Pillar[5]arene-Decorated Single-Walled Carbon Nanotubes.

Control of single-walled carbon nanotube dispersion properties is of substantial interest to the scientific community. In this work, we sought to investigate the effect of a macrocycle, pillar[5]arene, on the dispersion properties of a polymer-nanotube complex. Pillar[5]arenes are a class of electron-rich macrocyclic hosts capable of forming inclusion complexes with electron-poor guests, such as alkyl nitriles. A hydroxyl-functionalized pillar[5]arene derivative was coupled to the alkyl bromide side chains of a polyfluorene, which was then used to coat the surface of single-walled carbon nanotubes. Noncovalent functionalization of carbon nanotubes with the macrocycle-containing conjugated polymer significantly enhanced nanotube solubility, resulting in dark and concentrated nanotube dispersions (600 μg mL-1), as evidenced by UV-vis-NIR spectroscopy and thermogravimetric analysis. Differentiation of semiconducting and metallic single-walled carbon nanotube species was analyzed by a combination of UV-vis-NIR, Raman, and fluorescence spectroscopy. Raman spectroscopy confirmed that the concentrated nanotube dispersion produced by the macrocycle-containing polymer was due to well-exfoliated nanotubes, rather than bundle formation. The polymer-nanotube dispersion was investigated using 1H NMR spectroscopy, and it was found that host-guest chemistry between pillar[5]arene and 1,6-dicyanohexane occurred in the presence of the polymer-nanotube complex. Utilizing the host-guest capability of pillar[5]arene, the polymer-nanotube complex was incorporated into a supramolecular organogel.

Introduction

Since their discovery in 1991,[1] single-walled carbon nanotubes (SWNTs) have attracted significant interest within the scientific community because of their extraordinary structural,[2−4] mechanical,[5−7] and optoelectronic properties.[8−10] However, intertube π–π interactions result in the formation of insoluble SWNT bundles,[11] making nanotube processing challenging. In addition, all commercial SWNT production processes result in a complex mixture of semiconducting and metallic SWNTs, which are contaminated with amorphous carbon and metal catalyst particles.[12−16] To improve processability and purity, selective covalent and noncovalent functionalization methods have been investigated.[17−19] Covalent modification utilizes reactive intermediates that form bonds with the sp2-hybridized carbon framework of the SWNT sidewall, resulting in sp3-carbon defects that deteriorate their intrinsic properties.[17] Conversely, noncovalent functionalization utilizes a dispersant that adsorbs onto the SWNT surface, forming a supramolecular complex that does not disrupt the extended π-system.[20,21] A variety of dispersants can be employed, including surfactants,[22−24] aromatic small molecules,[25−27] conjugated polymers,[28,29] and biomacromolecules.[30−36] Of these, conjugated polymers have attracted significant attention as it is synthetically feasible to introduce a broad range of structural variation, which has allowed demonstration of their ability to selectively disperse specific SWNT subtypes.[36] To date, most structural modifications have aimed to achieve enriched dispersions of semiconducting SWNTs (sc-SWNTs), with recent efforts aimed toward developing polymer backbone structures capable of conformational changes or depolymerization in response to a stimulus.[37−41] In addition, stimulus-responsive pyrene derivatives have been utilized for the noncovalent modification of SWNTs to control dispersion properties. For instance, Feng and co-workers developed an intriguing CO2-responsive[42] and light-switchable dispersant–SWNT complex,[43] which was shown to reversibly alter the dispersibility of SWNTs.

Recently, we have prepared polyfluorene derivatives possessing azide groups in their side chains, and then used these derivatives to noncovalently functionalize SWNTs.[44−46] We have demonstrated that it is possible to decorate the resulting polymer–SWNT complex using the strain-promoted azide–alkyne cycloaddition[46] or the copper-mediated azide–alkyne cycloaddition.[45] These results indicate that conjugated polymer side chains can be tuned to impart interesting characteristics to the polymer–SWNT complex. Here, we utilize the same design principle to produce polymer–SWNT complexes that contain macrocycles in the polymer side chains to influence the dispersion properties of the resulting polymer–SWNT complex.

