Crystal structure. Red rod crystals of 1 (Supplementary Fig. 1) were synthesized via the solvothermal method (see the Experimental section in the Supporting Information for details). Single-crystal X-ray diffraction (SCXRD) analysis revealed that 1 crystallized in a highly symmetrical trigonal system with an R-3 space group (Supplementary Table 1), and exhibited a unique nanotubular structure (Fig. 1). The asymmetric unit of 1 contains 19 crystallographically independent sites, comprising two Cu, three Ge, and nine Se, in addition one water guest molecule, four potassium counter-cations (Supplementary Fig. 2). In addition, the valence state of Cu was confirmed to be monovalent by X-ray photoelectron spectroscopy (XPS) (Supplementary Fig. 3), while the phase purity was verified using powder X-ray diffraction (PXRD) (Supplementary Fig. 4). Through combination of the results of SCXRD, energy-dispersive spectroscopy (Supplementary Fig. 5), thermogravimetric analysis (TGA) (Supplementary Fig. 6) and elemental analysis, the empirical formula of 1 was determined to be K4Cu2Ge3Se9(H2O).
As shown in Fig. 1, the primary building unit (PBU) of 1 is a heterometallic supertetrahedral T2-CuGeSe cluster, which can be viewed as a homometallic supertetrahedral T2-GeSe cluster with one of the four corner germanium sites occupied by copper (Cu(1) in Supplementary Fig. 2). The [Cu(1)Se4] tetrahedron unit is slightly distorted in comparison with the [GeSe4] tetrahedron unit in the homometallic supertetrahedral T2-GeSe cluster (Fig. 1a and 1b), likely due to the mismatch of the local charge caused by the lower valence state of Cu compared with that of Ge. In the traditional view, these six staggered PBUs are connected end-to-end by six Cu ions to form a novel giant hexagonal wheel-shaped cluster ([Cu6(CuGe3Se10)6]) with the C3i point group (Fig. 1c and Supplementary Fig. 7), which can serve as the secondary building units (SBUs) for further assembly. More specifically, in this wheel-shaped cluster, each single independent Cu(2) atom is located close to the Cu(1) site and interlinked two adjacent T2-CuGeSe clusters via three Cu-Se bonds (Fig. 1c). Among the three Cu-Se bonds, two originate from the bonding of the Cu(2) atom with two edge Se atoms next to the Cu(1) site of one T2 cluster, while the remaining Cu-Se bond comes from the bonding of the Cu(2) atom with the corner Se(4)2− of the other T2 cluster. Furthermore, the three Cu-Se bonds are coplanar with unequal length, and the corresponding Se-Cu-Se bond angles are also diverse from one another (Supplementary Fig. 8). It should be noted that the planar trigonal coordination mode of Cu between two supertetrahedral clusters is rare,32 in contrast to the single linearly-coordinated Cu+ species between clusters33–35 and the tetrahedrally-coordinated Cu+ residing in the cluster. According to Pauling’s electrostatic valence rule, the theoretical residual charges of the Se2− ions coordinated with the Cu+ ions in an isolated T2-CuGeSe cluster, all increased by -0.75, when replacing one Ge4+ cation from an isolated T2-GeSe cluster with one low-valent Cu+ cation (Fig. 1a and 1b), thereby resulting in a serious mismatch of local charge. To address this issue, a single low-valent Cu+ cation was introduced and triangularly coordinated with three Se2− ions between two T2 clusters to maximally balance the excess local negative charge. Meanwhile, accompanying the looser geometrical demand of selenide compared to sulfide, an exceptional heterometallic supertetrahedral T2 cluster-based wheel-shaped ring was formed, which contrast to the 3D frameworks constructed by homometallic T2-GeS sulfide clusters with linearly-, trigonally- or tetrahedrally-coordinated low-valent transition metal ions (such as Cu+, Ag+, and Mn2+).32–37 To further lessen the excess high negative charge at the terminal Se site that is linked with the Cu(1), and facilitate the global charge balance, these large wheels are connected end-to-end through sharing of the terminal Se2− ion close to the Cu(1) site with the Ge(3)4+ of the other T2 cluster, ultimately forming an unprecedented supertetrahedral chalcogenide-cluster-based infinite nanotube ([Cu6(CuGe3Se10)6]n) along the c direction, with an outer diameter of 25.97 Å and an inner opening window of 12.91 × 19.86 Å (Fig. 1c and 1d). To the best of our knowledge, this structure represents the largest example of a crystalline inorganic nanotube to date (Supplementary Table 2). Moreover, the wall of the nanotube was composed of 16-membered ring (16 MR) windows formed by four T2-CuGeSe clusters with pore sizes of 3.59 × 6.47 Å (Supplementary Fig. 9). Because 1 cannot be dispersed in common solvents, an ultrathin section sample of 1 was observed by high resolution transmission electron microscopy (HRETM) (Supplementary Fig. 10), revealing the high crystallinity of 1 with distinct interplanar lattice fringes of 0.78 nm, which is consistent with the observation of an XRD peak at 2θ = 11.7° (Supplementary Fig. 4).
