1 s-1 to 7.8 × 10-3 mol-1 L-1 s-1 when the reaction temperature was increased from 353 K to 375 K (Table 3). Next, the activation energy Ea, activation enthalpy ΔH‡, activation entropy ΔS‡, and activation Gibbs free energy ΔG‡ were obtained from Arrhenius plots (ln k vs. 1/T) and Eyring plots (ln(k/T) vs. 1/T) on the basis of following equations, respectively (Fig. 3d):9,10
ln k = – Ea/RT + ln A (1)
ln (k/T) = – ΔH‡/RT + [ln (kB/h) + ΔS‡/R] (2)
Here, k is the rate constant, T is the temperature, R is the gas constant, ln A is a constant, kB is the Boltzmann constant, h is the Planck constant. The results summarized in Table 1 indicate that this one-step direct oxidation has an Ea of 120.6 kJ mol-1, with an endothermic ΔH‡ of 116.4 kJ mol-1, a positive ΔS‡ of 6.2 J mol-1 K-1.
Mechanistic studies. To gain more understandings on this one-step oxidation reaction, further investigations were carried out to understand the additional products and active intermediates. In situ proton nuclear magnetic resonance (1H NMR) was applied to analyze the leaving form of the methyl group in 1a (Fig. 4a). As shown in Fig. 4a, 1H NMR of 1a clearly depicted a methyl peak with a chemical shift (δ) at 4.252 ppm. After the reaction was fished, the in situ 1H NMR of reaction mixture indicated a disappearance of methyl peak in 1a, while a new singlet peak appeared at δ = 2.619 ppm. Compared to the methyl peak in methanol (δ = 2.827), which we hypothesized as the potential leaved form of methyl in this reaction, the reaction mixture showed substantially up-field shifted. Additionally, the reaction mixture was examined to be acidic, indicating the generation of acid during the reaction. Accordingly, we hypothesized the 1H NMR signal (δ = 2.619) of reaction mixture should be derived from the methyl in CH3Br,12 which was produced by the reaction between generated HBr and MeOH leaved from 1a especially reacting at high temperature. Meanwhile, when CH3Br was formed, H2O was simultaneously generated, which could then serve again as an oxygen source for this oxidation. Also, this result explained why non-18O labelled 2a was still detected even when we used a large excess of H218O. Therefore, the methyl group in 1a leaved in a methanol form, which further suggests that this oxidation reaction should involve a hemiketal intermediate. Also, the slightly positive ΔS‡ of this reaction reasonably explained the increased disorder because of the additional products of MeOH and HBr. Besides the analysis of additional products that generated during the reaction, further experiments were performed to confirm which active intermediate that mediated this one-step oxidation. Then, the radical scavenger 2,2,6,6-tertramethyl-1-piperidinyloxyl (TEMPO) was applied to confirm the generation of C60•+ intermediate from the single electron transfer between fullerene and CuBr2 (Fig. 4b). When the reaction was run in the presence of 4.0 equiv. of TEMPO, the yield of 2a was dramatically decreased from 95% to 8%. A further increase in amount of TEMPO to 10.0 equiv. stopped the reaction, suggesting that the electron transfer process was completely suppressed. Therefore, the one-step direct oxidation of alkoxy is initiated by electron transfer from C60 to CuBr2, and C60•+ plays a key role in the following oxidation steps.
Our mechanistic insights regarding to the C60•+ intermediate mediated one-step oxidation are provided in Fig. 5. Based on the above experimental results and our previous research,7 we considered that the oxidation of fullerene to C60•+ through single electron transfer in the presence of copper bromide demonstrate a critical role in this reaction. As depicted in Fig. 5, in this one-step oxidation, we hypothesized that the fullerene pendant in 1a is initially oxidized by CuBr2 via single-electron transfer, producing the key active specie, indanofullerenyl radical cation I. Owing to the electron deficiency of C60•+, the neighbouring C–H bond is then cleaved to generate neutral radical II with the release of one proton, which then spontaneously reacting with the isolated bromide anion to form HBr. Next, CuBr2 further oxidizes II to generate corresponding cation III, which undergoes nucleophilic addition by H2O, producing hemiketal intermediate IV. Finally, fullerene-fused ketone 2a is produced through the loss of methanol and deprotonation. Meanwhile, the methanol produced can react with HBr to generate CH3Br and H2O, which then quickly reacts with benzyl cation III (See Supplementary Fig. 6 for details). Therefore, fullerene pendant can facilitate the one-step direct oxidation of the alkoxy group to ketone by serving as an electron pool.
