Gram-scale synthesis of a covalent nanocage that preserves the redox properties of encapsulated fullerenes

Discrete nanocages provide a way to solubilize, separate, and tune the properties of fullerenes, but these 3D receptors cannot usually be synthesized easily from inexpensive starting materials, limiting their utility. Herein, we describe the first fullerene-binding nanocage (Cage4+) that can be made efficiently on a gram scale. Cage4+ was prepared in up to 57% yield by the formation of pyridinium linkages between complemantary porphyrin components that are themselves readily accessible. Cage4+ binds C60 and C70 with large association constants (>108 M−1), thereby solubilizing these fullerenes in polar solvents. Fullerene association and redox-properties were subsequently investigated across multiple charge states of the host-guest complexes. Remarkably, neutral and singly reduced fullerenes bind with similar strengths, leaving their 0/1− redox couples minimally perturbed and fully reversible, whereas other hosts substantially alter the redox properties of fullerenes. Thus, C60@Cage4+ and C70@Cage4+ may be useful as solubilized fullerene derivatives that preserve the inherent electron-accepting and electron-transfer capabilities of the fullerenes. Fulleride dianions were also found to bind strongly in Cage4+, while further reduction is centered on the host, leading to lowered association of the fulleride guest in the case of C602−.


S5
rotary evaporation, followed by suspending the red-brown residue in water, filtering, and washing with deionized H2O (3 x 100 mL). At this stage, TLC (silica, 95:5 DCM:MeOH) shows several spots including four purple bands, the slowest of which corresponds the desired product. The purple solid was then dissolved in DCM, combined with silica gel, and evaporated to dryness. Column chromatography was then used to isolate the desired porphyrin by first eluting with 5% MeOH in DCM to flush out most impurities, followed by 8-10% MeOH to isolate the desired product, which is the last purple fraction eluted. Fractions containing the desired product were identified by TLC and rotovapped to dryness. The solid residue was suspended in water, filtered, and washed with water and cold methanol to obtain a dark purple powder (4.98 g, 83.7 %). 1

Synthesis of tetrakis-3-(N-methyl-4-pyridinium)phenylporphyrin.
In a Schlenk flask, tetrakis-3-(4pyridyl)phenyl)porphyrin (0.100 g, 0.108 mmol) was dissolved in 10 mL of dry DMF under a nitrogen atmosphere. Methyl iodide (0.113 g, 0.05 mL, 0.800 mmol) was added and the reaction mixture was heated under reflux for 16 h. After cooling, the solvent was evaporated under vacuum and the resulting residue was dissolved in water and precipitated out by addition of a large excess (0.5 g, 3.07 mmol) of ammonium hexafluorophosphate. The resulting precipitate was collected by filtration, washed with water, and dried under air to obtain a purple-brown powder (0.151 g, 89.2%). 1

Synthesis of [N-(4-phenylbenzyl)-4-phenylpyridinium]PF6.
In a 50 mL round bottom flask, 4phenylpyridine (244 mg, 1.57 mmol) and 4-phenylbenzyl bromide (776 mg, 3.14 mmol) were combined in 20 mL DMF and stirred at room temperature for 72 hours. The reaction was monitored by TLC until no more 4-phenylpyridine remained in solution. Solvent was removed under vacuum and the solid was further purified via hot recrystallization from 3 mL of DMF and 2 mL of toluene. The suspension was filtered and rinsed with diethyl ether (3 x 10 mL) and hexanes (3 x 10 mL). A while solid, [N-(4phenylbenzyl)-4-phenylpyridinium]Br was obtained in an 82% yield (519 mg, 1.29 mmol). A salt exchange was performed on [N-(4-phenylbenzyl)-4-phenylpyridinium]Br to exchange the bromide for a hexafluorophosphate ion. In a 20 mL scintillation vial, 80 mg (198.8 μmol) was dissolved in MeOH (5 mL) and deionized water was added (2 mL) to decrease the solubility of the final product. The water/methanol solution was filtered though a 0.2 micron PFTE syringe filter to remove undissolved impurities. Then 100 mg of NH4PF6 was dissolved in 5 mL of deionized water and added to the scintillation vial to precipitate out [N-(4-phenylbenzyl)-4-phenylpyridinium]PF6 as a white powder. The suspension was filtered and dried under vacuum to obtain 88 mg of product (94.6%).%