Macrocycles have received intense interest in the literature,[47−52] and encompass several key scaffolds that include cyclodextrins,[53−55] calixarenes,[56−58] cucurbiturils,[59−61] and pillararenes.[62−64] Owing to the seminal work by Ogoshi and co-workers,[65] pillar[5]arenes have attracted significant attention because of their facile one-step synthesis and their unique “pillar-like” structure, composed of hydroquinone units connected by methylene bridges at the para-positions. These pillar[5]arenes can be further derivatized to install a single functional handle that can be used in subsequent derivatization. The electron-rich dialkoxybenzene moieties of pillar[5]arenes form stable inclusion complexes with various electron-poor molecules, including viologens,[66−68] alkanediamines,[69,70] and alkylnitriles.[71−73] Host–guest interactions with pillar[5]arenes have been used in polymers,[74−76] sensors,[70,77,78] and organogels.[66,79,80] In this study, we sought to explore the incorporation of pillar[5]arenes onto the surface of a polymer–SWNT complex. We characterize the dispersion properties before and after macrocycle incorporation using UV–vis–NIR, Raman, and fluorescence spectroscopy, and we determine the SWNT concentration in solution using thermogravimetric analysis (TGA). We then incorporate the macrocycle-containing polymer–SWNT dispersion into an organogel using a bis(alkylnitrile)-functionalized poly(ethylene glycol) (PEG) polymer as a cross-linking agent.

Results and Discussion

To prepare the macrocycle-containing polymer–SWNT complex, we first synthesized a polyfluorene derivative containing alkyl bromides (PF-Br) according to literature procedures (see the Supporting Information for details).[81] Gel permeation chromatography (GPC) analysis of PF-Br revealed that the polymer had an average molecular weight (Mn) of 31 kDa and a dispersity (Đ) of 2.02. Separately, the ethoxy derivative of pillar[5]arene was prepared via condensation of 1,4-diethoxybenzene with paraformaldehyde in the presence of BF3·OEt2 (Scheme S3). The resulting structure was then mono-de-ethylated using BBr3, resulting in the monohydroxylated pillar[5]arene 6. This moiety was introduced into the conjugated polymer scaffold by alkylating PF-Br with 6 via phase transfer alkylation, affording the pillar[5]arene-containing polyfluorene, PF-Pillar (Scheme ). As shown in Figure , the 1H NMR signal at 3.30 ppm in PF-Br, which corresponds to the methylene group adjacent to the alkyl bromide, shifts to 3.76 ppm in PF-Pillar. This downfield shift of the resonance upon alkylation is indicative of the quantitative conversion from the alkyl bromide to the alkyl ether product.

Scheme 1

Postpolymerization Functionalization of PF-Br with 6 to Afford PF-Pillar

Figure 1

1H NMR spectra overlay of PF-Br (blue) and PF-Pillar (orange) in CDCl3.

With our polymers in hand, polymer–SWNT dispersions were prepared with raw HiPCO SWNTs following modified literature procedures.[82] Briefly, HiPCO SWNTs (5 mg) were added to a solution of polymer (10 mg of PF-Br or PF-Pillar) in tetrahydrofuran (THF) (10 mL). The mixture was then sonicated for 2 h in a bath sonicator chilled with ice. The resultant black suspension was centrifuged at 8346g for 30 min, and the supernatant was carefully isolated to obtain the polymer–SWNT dispersion, which was stable on the benchtop for at least several months.

