We have focused on an angular tetraamine and a linear dialdehyde to construct the covalent organic nanotubular architecture (Fig. 1). The tetraaminotriptycene (TAT) possesses its opposite terminal pairs at a dihedral angle of ~120º, which mandates the construction of one-dimensional (1D) CONTs (Fig. 1a). The precise orientation of amine functionalities in the TAT units enables the propagation of covalent linkages in one dimension. Our model structure (Fig. 1c) indicates that the two geminal imine bonds connected with a triptycene unit are trans to each other. TAT units' orientation ensures the framework's tubular propagation during the condensation, keeping the non-functionalized benzene ring towards the tube wall. To check the reaction's feasibility, we have first synthesized a monomeric unit by reacting the TAT with 2-methoxy benzaldehyde. High-resolution mass spectrometry (HRMS) analysis indicates that the stoichiometric condensation of TAT and 2-methoxy benzaldehyde results in a mixture of three products. These products are monomer-1; with two imidazole rings, monomer-2; with one imidazole ring and two imine bonds, and monomer-3; with four imine bonds (Supplementary Section 2). By carefully examining the monomers, we concluded that imidazole formation is the competitive reaction preventing the nanotube formation. Thus, to suppress the imidazole formation, we have used argon atmosphere, room temperature condition, high dilution, and dropwise addition of TAT during the CONT synthesis.
We have synthesized two CONTs (CONT-1 & CONT-2) via imine condensation reactions by combining a mixture of one equivalent of TAT (15.7 mg, 0.05 mmol) and two equivalents of 2,5-dimethoxybenzene-1,4-dicarboxaldehyde (DMDA) (19.4 mg, 0.1 mmol) for CONT-1 and two equivalents of terephthalaldehyde (TA) (13.4 mg, 0.1 mmol) for CONT-2. The dropwise addition of amine solution in dichloromethane (DCM) into the aldehyde solution (in DCM) in presence of 0.5 mL 6 (M) acetic acid results in the cloudy precipitate (Supplementary Section 1). The resulting precipitate was collected by centrifugation followed by washing with anhydrous THF and then evacuated at 120 °C for 12 hours to yield 25.05 mg of CONT-1 and 13.3 mg of CONT-2 [78%, and 51% yield of CONT-1 and CONT-2 respectively based on TAT] as off-white solid. The FTIR spectra of both CONTs show the characteristic peaks at 1610 cm-1, which are characteristic -C=N- stretching modes for imine bonds (Fig. 2a, Supplementary Figure 3). Solid-state Cross-Polarization Magic angle Spinning (CPMAS) NMR spectroscopy of the 13C and 15N nuclei was performed to validate the formation, connectivity, and atomic level construction of the CONT-1 backbone (Fig. 2b, Supplementary Figure 4).
Solid-state 13C CP-MAS NMR spectroscopy shows the characteristic peaks of the imine (−C=N−) bonded carbon atoms at 155.5 ppm, whereas the methyl carbon appears at 53.5 ppm. The NMR spectrum also displayed discrete resonances in the aromatic region between 150 and 110.5 ppm (Fig. 2b). Solid-state High-Power Decoupled (HPDEC) 13C spectrum quantifies the number of different carbon atoms in the tube backbone. According to the reaction scheme, the basic unit consists of 20 carbons from TAT and 40 carbons from DMDA. Out of these, 18 carbons from TAT and 32 carbons from DMDA appear in the aromatic and carbonyl region (200 to 100 ppm), two carbons of TAT and eight carbons of DMDA appear in the aliphatic region (~55 ppm) (Fig. 2c). Thus, one would expect an aromatic to the aliphatic carbon ratio of 5:1 for a complete reaction. The ratio obtained from the 13C HPDEC spectrum is close to 5:1. The peak at 186 ppm could be assigned as free aldehyde functionality generated from the defect sites at the nanotube surface (Supplementary Figure 4). The 15N CP-MAS spectrum of CONT-1 displayed two discrete peaks, a resonance at 240 ppm and another at 148 ppm, indicating two distinct nitrogen sites. The characteristic peak at 240 ppm suggests the formation of the imine (−C=N−) bonded nitrogen atoms. The resonance at 148 ppm indicates the presence of N-H group, which could be due to imidazole ring formation at the defect sites (Supplementary Section 17). Thermogravimetric analysis (TGA) of the activated CONT-1 under N2 atmosphere indicates that the framework has thermal stability up to 400º C, and there is no guest molecule inside the nanotubes (Supplementary Figure 25). The porosity of the CONTs was evaluated by measuring N2 adsorption isotherm at 77K (Fig. 2d, Supplementary Figure 5). Activated CONTs at 140 ºC showed reversible type II adsorption isotherm. The BET surface areas of the activated CONT-1 and -2 were 321 m2g-1 and 52 m2g-1, respectively (Supplementary Figure 5). The pore size distribution of both the CONTs was calculated based on the nonlocal density functional theory (NLDFT) (Fig. 2d and Supplementary Figure 6). The pore size distribution, which indicates two types of pores of 1-2 nm and 3.5-4.5 nm of size, is in good agreement with the theoretically predicted structure (Supplementary Figure 6). The 3.5-4.5 nm pore distribution refers to the main hollow tubular channels running along the length of the CONTs. The 1-2 nm pore distribution corresponds to the side pores located on the nanotubes' walls.