S7
Synthesis of Cage•4PF6. In a 1 L two-neck round-bottom flask, 500 mL of dry benzonitrile (PhCN) was sparged with N2 for 20 min. A reflux condenser was then attached to one neck of the flask, and the other neck was fitted with a rubber septum that was secured with copper wire. The PhCN was heated under reflux (ca. 200 °C external temperature) with stirring. Syringe pumps were then used to slowly and simultaneously add 20 mL benzonitrile solutions of 2 (300 mg, 0.232 mmol) and 1 (214.6 mg, 0.232 mmol) over 5 days (0.16 mL/h for each solution). Heating was continued for one more day after completing the addition. The mixture was then allowed to cool to room temperature, and a dark solid was collected by vacuum filtration and washed with dichloromethane, acetone, water, and DMF until the washings were clear. The remaining solid was heated to 100 °C for 10 min in DMF containing excess ammonium hexafluorophosphate. Undissolved solids were removed by filtration and washed with DMF to extract additional product until the filtrate was clear. The filtrates were combined and solvent was removed by rotary evaporation to produce a green residue. The green solid was suspended in water, collected on a filter, and washed with hot water until changing color to purple.  General Procedure for Anion Exchange: In a 20 mL scintillation vial, Cage•4PF6 (10 mg, 4.04 μmol) was dissolved in a minimal amount of acetonitrile (5 mL). 10 mL of a saturated aqueous solution of the sodium salt of the desired anion (NH4BF4 or NaBPh4) was added until a red-brown precipitate formed. The solid was centrifuged, filtered, and washed with deionized water (2 x 125 mL) to remove excess salt (88% for the BF4 − salt, 90% for the BPh4 − salt).

NMR Characterization
Instrumentation: Nuclear magnetic resonance (NMR) spectra were measured using a Bruker AVANCE Neo spectrometer or a Varian VNMR spectrometer, both with a 500 MHz working frequency for 1 H nuclei. 1 H NMR spectra were referenced to the residual proteo signal of the solvent ( 1 H δ 1.94 ppm for CD2HCN in CD3CN; 1 H δ 2.50 ppm for (CD3)(CD2H)SO in DMSO-d6; 1 H δ 7.26 ppm for CHCl3 in CDCl3 ). 13 C{ 1 H} NMR in DMSO-d6 spectra were acquired with 256 scans and a 2 s relaxation delay, 13 C{ 1 H} NMR in CD3CN were acquired with 10000 scans and a 4 s relaxation delay. Figure S1. 1 H NMR spectrum of 1 (298K, CDCl3, 500 MHz). S10 Figure S2. 13 C{ 1 H} NMR spectrum of 1 in DMSO-d6 . Note that resonances for the quaternary carbons on the porphyrin macrocycle were not observed. Figure S3. 1 H NMR spectrum of 2 (298K, CDCl3, 500 MHz). S11 Figure S4. 13 C{ 1 H} NMR spectrum of 2 in DMSO-d6. Note that some resonances for quaternary carbons on the porphyrin macrocycle were not observed.

3b. Diffusion Ordered (DOSY) 1 H NMR Spectroscopy Measurements
Diffusion order spectroscopy measurements were conducted on a Bruker AVANCE Neo spectrometer using a wait time of 0.04 s. Samples were prepared at roughly 6 mM in 0.6 mL DMSO-d6. Raw data was processed using Dynamic Center for TopSpin 4.0.7. The decay of each resonance was monitored to determine the diffusion constant (D) for each proton corresponding to Cage 4+ . The diffusion constants of Cage 4+ and host-guest complexes were determined by averaging all D values for proton resonances. Average diffusion constants were used in conjunction with the Stokes-Einstein equation to calculation effective hydrodynamic radii in solution. Binding competition experiments between C60 and C70 were performed by tracking the 1 H NMR resonances of empty Cage 4+ , C60@Cage 4+ , and C70@Cage 4+ in CD3CN after sonicating a 10:1 mixture of the fullerenes in a CD3CN solution of the host (see Figure S30).

S27
The association of the fulleride anions C60and C60 2in Cage 4+ was measured by titrating DMF solutions of the Cp*2Co + salts of the fullerides into solutions of Cage•4PF6 in CD3CN under an N2 atmosphere. Despite signal broadening caused by the paramagnetic guests, new 1 H NMR resonances for the host-guest complexes were clearly observable as distinct from those of the empty host ( Figures S31  and S32), indicating slow exchange of the guests. The benzylic CH resonances of the host and host-guest complexes were used for determining the ratio of empty vs. occupied host. The total concentration of the fulleride guest in solution was determined by integration of the resonance of the Cp*2Co + countercations (1.7 ppm), and the concentration of free guest was determined by subtracting the concentration of hostguest complex. The complex [C60 2-@Cage]•2PF6 has a reduced solubility in the CD3CN, resulting in some precipitation of this complex during the titration, so a sealed capillary of ferrocene in CD3CN was used as an internal standard to provide accurate quantification of the concentration of the host-guest complex over the course of the titration. Association constants for C60and C60 2in Cage 4+ were determined by integration of the spectra at the mid-point of each titration, at which point the concentrations of free and occupied host are nearly equal.

