Report

Finite phenine nanotubes with periodic vacancy defects

See allHide authors and affiliations

Science  11 Jan 2019:
Vol. 363, Issue 6423, pp. 151-155
DOI: 10.1126/science.aau5441

Toward nanotubes with periodic gaps

Carbon nanotubes consist of a continuous array of benzene rings fused along their edges. It is not straightforward to excise regular fragments in a top-down fashion to produce periodic gaps. Sun et al. showcase the beginnings of a bottom-up strategy toward this end. They used borylations and catalytic cross-coupling chemistry to prepare a discrete cylindrical carbon compound composed of 40 benzene rings bonded to one another at the 1, 3, and 5 positions to leave regular void spaces in the walls. Catenation of multiple similar segments could ultimately lead to an extended nanotube with periodic wall defects.

Science, this issue p. 151

Abstract

Discrete graphitic carbon compounds serve as tunable models for the properties of extended macromolecular structures such as nanotubes. Here, we report synthesis and characterization of a cylindrical C304H264 molecule composed of 40 benzene (phenine) units mutually bonded at the 1, 3, and 5 positions. The concise nine-step synthesis featuring successive borylations and couplings proceeded with an average yield for each benzene-benzene bond formation of 91%. The molecular structure of the nanometer-sized cylinder with periodic vacancy defects was confirmed spectroscopically and crystallographically. The nanoporous nature of the compound further enabled inclusion of multiple fullerene guests. Computations suggest that fusing many such cylinders could produce carbon nanotubes with electronic properties modulated by the periodic vacancy defects.

The molecular structure of fullerenes (1) has inspired synthetic approaches toward curved geodesic polyarenes and larger congeners through assembly of hexagons and contorting nonhexagons, such as pentagons (24). The curved carbon networks of fullerenes have also led to the discovery of carbon nanotubes (CNTs), extended geodesic cylinders composed exclusively of hexagonal units (57). Synthesis of smaller discrete cylindrical molecules, particularly those with rigid graphitic walls, are deepening our understanding of CNT features such as chiroptical properties (8). However, discrete carbonaceous cylinders are not necessarily restricted to defect-free graphitic morphologies (9, 10), and their structural diversity could be further enriched by congeners designed with structural defects. We report the design and synthesis of a molecular cylinder composed of multiple benzene rings. By replacing the trigonal sp2-carbon atoms of the CNT with trigonal 1,3,5-trisubstituted benzene (phenine) units, we designed nanometer-sized cylindrical molecules, named phenine nanotube (pNT) molecules, that possess 240 sp2-carbon atoms with six-atom vacancy defects appearing periodically throughout the cylindrical graphitic sheets.

Our previous geodesic phenine designs of a bowl and a saddle mimicked frameworks of geodesic polyarenes and adopted [n]cyclo-meta-phenylenes ([n]CMP) as the starting nonhexagonal omphalos (11, 12). By contrast, the present target required a distinct synthetic strategy. For the design of pNT, we selected a homothetic framework from the hypothetical but long-sought Vögtle belt polyarenes (Fig. 1) (13, 14). Shaping curved networks of the Vögtle belt, however, poses a difficult synthetic problem that has hampered synthetic approaches with trigonal sp2-carbon atoms.

Fig. 1 pNT molecules and Vögtle belt.

The positions (a to h) are labeled for assignments of the 1H NMR resonances in Fig. 2.

Departing from our [n]CMP-omphalos strategy (11, 12), we examined several synthetic strategies and finally succeeded with a concise, nine-step route (Fig. 2A and fig. S1). A two-step transformation of 1,3-dibromobenzene (1) through 2 afforded the terphenyl 3, which possessed linking moieties installed on both ends, and after [3+3] cyclization (15), Ir-catalyzed C–H borylation (16) produced the [6]CMP derivative 5, with boryl linkers at the para positions. Subsequent application of Yamago’s Pt-macrocyclization (17) with our boryl-route modifications (18) gave a macrocyclic precursor (6) comprising a tetrameric [6]CMP array. The final three steps—iododesilylation, Suzuki coupling, and Ni-mediated Yamamoto-type coupling (19)—were performed with 6 to obtain the target (12,12)-pNT molecule (the nomenclature is provided below). The synthesis was completed in nine linear steps from dibromobenzene, and five coupling steps successfully resulted in 52 biaryl bonds that linked 40 benzene rings in a cylinder. Overall, the yield from dibromobenzene was 0.7%, which corresponded to an average yield of 91% per biaryl bond. This synthetic strategy manipulated the common phenine unit adopted in the previous three-dimensional (3D) geodesic phenine frameworks into one dimension (fig. S1) (11, 12, 20). Structural variations of the pNT molecules were preliminarily investigated by using minor byproducts (supplementary materials). In short, as byproducts of the macrocyclic precursor 6, we obtained minor congeners composed of three and five [6]CMP panels. These congeners were subjected to identical synthetic transformations and allowed us to spectroscopically detect narrower and wider cylinders: (9,9)- and (15,15)-pNT molecules (Fig. 1), respectively (supplementary materials). Although our repetitive efforts failed to complete the synthesis with full requisite data, the spectroscopic detection showed that a variety of pNT congeners should become synthetically accessible in the future (21, 22).