Scanning Electron Microscopy (SEM) images of CONTs reveal uniform tubular morphology, with an average diameter of ~5 nm (Supplementary Figure 7). High-Resolution Transmission Electron microscopy (HRTEM) (Fig. 2e) identifies the hollow tubular nature at the interior with a constant diameter of ~5 nm throughout the entire length of the nanotube (Supplementary Figure 8). Atomic Force Microscopy (AFM) further corroborates this finding (Supplementary Figure 9). The AFM height profile shows the uniform diameter of ~5 nm of the single nanotube, which is in good conjunction with the TEM results and nanotubes' theoretically predicted structure (Fig. 2f). Additionally, electronic microscopic images show that the single-walled CONTs are intertwined, which could be due to their high length to width ratio (~avg. 300:1).
The microscopic analysis further confirms that the individual units of the intertwined nanotubes' diameter having a close match (~5 nm) with isolated single nanotubes (Supplementary Figure 13).
We have isolated the reaction mixtures at different time intervals and characterized them via SEM analysis to shed more light on the CONT-1 formation (Supplementary Figure 10). CONT-1 formation starts within 6 hours of reaction. These CONTs are 100-200 nm in length and ~5 nm in diameter. After 12 hours, the CONTs grow up to 500-800 nm in length with a diameter of ~5 nm (avg. length to diameter ratio 130:1). The intertwining starts after 24-30 hours when the length to diameter ratio of the CONTs increases significantly (~avg. 200:1).
The rapid increase in the length of CONTs induces high flexibility, promoting intertwined structures. After 36 hours of reaction, almost all CONTs become completely intertwined, and no significant morphological changes were observed (Supplementary Figure 11). We speculate that the defect centers could trigger this intertwining (Fig. 3). After 6 hours, two nanotubes uniformly intertwine, following a particular pattern where the twining pitch is 70 (±10) nm (Supplementary Figures 13, 14). The resulting intertwined nanotube thread again entangles with the available mesh of CONTs. The width of intertwined nanotubes reaches up to ~100 nm (maximum) with time (Supplementary Figures 11, 12, and 15). However, the intertwining pitch remains constant (~70 nm) irrespective of their size and diameter.
To our surprise, the nanotubes retain their morphology in a broad range of solvents of various polarities (Figs. 4a, b). However, the width of intertwined CONT-1 varies with the solvents' nature. The N2 adsorption isotherm confirms the structural stability in water after seven days (Fig. 4d). We drop-casted the well-dispersed CONTs on a silicon wafer and heated it at different temperatures to validate the material's temperature stability. The morphology remains unaltered even at 150 ºC (Fig. 4c).