Mass Spectrometry 4a. ESI-MS characterization of Cage 4+ and Host-Guest Complexes
High-resolution mass spectra were obtained using a Waters Xevo G2-XS QTof mass spectrometer using an ESI source and positive ion detection. Samples of the host-guest complexes in MeCN were introduced by direct infusion (5 µL/min). A standard of Leu-Enkephalin (Waters, Milford, MA) was injected in tandem with the sample for in-process calibration. In general, capillary and sample cone voltages were kept between 2 and 3 V, and 120 and 140 V, respectively, to optimize signal to noise ratio for free Cage 4+ and host-guest species. Trifluoroacetic acid was included in samples as an H + source for acquiring the mass spectra of neutral porphyrins.

Electrochemical Studies
Cyclic voltammetry measurements were recorded using a CH Instruments 600E potentiostat. Dry DMF was used as the solvent, and 0.1 M tetrabutylammonium hexafluorophosphate (TBAPF6) was used as a supporting electrolyte. All measurements were recorded using a 3 mm glassy carbon working electrode, a platinum wire counter electrode, and a silver wire pseudo-reference electrode which was confined in a polypropylene body that provided contact with the analyte solution via a porous zeolite bead or a glass frit. Potentials were referenced to the Fc +/0 couple of an internal ferrocene standard or by calibrating the pseudo-reference electrode vs. the Fc +/0 couple of an external solution of ferrocene immediately before use (note: keeping the pseudo-reference electrode isolated from the bulk sample prevents significant drift of the reference potential over numerous CV experiments). For clarity, we present the CVs that do not include the Fc +/0 wave. Positive feedback iR compensation was applied during all CV experiments. The 2nd cycle out of three is presented unless otherwise noted.       1 mM) in benzonitrile recorded at a scan rate of 1 V/s. Note that the first two fullerene-centered reductions show good reversibility, and the C60 1-/2couple is better defined than was observed using DMF as the solvent.  1 mM) in benzonitrile recorded at a scan rate of 1 V/s. As seen for data collected in DMF, the first fullerene-centered reduction shows good reversibility within a potential window that avoids reduction of Cage 4+ . However, the second fullerene-centered reduction is less well-defined than seen in DMF, even when reduction of Cage 4+ is avoided.   Table S1. Comparison of redox potentials for free fullerene, Cage 4+ , and host-guest complexes in benzonitrile. Encapsulation within Cage 4+ minimally perturbs the redox potentials of both fullerenes. Similarly, the fullerenes do not have a significant impact on the cage-centered reduction events. These small changes in potentials in both DMF and PhCN suggests solvents do not play a significant role in mediating the electrostatic interaction between reduced fullerenes and Cage 4+ .

UV-vis-NIR Spectroscopy
UV-vis-NIR spectra were recorded at room temperature using a Shimadzu UV-2600i spectrophotometer equipped with a an ISR-2600Plus integrating sphere detector that provides a 220 -1400 nm wavelength range for solution absorbance measurements. Samples for UV-vis-NIR spectroscopy were prepared at an analyte concentration of 0.125 mM and measured in 1 mm path length quartz cuvettes.
For experiments examining complexes of fulleride anions bound in Cage 4+ , the fulleride@Cage 4+ complexes were prepared in situ from deoxygenated solutions of fullerene@Cage 4+ by addition of Cp*2Co as a stock solution (160 mM) in C6D6.
For assessing the ability of different strength reducing agents to reduce encapsulated C60, two viologens, methyl viologen (MV 2+ ) and methoxyphenyl viologen (MPV 2+ ), were treated with Cp2Co to generate the radical-cation states of the viologens, which were then examined as chemical reductants for accessing the first fullerene-centered reduction of C60@Cage 4+ . Stock solutions of each viologen radical cation were prepared via the addition of Cp2Co in benzene (27.8 M) to viologen solutions in DMF (7.6 mM for MPV 2+ , 15.1 mM for MV 2+ ). To avoid over-reduction of the viologens and ensure no unreduced Cp2Co remained in solution, 0.8 equiv of Cp2Co was used for each viologen. Reduced viologen solutions were then added to solutions of C60@Cage 4+ in DMF (0.125mM) in order to introduce 1 equiv of reduced viologen per host-guest complex. Cuvettes sealed under a nitrogen atmosphere were used to perform UV-vis absorbance experiments to monitor the reduction of the fullerene.
To assess the association of C60 2− and C70 2− in the reduced Cage 2− state of the host, 0.50 mM solutions of C60@Cage 4+ and C70@Cage 4+ in MeCN were treated with 8 equiv of Cp*2Co that was added as a stock solution (170 mM) in benzene in an N2 atmosphere glovebox. For both complexes, a precipitate formed immediately upon adding the reductant and the resulting suspension was stirred for 15 min before filtering through a 0.22 micron PTFE filter into a quartz cuvette. The samples were sealed under N2 with a PTFE stopper and UV-vis-NIR spectra were measured immediately after removing the cuvettes from the glovebox. The spectra confirmed complete disappearance of signals corresponding to the host, enabling quantification of the amount of free C60 2− or C70 2− remaining in solution.      Figure S54). Figure S70. UV-vis-NIR spectra of a solution of [C60@Cage]•4PF6 before and after treatment with 1 and 2 equiv of the methoxyphenyl viologen radical cation (MPV +• ), which has a reducing strength of ca. −0.71 V vs. Fc +/0 in DMF as determined by examining the MPV 2+ /MPV +• redox couple by cyclic voltammetry (see Figure S53). Partial reduction of the C60 guest is evident from the NIR absorption band with lmax = 1071 nm, but the intensity of this absorption is much lower than that observed upon reduction of C60@Cage 4+ with Cp*2Co or MV +• (see Figures S58 and S63), confirming that the encapsulated guest has a reduction potential that is not perturbed much from that of unencapsulated C60 (-0.82 V vs. Fc +/0 in DMF).   Figure S73. Overlaid absorbance spectra of solutions of C60@Cage 4+ in DMF after treatment with 1 equiv of the methoxyphenyl viologen radical cation (MPV •+ , E1/2 = -0.71 V vs Fc +/0 ) to establish a small equilibrium concentration of the reduced complex C60 -@Cage 4+ . Spectra were acquired at different concentrations of C60@Cage 4+ and with various concentrations of electrolytes added. The NIR absorption band (see inset) of the C60guest indicates similar ratios of C60 -@Cage 4+ to C60@Cage 4+ were formed in each sample, suggesting that ions in solution have little impact on the reduction potential of the encapsulated guest.