Fig. 2 Nine-step synthesis of pNT.

(A) Reagents and conditions. (i) Bis(pinacolato)diboron [(Bpin)2], PdCl2(dppf)•CH2Cl2, KOAc, dimethyl sulfoxide, 80°C, 91%. (ii) 1,3-dibromo-5-trimethylsilylbenzene, Pd(PPh3)4, K2CO3, N,N′-dimethylformamide (DMF)/H2O, 80°C, 66%. (iii) Ni(cod)2, 2,2′-bipyridyl (bpy), 1,5-cyclooctadiene (cod), DMF/toluene, 80°C, 84%. (iv) (Bpin)2, [Ir(cod)OMe]2, 4,4′-di-t-butyl-2,2′-bipyridyl (dtbpy), tetrahydrofuran (THF), 65°C, 92%. (v) PtCl2(cod), CsF, THF, 65°C. (vi) PPh3, toluene, 100°C, 25% (two steps). (vii) ICl, CHCl3, ambient temperature. (viii) 3-t-butyl-5-chlorophenylboronic acid pinacol ester, Pd(PPh3)4, K2CO3, THF/H2O, 30% (two steps). (ix) Ni(cod)2, bpy, cod, DMF/toluene, 80°C, 20%. (B) Spectra of the (12,12)-pNT molecule. (Top) MS spectrum. (Inset) HRMS; pyrene matrix, positive, [M]+ calcd for C304H264 = 3917.0754. (Bottom) 1H NMR spectrum in 1,1,2,2-tetrachloroethane-d2 at 298 K. Positions corresponding to the peak assignments are shown in Fig. 1. The assignment procedures are available in the supplementary materials.

Considering the cylindrical structure of pNT, we can apply CNT vector nomenclatures to describe the carbon arrangement (6, 23). Thus, the sp2-carbon atoms of the pNT molecules are arranged in networks of (9,9)-, (12,12)-, and (15,15)-CNTs with periodic six-atom vacancy defects. The lengths and defects in the cylinders can also be described by other vector indices, such as length index (tf), bond-filling index (Fb), and atom-filling index (Fa) (Fig. 1) (23). The length index of all the pNT molecules is commonly 7.0, which is more than double the preceding largest value of 3.0 (23). The values of Fa (65%) and Fb (57%) assigned for pNT molecules quantitatively confirm their defect-rich nature.

The structure of the (12,12)-pNT molecule was first confirmed by means of spectroscopy. Matrix-assisted laser desorption/ionization (MALDI) mass spectrometry showed a mass/charge ratio (m/z) value of 3917.0693 (Fig. 2), which matched well with the value calculated for the chemical composition of the molecule (C304H264). Other congeners, (9,9)- and (15,15)-pNT molecules, were also detected by means of MALDI mass spectrometry as m/z = 2938.8 (C228H198) and m/z = 4898.1 (C380H330), respectively (fig. S17). Despite their enormous molecular weight, the molecules were amenable to solution-phase analysis with nuclear magnetic resonance (NMR) spectroscopy in 1,1,2,2-tetrachloroethane-d2. The simple 1H spectrum was consistent with D4h point symmetry (Fig. 2B). For example, only a single methyl 1H resonance appeared for all 16 t-Bu groups appended to both ends of the cylinder. This symmetrization arises from elementary D4h symmetry operations (6, 24); eight t-Bu groups on one end of the cylinder are rendered equivalent by the C4-axis/σv-plane along the axial direction of the cylinder, and the t-Bu groups on the “top” and “bottom” ends become symmetric by the orthogonal C2-axes/σh-plane located at the middle of cylinder. These symmetry operations also converged the signals of the 120 aromatic protons into 8 1H resonances (Fig. 2B, a to h), which were fully assigned by a combination of nuclear Overhauser effect spectroscopy and long-range correlation spectroscopy spectra (supplementary materials). Unlike the 3D nanocarbon molecules with the phenine frameworks (bowl/saddle) (11, 12), aggregation behaviors were not observed with the present 1D nanocarbon cylinder.