To characterize the polymer–SWNT dispersion, we first performed UV–vis–NIR spectroscopy. The absorption features within the spectral range can be grouped into three categories: two semiconducting regions, S11 (830–1600 nm) and S22 (600–800 nm), and a metallic region, M11 (440–645 nm).[83] As shown in Figure , the as-produced PF-Br–SWNT dispersion possesses sharp absorption features that reach a maximum intensity of ∼0.9. The absence of a broad exponential background suggests at least some degree of sc-SWNT species enrichment. In contrast, the as-produced PF-Pillar–SWNT dispersion was highly concentrated and exhibited absorptions in the metallic as well as the semiconducting regions (Figure ). The as-produced PF-Pillar–SWNT dispersion was actually too concentrated for measurement, and required a 5-fold dilution to obtain the absorption spectrum shown in Figure . We hypothesize that steric bulk imparted by the pillar[5]arene side chains allows for effective steric stabilization of the colloidal polymer–SWNT dispersion.

Figure 2

UV–vis–NIR absorption spectra of polymer–SWNT dispersions (2:1 polymer/SWNT mass ratio) in THF for PF-Pillar–SWNT (orange) and PF-Br–SWNT (blue). The absorption spectrum of PF-Pillar–SWNT was diluted 5-fold in THF.

To further probe the polymer–SWNT dispersions, we employed TGA to calculate SWNT concentration. Samples for TGA were prepared by filtering a known aliquot of polymer–SWNT dispersion (0.5–6 mL) through a Teflon membrane with 0.2 μm pore diameter and then washing with THF until the filtrate did not fluoresce at 365 nm. The samples were then transferred to the TGA crucible to determine the recovered mass of polymer–SWNT complex. As shown in Figure S1, samples were heated to 500 °C under an argon atmosphere at a rate of 15 °C min–1. Polymer-only samples were also recorded under identical experimental conditions, and the mass losses, which correspond to polymer side chain degradation, could be used to calculate the SWNT mass fraction (fSWNT) (see the Supporting Information for calculations). Given a known volume of polymer–SWNT dispersion (Vpolymer–SWNT) containing a known mass of polymer–SWNT complex (mpolymer–SWNT), the SWNT concentration (cSWNT) could be calculated asThe relevant data are tabulated in Table S1, and cSWNT for the PF-Br–SWNT and PF-Pillar–SWNT dispersions were determined to be 20 and 600 μg mL–1, respectively. These results demonstrate that the introduction of macrocyclic structures onto the polymer side chains is effective at producing unusually concentrated SWNT dispersions using small amounts of the polymer (∼30-fold increase in SWNT concentration using a 2:1 polymer/SWNT mass ratio). In comparison, a number of reports describe polymer/SWNT mass ratios in excess of 50:1, only to produce relatively dilute dispersions.[84,85]

We next performed Raman spectroscopy to investigate the differences in the SWNT populations dispersed by PF-Br and PF-Pillar. The samples were prepared by drop-casting the dispersions onto a silicon wafer and evaporating the solvent at room temperature. A reference sample containing raw HiPCO SWNTs was prepared by sonicating the raw SWNTs in CHCl3 and then drop-casting the suspension onto a silicon wafer. Raman scans were obtained using excitation wavelengths at 514, 633, and 785 nm, as it has been shown that these excitation wavelengths are sufficient for the characterization of the electronic properties of HiPCO SWNTs.[86]Figure shows the radial breathing mode (RBM) of the Raman spectra at the three excitation wavelengths (full Raman spectra are provided in Figure S2). The spectra were normalized to the G band at ∼1590 cm–1 and offset for clarity. In the RBM region at 514 nm, predominantly m-SWNT features (225–290 cm–1) are observed.[87] The PF-Pillar–SWNT sample exhibits peaks in this region, while the PF-Br–SWNT sample does not (Figure a). This suggests that, under identical dispersion preparation conditions, PF-Pillar disperses m-SWNTs, while PF-Br does not. This result is further corroborated by G-band analysis (the inset of Figure a). The G band consists of two peaks: a lower-frequency G– and a higher-frequency G+. For sc-SWNTs, both the G– and G+ have Lorentzian line shapes, while for m-SWNTs, the G– exhibits a broader Breit–Wigner–Fano (BWF) line shape.[88] A broad G– band is observed for the PF-Pillar–SWNT sample, which is consistent with the presence of m-SWNTs. Meanwhile, the PF-Br–SWNT sample lacks the BWF line shape in the G band, confirming that m-SWNTs are absent. At 633 nm, both m-SWNTs (175–230 cm–1) and sc-SWNTs (230–300 cm–1) are in resonance.[88,89] As shown in Figure b, the PF-Pillar–SWNT sample exhibits peaks corresponding to both sc- and m-SWNTs. However, only sc-SWNTs peaks are present in the PF-Br–SWNT sample. Finally, at 785 nm, sc-SWNTs are primarily in resonance.[87,90] A few large diameter metallic species, most notably the (16,7) and (12,9) species, are also observed in the low-frequency region. Figure c shows that sc-SWNTs are present in both the PF-Br–SWNT and PF-Pillar–SWNT samples. Again, this demonstrates that PF-Br disperses only sc-SWNTs, while PF-Pillar disperses both sc- and m-SWNTs. We hypothesize that the additional steric bulk of the pillar[5]arene macrocycle in the polyfluorene side chains improves the colloidal stability of the polymer–SWNT complex, resulting in more concentrated and less-selective polymer–SWNT dispersions. Beyond sc-SWNT peaks at 785 nm, a prominent peak at 265 cm–1 arises from the (10,2) sc-SWNT species when trapped in a SWNT bundle. Assuming (10,2) is present, this “bundling peak” can be used to identify the degree of bundling in a nanotube sample.[82] It was found that the bundling peaks for both polymer–SWNT samples were substantially suppressed compared to that for the raw SWNT sample. Overall, the Raman data demonstrates that the concentrated PF-Pillar–SWNT sample is a function of unusual colloidal stability, rather than due to the suspension of SWNT bundles.