Fluorescence measurements of C 60 and C 70 association in Cage 4+
Scintillation vials were treated to silylate exposed surface OH sites by leaving the vials in an evacuated desiccator with 10 mL of hexamethyldisilazane for at least 30 minutes before use. The vials were then filled with benzonitrile, 25.0 L of a 4.0 x 10 -5 M solution of Cage•4PF6 in benzonitrile, and appropriate amounts of a stock solutions of 4.16 x 10 -5 M C60 in benzonitrile for a total volume of 5 mL per sample, a Cage 4+ concentration of 2.0 x 10 -7 M, and C60 concentrations ranging from 1.0 x 10 -7 M to 9.0 x 10 -6 M. Samples with no C60 were prepared as references. All vials were sealed with parafilm and sonicated in a Bransonic CPX2800H ultrasonic bath for 1 h. The fluorescence emissions of each sample were then promptly measured using a Horiba Aqualog combination spectrophotometer-fluorimeter. Absorbances were also measured to ensure that they did not exceed an absorbance of 0.1. The same procedure was followed for C70, except that lower concentration stock solutions were used, producing final solutions withC70 concentrations ranging from 5.0 x 10 -9 M to 1.5 x 10 -7 M and a calculated Cage 4+ concentration of 3.98 x 10 -8 M. Repeat data points around the inflection point were taken to help define the shape of the curve.

S53
The concentration of host-guest complex in each sample was calculated according to Eq S1, where [H]0 is the total host (Cage 4+ ) concentration, I is the measured reduction of fluorescence intensity, and I0 is the fluorescence intensity of a sample of the host with no guest (fullerene) present. Since small amounts of Cage 4+ were found to adsorb on the surface of the glass vials used for sample preparation, the value of [H]0 was determined by comparing the fluorescence of the sample of Cage 4+ without added guest vs. the fluorescence intensity of freshly prepared samples of Cage 4+ . The concentrations of host-guest complex were then plotted as the dependent variable against the added guest concentration as the independent variable. OriginLab 2020 was used for baseline correction and fitting the resulting plots to the equations outlined by Thordarson 10 (Eq S2)

DFT Calculations
All density functional theory calculations were performed using Gaussian 16 revision A.03. The b3lyp functional was employed in all calculations along with the SMD solvation model of Truhlar and Cramer. 11 For optimizing the geometry of Cage 4+ , the 6-31+G(d,p) basis set was employed for all atoms (C,N,H), and the structure was optimized from an initial guess based on reasonable bond lengths and angles for the porphyrin units and benzylpyridinium linkers. The optimized geometry of Cage 4+ was used along with reported structures of C60 and C70 to construct starting points for optimizations of the host-guest complexes C60@Cage 4+ and C70@Cage 4+ . The host-guest complexes were optimized using the 3-21+G basis set because calculations could not be completed in a reasonable time (≤30 days) using a larger basis set. Figure S76. Depictions of the LUMO to LUMO+7 orbitals of Cage 4+ determined via DFT calculations. Orbital images were generated using GaussView at a surface isovalue of 0.02.