To obtain a molecular-level insight into the structure and interactions of CONTs, we used multiscale molecular models. First, we simulated the behavior of the CONTs in different conditions.16,17 We developed all-atom (AA) models of CONT-1 composed of 15 TAT layers (Fig. 5a). This AA model was immersed in explicit DCM or THF solvent molecules (Fig. 4b) and was then equilibrated via 200 ns of all-atom molecular dynamics (AA-MD) at 20 °C of temperature (Supplementary Section 24). For comparison, we also equilibrated via the AA-MD model of the CONT-1 in water and the absence of solvent (Fig. 5b). The simulations in DCM and THF showed that the CONT-1 equilibrates to configurations just slightly deviating from the initial perfect one (Fig. 5b). This can be inferred by the distributions of the characteristic angles 1 and 2 (Fig. 5a: the angles formed by neigh-boring TAT units), which equilibrate at values similar to those of the starting conformation (Fig. 5c: green and orange distributions vs. black dotted lines). On the contrary, in water and the absence of solvent, the tubules adopt a much more distorted configuration, as the solvophobic effects are so strong that the CONT-1 structure is compressed (Fig. 5b and Fig. 5c). We then used these AA models as a guideline to develop a minimalistic Coarse-Grained (CG) model18 that, although more approximated, allowed us to study the behavior and interaction between the CONTs on a much higher scale. In particular, the CG model is composed of one CG particle per TAT unit in the CONT-1 structure, which are interconnected with each other via explicit bonds (Fig. 5d).
We parametrized this CG model consistently with the behavior observed through the AA-MD simulations (Supplementary Section 24). Such CG model allowed us to simulate long CONT-1 composed of 500 TAT hexagonal layers, corresponding to a tubule length of ~820 nm (Fig. 5c). Thus, we compared via CG-MD simulations the behavior of two long CONT-1 tubules that interact between them in different conditions. In particular, the non-bond interaction between the CG particles that compose the two CONT models (modeled by a Lennard-Jones, LJ, potential in the CG model) was changed to modulate the nanotube-nanotube interaction. This allowed us to simulate the effect of the solvent's change on the system by increasing/decreasing the solvophobic interactions between the CONTs. In particular, for ɛ (depth of the LJ potential, determining the strength of non-bonding the interaction between the CG particles) values below 1.0 kJ mol-1, the two CONTs interacted only in an unstable way, and no intertwining was observed. Instead, for ɛ≥ 2 kJ mol-1, we could observe stable interaction between the CONTs and inter-tubule wrapping. As indicated by umbrella sampling19 calculations showed (Supplementary Section 24), the CG models with ɛ values of 2 and 2.5 kJ mol-1 provided a CONT-CONT attraction compatible with that of AA models in explicit DCM and THF solvents, respectively (Supplementary Figure 29). In these cases (Figs. 5e-f), the CG-MD showed an average intertwining pitch consistent with that observed experimentally (~70±10 nm). Umbrella sampling results also demonstrated that the CONT-CONT interactions of the AA model in water, or in the absence of solvent, are compatible with much higher ε values in the CG model (stronger solvophobic effects). However, CG-MD simulations showed that at ɛ>2.5 kJ mol-1, the intertwined CONTs are not stable on their own, and the tubules tend to fold multiple times further.
The intertwined CONTs further self-assemble to form a toroidal superstructure upon ultrasonication (Figs. 6a-c).20-22 We have observed that tetrahydrofuran (THF) is the best solvent to obtain the toroids in high yields (up to 60%) (Supplementary Figure 21). Toroidal micro ring formation also proceeds in other solvents like o-xylene and o-dichlorobenzene (DCB), with a yield of ˂5%. The toroids are purified from the mixture of intertwined nanotubes by filtering through a Whatman 42 filter paper (pore size of 2.5 µm) (Supplementary Section 17). The Dynamic Light Scattering (DLS) study of the filtrate provides the average outer diameter of 600 nm at 20 ºC with a polydispersity index less than 0.15 (Fig. 6d). FESEM images show that the toroid diameters ranging from 300 nm to 900 nm (Fig. 6f). However, the rings' thickness remains constant (~50 nm) around their circumference (Supplementary Figure 21).
Topographical analysis using AFM further confirms the same toroidal morphology (Supplementary Figure 23). HRTEM images of toroids prove that the walls of toroids are composed of intertwined hollow nanotubes with a ~5 nm individual tube diameter and a constant pitch of ~70 nm (Supplementary Figure 22). Detailed SEM, TEM, and AFM analyses reveal that the intertwined nanotubes first bend to form non-uniform loops (diameter 100-1000 nm) (Fig. 6f). The most probable mechanism involves the creation of bubbles in THF.23 The bubbles act as the template for hydrophobic CONTs, which eventually orient themselves around such bubbles' circumference. Being bent at the bubble–THF interface, nanotubes form loops when the bubble collapses. The untied intertwined nanotubes coil up in both the transverse and longitudinal directions to create a closed and coiled loop-like structure of various diameters (0.1-1 µm). These spiral loop structures were later transformed into toroidal structures (Fig. 6f).