Figure 3

Raman spectra showing RBM regions collected using (a) 514 nm, (b) 633 nm, and (c) 785 nm excitation wavelengths. The tray boxes denote the signals arising from sc-SWNTs, while the pink boxes indicate the locations of signals from m-SWNTs. The inset in (a) shows the G-band region, located at ∼1590 cm–1, upon excitation at 514 nm.

To further characterize the polymer–SWNT dispersions, photoluminescence (PL) maps were obtained. The polymer–SWNT dispersions were diluted in THF to obtain an absorption intensity of ∼0.11 for the peak centered at ∼1279 nm (Figure S3). The excitation and emission energies of various sc-SWNTs were obtained from experimental Kataura plots and plotted on the PL map.[91] As shown in Figure a, high-intensity PL signals were observed in the PF-Br–SWNT dispersion, with the most-intense peak corresponding to the (7,6) sc-SWNT species. Other prominent species include (8,7), (8,6), and (7,5). For the PF-Pillar–SWNT dispersion, the (7,6) species is also the most-intense species. Other prominent species include (9,4), (7,5), (6,5), and (8,4). Interestingly, the relative fluorescence intensity of the PF-Pillar–SWNT sample is roughly 2 orders of magnitude lower compared to that of the PF-Br–SWNT sample. This observation may be attributed to either SWNT bundles or m-SWNT species present in the sample, as both are known fluorescence quenchers.[9] As Raman analysis indicates the lack of significant bundling in the polymer–SWNT dispersions, we attribute the observed fluorescence quenching to an increased amount of m-SWNTs dispersed by PF-Pillar, which is consistent with the aforementioned analyses.

Figure 4

Photoluminescence maps of (a) PF-Br–SWNT and (b) PF-Pillar–SWNT, concentration-matched by UV–vis–NIR and plotted on the same scale.