The cylinder shape of the (12,12)-pNT molecule was unambiguously confirmed by means of crystallography. A single crystal was obtained from chlorobenzene (PhCl)/2-propanol and analyzed with a monochromated x-ray beam at SPring-8 (BL38B1). The crystal structure revealed a nanometer-sized hydrocarbon cylinder composed of 240 sp2-carbon atoms (Fig. 3). The 240 sp2-carbon atoms were connected to form 40 benzene rings, and the larger hexagons of [6]CMP were fused in the homothetic network of the Vögtle belt. The cylinders were packed in the I4/m space group, and the molecular structure had C4h point symmetry with a C4 axis and a σh plane. The C4h point symmetry lacked only C2v operations from the inherent D4h symmetry and showed that the molecules in the crystal have minute structural deformations. Such packing characteristics should benefit further development of materials design via crystal engineering. The length and diameter measured at the sp2-carbon edges were 1.71 and 1.64 nm, respectively (fig. S19), which matched well with the expected geometries of (12,12)-CNT with tf = 7.0 (geometric length Tf = 7.0 × 0.249 = 1.74 nm and geometric diameter dt = (3 × 122)1/2 × 0.249/π = 1.65 nm) (6, 23). With van der Waals radii of t-Bu groups, the overall length and diameter were ~2.57 and 2.54 nm, respectively. Measurement of the π-orbital axis vectors (POAVs), an indicator of structural deformations of the sp2-carbon from planarity (25), revealed small values between 0.0° and 2.9° at the benzene-benzene bonds, with an average value of 1.5° (fig. S20). The average POAV angle was smaller than in preceding macrocycles of benzene, such as [12]cyclo-para-phenylene (average POAV angle = 2.7°), which showed the presence of small structural strains in pNT (12). The POAV and dihedral angles of the pNT molecule were particularly small at the circumferences (fig. S20).

Fig. 3 Crystal structures of (12,12)-pNT.

(A) Side and top views of the molecular structure. Carbon frameworks are highlighted in red, with hydrogen atoms colored in white. (B) Top and side views of the packing structure. Residual electron densities of solvent molecules were removed during the refinement process (fig. S21). (C) Top and side views of the packing structure of Embedded Image. Fullerene C70 molecules were found at four inequivalent locations. Two interior C70 molecules are shown in green and blue, and two interstitial C70 molecules are shown in orange and pink (fig. S23). Multiple PhCl solvent molecules and minor conformational disorders of t-Bu groups were omitted for clarity.

The crystal data further allowed for in-depth analyses of higher-order structures, which revealed a distinct bundle assembly of pNT molecules. Unlike trigonal/hexagonal bundles conceived for CNTs (2628), tetragonal bundles were found in the single crystal of (12,12)-pNT molecules. Thus, aligning the cylinder axes in parallel, the pNT molecules were bundled on a square plane with tetragonal packings (Fig. 3B). The minimum intercylinder distance was 2.28 nm, slightly larger than 1.98 nm, which was expected for van der Waals contacts of cylinders of the 1.64-nm diameter (1.64 + 0.34 = 1.98 nm). In the intercylinder 0.30-nm space, residual electron densities were observed (fig. S21), which was most likely due to intercalated solvent molecules (PhCl). The tetragonal packing also resulted in large interstitial sites between the cylinders at the diagonal positions (28, 29), and the presence of entrapped substances there was also indicated by residual electron densities. The nanometer-sized void spaces at the interior and interstitial sites resulted in a large void fraction of 63% (fig. S22) (29). The tetragonal bundles persisted when multiple C70 molecules were entrapped in a form of Embedded Image. When the crystal was grown in the presence of an excess amount of C70 (approxinately four equivalents), we obtained a single crystal that included C70 molecules both at the interior and interstitial sites of the tetragonal bundles (Fig. 3C and fig. S23). The cylinder structure was essentially identical to that found in the C70-free crystal, with a diameter of 1.67 nm (fig. S19), and the minimum intercylinder distance was 2.00 nm. The intercylinder distance was comparable with the van der Waals contacts of 1.67-nm cylinders (1.67 + 0.34 = 2.01 nm), and the intercalated molecules were not observed in this crystal. We may interpret that the packing behaviors originate from donor-acceptor interactions between pNT and guest molecules. Supramolecular encapsulation in these constructs is a promising application to explore in the future (30, 31). We believe that the persistent tetragonal packing indicates that the bundle structures are under the influence of the symmetry of the nanotubes—the D4h-(12,12)-pNT molecules in this study. There are contradictory viewpoints on roles of the interstitial sites in the adsorption behaviors of CNTs (28, 29), which could also be due to the presence of various inhomogeneous bundles of different symmetry. The present pNT crystal may be a useful reference substance in this context.