Computation studies. To provide further support for the proposed mechanism, density functional theory (DFT) calculations were performed to understand the key species and reaction barriers (Fig. 6, See Supplementary Table 7 for details). The DFT results indicated that the rate-determining step is proton transfer from H2O to the methoxy group, which had computed potential energy barrier of 124.5 kJ mol-1, in fair agreement with the experiment value. It should be noted that Br– efficiently accelerated this proton transfer, as shown by DFT calculations in the absence of Br– (See Supplementary Fig. 7 and Table 9 for details). In addition, the calculations showed facile oxidation of 1a by copper (II) and relatively easy deprotonation of I to form the benzyl cation III and HBr, with an energy barrier on the order of 38.6 kJ mol-1 (See Supplementary Fig. 8 for details). Therefore, CuBr2 plays two roles in this one-step oxidation reaction: a) oxidation of fullerene via electron transfer with assistance of bromide anion; b) proton transfer for formation of the hemiketal through the formation of Br–.
Performance of evaporable fullerene-fused ketone. So far, fullerene derivatives have been demonstrated as versatile and high-performance electron-transport materials in perovskite solar cells,13,14 but the high-performance fullerene electron-transport materials have never been achieved using vacuum-deposition process.15–17 Accordingly, both the indanofullerene 1a and the produced fullerene-fused ketone 2a were further processed through the vacuum-deposition process to fabricate the electron-transport layer. HPLC analyses of vacuum-deposited 1a-film indicated that 1a is instable in the vacuum-deposition process (Fig. 7a). To our delight, we observed that no thermally decomposed components were detected when vacuum-depositing 2a, which indicated fullerene-fused ketone has stronger stability (Fig. 7b). Further thermogravimetric analyses (TGA) manifested that 2a show high thermal stability with an initial decomposing temperature at 409.5 oC, which is much thermally stable than that of 1a or PC61BM (See Supplementary Fig. 9 for details). Transmission electron microscopy (TEM) was carried to compare the morphology of spin-coated and vacuum-deposited 2a-films, respectively. Fig. 7c indicated that the spin-coated 2a-film show obvious pinholes with substantial crystalline found in the selected area electron diffraction (See Supplementary Fig. 10a for details). In stark contrast, the vacuum-deposited 2a-film exhibits a highly uniform and amorphous morphology, which benefits the high electron-transport performance (Fig. 7d and See Supplementary Fig. 10b for details).6 To evaluate the charge carrier mobility of fullerene-fused ketone 2a, space-charge-limited current (SCLC) measurements was applied to compare the trap-filling limit voltage (VTFL) and trap density (nt) of spin-coated and vacuum-deposited 2a-films, respectively (Figs. 7e & 7f). In well accordance with TEM observations, the spin-coated 2a-film showed more defects with higher VTFL (1.49 V) and nt (1.4 × 1018 cm-3), compared with VTFL (1.01 V) and nt (9.3 × 1017 cm-3) of the vacuum-deposited 2a-film. Moreover, additional SCLC measurements further compared vacuum-deposited C60- and 2a-films (Figs. 7g and 7h). The 2a-film exhibited an equally high electron mobility (2.16 × 10-6 cm2 V-1 s-1) compared with C60-film (2.33 × 10-6 cm2 V-1 s-1), which suggesting that fullerene-fused ketone can be applied as an efficient electron-transport layer to replace the pristine fullerene in perovskite solar cells.
In summary, here we report a facile CuBr2-promoted one-step direct oxidation of alkoxy to ketone with the aid of an oxidizable fullerene pendant. The mechanistic investigation demonstrates in situ generated fullerenyl radical cation (C60•+) behaves as an electron pool to facilitate the one-step direct oxidation: a) initiating oxidation via electron transfer from C60 to CuBr2 to form C60•+ and b) activating cleavage of the neighbouring C–H bond by withdrawing electrons from the bond and subsequently affording the key hemiketal intermediate. Moreover, we found that produced fullerene-fused ketone can form the high-quality electron-transport film using the vacuum-deposited process. Therefore, this reaction will not only provide a useful method in fundamental organic chemistry regarding the direct oxidation of alkoxy to ketone and in fullerene cation chemistry, but also provide an evaporable fullerene material for high-performance electron-transport material in perovskite solar cells.