With our polymer–SWNT dispersions fully characterized, we sought to explore the dispersion properties of the PF-Pillar–SWNT sample using 1H NMR spectroscopy. We employed 1H NMR to investigate the host–guest interactions in the presence of our polymer–SWNT sample. We first examined the host–guest interactions between PF-Pillar and 1,6-dicyanohexane in THF-d8 (Figure S4). The resonance at 6.84 ppm, which corresponds to the aromatic protons in pillar[5]arene, was shifted downfield by 0.07 ppm upon the addition of 1,6-dicyanohexane. This downfield shift is consistent with the association between 1,6-dicyanohexane and a pillar[5]arene-functionalized conjugated polymer.[92] To investigate the properties of the PF-Pillar–SWNT complex, we prepared a PF-Pillar–SWNT dispersion in THF-d8 (0.75 mL), following the previously outlined protocol (vide supra), and added hexamethyldisilane (0.5 μL) as an internal standard. NMR spectra were normalized to this internal standard (for full 1H NMR spectra, see Figure S5). Compared to that of PF-Pillar, the 1H NMR spectrum of the PF-Pillar–SWNT dispersion in THF-d8 shows broad signals that correspond to the fluorene backbone as well as the pillar[5]arene moiety (Figure ). Upon addition of 1,6-dicyanohexane, the resonance corresponding to the aromatic pillar[5]arene protons shifts from 6.83 to 6.91 ppm, with slight broadening. Thus, it is apparent that the host–guest chemistry between the pillar[5]arene moiety and 1,6-dicyanohexane occurs in the presence of the polymer–SWNT complex.

Figure 5

1H NMR spectra (THF-d8, 298 K) of the PF-Pillar–SWNT dispersion recorded after successive additions of 1,6-dicyanohexane (0–65 equiv).

Having confirmed the formation of inclusion complexes in the presence of the PF-Pillar–SWNT dispersion, we sought to incorporate the concentrated dispersion into supramolecular organogels. To prepare these organogels, we first synthesized a homobifunctional poly(ethylene glycol) (PEG) polymer with terminal alkyl nitriles (PEG600-(CN)2) to act as a cross-linker, according to literature procedures.[92]Figure depicts an idealized cartoon representation of the supramolecular gel. We envisioned a system where the pillar[5]arene units interact with the alkyl nitrile groups in the PEG600-(CN)2 cross-linker to form a uniform cross-linked network. Using this concept, we prepared 40 wt % organogels in 1,2-dichlorobenzene by adding PEG600-(CN)2 cross-linker to the PF-Pillar–SWNT dispersion (see the Supporting Information for details). After 30 min of incubation under ambient conditions, both the PF-Pillar–PEG600-(CN)2 (native) and PF-Pillar–SWNT–PEG600-(CN)2 (hybrid) mixtures formed gels that did not flow when the microcentrifuge tube was inverted, as shown in Figure . The resultant gels were soft and tacky in texture, with a marked difference in color between the native (yellow) and hybrid (black) gels. In a control experiment, PF-Pillar and PEG600 were mixed in 1,2-dichlorobenzene under conditions identical to those used to form the aforementioned organogels (see the Supporting Information for details). Upon inversion, this control mixture remained liquid and flowed down the wall of the container (Figure ). Mechanical testing of the organogels was conducted using a home-built apparatus that measures the contact mechanics between a glass hemisphere and the organogel samples (see the Supporting Information for details). Briefly, the hemispherical indenter was attached to a force transducer to measure the force applied (F) as a function of indenter vertical position (d). Young’s modulus was calculated using Hertzian theory, where Young’s modulus was obtained as the slope when plotting F versus d. We determined that Young’s moduli of PF-Pillar and PF-Pillar–SWNT gels were 6.0 ± 0.2 and 5.8 ± 1.3 kPa, respectively. We hypothesize that the wt % of SWNT in the PF-Pillar–SWNT organogel may be too low to observe substantial differences in mechanical properties. Given the limitation of our dispersion concentration, we could not further concentrate the PF-Pillar–SWNT dispersion to improve the SWNT loading. Overall, we demonstrate that supramolecular organogels incorporating PF-Pillar-coated SWNTs can be produced, and that gelation is unaffected by the presence of SWNTs.

Figure 6

Idealized cartoon representation of host–guest-driven gelation of PF-Pillar–SWNT (red) and PEG600-(CN)2 (green). The polyfluorene backbone (represented with a blue ribbon) is depicted to wrap around SWNTs in a helical fashion. However, the polymer backbone may also irregularly coat the SWNT surface.