The concise synthesis of pNT molecules should be adaptable to product lengthening. For example, installation of trimethylsilyl groups in place of t-Bu groups for the iododesilylation could facilitate lengthening from the halogen linkers. The lengthening reactions could readily tolerate polymerization protocols (20), and we theoretically investigated electronic structures of such infinite pNT by extrapolating the (12,12)-pNT molecule in density functional theory (DFT) calculations (fig. S1) (32). The DFT calculations of the (12,12)-pNT molecule revealed highly degenerate and dense characters of delocalized molecular orbitals (Fig. 4A and figs. S24 and S25), and a comparison with those of benzene revealed effects of π-conjugation of 40 phenines (240 p-orbitals) in a cylindrical framework (fig. S26). Further comparison of the pNT energetics with an experimental optical gap (3.59 eV, λedge = 345 nm) from an ultraviolet-visible absorption spectrum (Fig. 4B) qualitatively confirmed the validity of the calculations [highest occupied molecular orbital–lowest unoccupied molecular orbital (HOMO-LUMO) gap = 3.73 eV]. The (12,12)-pNT molecule was then used as a base template (2.5 layers) for simulation of an infinite (12,12)-pNT structure, with periodic defects (Fig. 4C). Under periodic boundary conditions, the electronic density of states (DOS) of infinite (12,12)-pNT was obtained to reveal semiconducting character with a bandgap of 2.68 eV (Fig. 4D). A reference DOS with an infinite, defect-free (12,12)-CNT was also investigated by using an identical method to show the gapless, metallic nature of the armchair-type CNT (Fig. 4D) (6). Therefore, the results show that installation of periodic defects in CNTs should allow for manipulations of their electronic character. Recently, both top-down and bottom-up methods have been developed for the preparation of graphenes with periodic defects (9, 10), and their anomalous structures (antidot lattices and nanomeshes) have attracted attention in the interest of manipulating the electronic characteristics of nanocarbons (33, 34). Our experimental and theoretical results show that the rolled variants (5, 6)—CNTs with periodic defects—could be an intriguing next target for exploration.

Fig. 4 Theoretical and experimental views of (12,12)-pNT with periodic defects.

(A) Histogram sorting 192 orbitals of the (12,12)-pNT molecule by energy, from the DFT calculations. Representative orbitals are provided in figs. S24 and S25. (B) An absorption spectrum of a 2.0 × 10−6 M solution of (12,12)-pNT in CH2Cl2 at 25°C. (C) Posited structure of infinite (12,12)-pNT from extrapolation of the (12,12)-pNT molecule (red, 2.5 periodic layers). A ruler shows the periodicity of the pNT structure. (D) Electronic density of states from DFT calculations.

Supplementary Materials

www.sciencemag.org/content/363/6423/151/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S26

Tables S1 to S4

References (3547)

References and Notes

Acknowledgments: We thank H. Yamamoto (IMS/ERATO) for valuable discussion. We were granted access to the x-ray diffraction instruments of SPring-8 (2018A1064 and 2018A1257). Funding: This study is partly supported by JST ERATO (JPMJER1301) and KAKENHI (17H01033, 17K05773, 17K05772, 16K04864, 25102007). Author contributions: H.I. launched the project; Z.S, K.I., and H.I. conceived the design; and Z.S. and T.M.F. synthesized the target. Z.S., K.I., and S.S. performed the crystallographic studies, and Z.S., K.I., T.K., and R.A. performed the DFT calculations. All authors analyzed and discussed the results, and Z.S., K.I. and H.I. wrote the manuscript. Competing interests: Z.S., K.I., and H.I. are inventors on U.S. patent application no. 62/745,444 submitted by the University of Tokyo, which covers the syntheses described in the manuscript. Data and materials availability: Crystallographic data are available free of charge from the Cambridge Crystallographic Database Centre (CCDC 1844346 and 1868931). All other data are available in the main text or the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article