Figure 7

Photographs of supramolecular organogels made from the host–guest interaction between PF-Pillar and PEG600-(CN)2 (native) and in the presence of SWNTs (hybrid).

Conclusions

Here, we demonstrate that a macrocycle-containing polyfluorene derivative can produce unusually concentrated polymer–SWNT dispersions. The quantitative alkylation of polyfluorene with mono-de-ethylated pillar[5]arene was confirmed by 1H NMR spectroscopy. We show that the facile modification of a polyfluorene backbone with a pillar[5]arene macrocycle has a significant impact on the polymer–SWNT dispersion properties. As evidenced by UV–vis–NIR spectroscopy, the dispersion postfunctionalization with a macrocycle was substantially more concentrated than the corresponding dispersion without the macrocycle. It was found that the polyfluorene derivative improved the concentration of SWNTs ∼30-fold, from 20 to 600 μg mL–1, as determined by TGA. Using Raman and fluorescence spectroscopy, it was determined that the initial polyfluorene derivative preferentially dispersed sc-SWNT species, while the pillar[5]arene-decorated polyfluorene produced dispersions containing both sc- and m-SWNTs. The formation of an inclusion complex between pillar[5]arene and 1,6-dicyanohexane was confirmed in the presence of the macrocycle-containing polymer–SWNT complex. This polymer–SWNT dispersion was then successfully incorporated into supramolecular organogels.

Experimental Section

All reagents were obtained from commercial sources and were used as received without further purification. Raw HiPCO SWNTs were purchased from NanoIntegris (batch #HR27-104, 10 wt % in anhydrous EtOH) and used without further purification. Flash chromatography was performed using an IntelliFlash 280 system from Analogix. Unless otherwise noted, compounds were monitored using a variable wavelength detector at 254 nm. Solvent amounts used for gradient or isocratic elution were reported in column volumes. Columns were prepared in Biotage SNAP KP-Sil cartridges using 40–63 μm silica or 25–40 μm silica purchased from Silicycle. 1H NMR spectra of small molecules and polymers were recorded on Bruker Avance 600 and 700 MHz spectrometers, respectively. Polymer molecular weights and dispersities were analyzed (relative to polystyrene standards) via GPC using a Waters 2695 Separations Module equipped with a Waters 2414 refractive index detector and a Jordi Fluorinated DVB mixed bed column in series with a Jordi Fluorinated DVB 105 Å pore size column. THF with 2% acetonitrile was used as the eluent at a flow rate of 2.0 mL min–1. Sonication was performed in a Branson Ultrasonic B2800 bath sonicator. Centrifugation of the polymer–SWNT samples was performed using a Beckman Coulter Allegra X-22 centrifuge. UV–vis–NIR spectra were recorded on a Cary 5000 spectrometer in dual beam mode, using matching 10 mm quartz cuvettes. Thermogravimetric analysis was performed on a Mettler Toledo TGA/DSC 3+, and all measurements were conducted under an argon atmosphere, with sample masses ranging from 0.5 to 1.0 mg. Raman spectra were recorded using a Renishaw InVia Laser Raman spectrometer, with three different lasers: a 25 mW argon ion laser (514 nm, 1800 L mm–1 grating), a 500 mW HeNe Renishaw laser (633 nm, 1800 L mm–1 grating), and a 300 mW Renishaw laser (785 nm, 1200 L mm–1 grating). For the raw SWNT sample dispersed in CHCl3, laser intensity was set to 1% for 514 and 633 nm, and 10% for 785 nm. For the polymer–SWNT samples, laser intensity was set to 1% for all excitation wavelengths. Fluorescence spectra were recorded on a Jobin-Yvon SPEX Fluorolog 3.22 equipped with a 450 W Xe arc lamp, a digital photon counting photomultiplier, and an InGaAs detector, also using a 10 mm quartz cuvette. Slit widths for both excitation and emission were set to 10 nm band-pass, and correction factor files were applied to account for instrument variations. Photoluminescence maps were obtained at 25 °C, with 5 nm intervals for both the excitation and the emission.