Furthermore, the negative charges on the skeleton of the nanotube are balanced by pure K+ ions located at the gaps between the nanotubes and by hydrated K+ ions filling in the channel. These K+ ions, along with guest water molecules residing inside and outside of the nanotube, play a key role in constructing and stabilizing the Nanotubes, and promote their further packing into a highly ordered honeycomb-like hexagonal symmetrical array (Fig. 1e) via complex weak interactions such as hydrogen bonding and electrostatic interactions (Supplementary Fig. 11).6, 7, 20 Thus, control experiments demonstrated that K2S is indispensable for the formation of 1. Figure 1f also displays the pillared stacking of the wheel clusters in the axial direction and the assembly of nanotubes to form a 1D tubular superlattice. The nanotube can also be viewed in a different way, where six 1D chains, formed by the end-to-end linkage of T2-CuGeSe clusters through sharing corner Se2− ions coordinated with Cu(1) and Ge(3) atoms, bind alternately with six Cu+ ions in the same manner as above to form nanotubes (Fig. 1c, 1g, and Supplementary Fig. 12). This assembly mode is supported by the observation of 2, a 1D chain structure based on T2-CdGeSe clusters (Supplementary Fig. 13), which forms upon replacing the copper salt with a cadmium salt during preparation. Compound 2 was comprehensively characterized (Supplementary Fig. 14–18 and Supplementary Table 3), and upon comparison with the structure of 1, Cd2+ was found to occupy the Cu+ site of the T2-CuGeSe cluster in 2, resulting in decreased theoretical residual charges from the surrounding Se atoms, which correspondingly reduces the further bonding capability of the Se2− ions on the edges of the cluster toward other metal ions, thereby leading to the formation of 1D chains rather than 1D tubules. We therefor speculated that the dissimilar ionic radii and coordination modes between Cd2+ and Cu+ may also contribute to such difference.
Electrical conductivity measurement. The optical indirect bandgap of 1 was calculated to be 1.03 eV from the transformed solid-state UV-Vis diffuse reflectance spectrum (Supplementary Fig. 19). This value was considered relatively narrow and largely red shifted by 0.63 eV compared with the corresponding value of 2 (i.e., 1.66 eV), thereby indicating the superior conductivity of 1.38 This was confirmed by electrical conductivity measurements on a single crystal of 1 through a direct-current two-terminal method (Fig. 2a and 2b).
As shown in Fig. 2c, the electric conductivity of 1 was determined to be 7.6 × 10− 6 S cm− 1 at 40°C along the c axis, and was positively related to temperature, exhibiting typical semiconductive characteristics. The corresponding activation energy (Ea) was calculated to be0.52 eV (Fig. 2d). The electrical conductivity of 1 was found to be approximately 10000-times higher than those of other crystalline nanotube arrays, and among one of the highest values for crystalline semiconductor materials containing copper or/and chalcogenide elements (Supplementary Table 4).6, 7, 39–42 Moreover, the photoconductivity of 1 was investigated. As shown in Fig. 2e, 1 exhibits a rapid wavelength-dependent response upon illumination with 400–700 nm light, without any apparent attenuation during the on/off switching cycles, thereby indicating the efficient separation of photogenerated charge carriers.43 The responsivity (Rλ), detectivity (D*) and external quantum efficiency (EQE) at different wavelengths are summarized in Supplementary Table 5. Interestingly, the largest value of Rλ was achieved at 600 nm (Fig. 2f), which is inconsistent with the maximum absorption in UV-Vis spectrum (Supplementary Fig. 19a), thereby suggesting a temporary unclear process that enhanced the photocurrent at longer wavelengths.43 Moreover, the Rλ and D* values of 1 increased as the light intensity decreased (Supplementary Fig. 20). Thus, with its outstanding conductivity, fast turn-over response, and good reproducibility, 1 displays potential for use in optoelectrical applications.43 In the context of 2, the conductivity was determined to be only 1.89 × 10− 9 S cm− 1 at 25°C, with an Ea of 0.64 eV (Supplementary Fig. 21), while its photoconductivity performance was also significantly poorer than that of 1 (Supplementary Fig. 22). The enormous conductivity disparity between 1 and 2 was mainly attributed to their different structures and compositions. We speculated that the substitution of Cd with Cu in the 1D T2-CdGeSe chain may largely improve the intrinsic conductivity due to the superior conductivity of Cu compared to Cd. Combined with the narrow optical bandgap, a good oriented photoconductive behavior can be observed in 1.
Theoretical DFT study of 1. To gain deep insight into the intrinsic electronic properties of 1, density functional theory (DFT) calculations on the band structure and the projected density of states (PDOS) were performed, whereby 1 was found to exhibit a quasi-direct bandgap of 0.92 eV at the gamma point (left of Fig. 3a), consistent with the experimental value.
Compared with the almost flat band lines close to the valence band maximum (VBM) along the whole Brillouin zone, which are mainly dominated by the Cu d orbital and the Se p orbital (right of Fig. 3a and Supplementary Fig. 23), the bands near the conduction band minimum (CBM), which are contributed primarily by the Ge s orbital and the Se p orbital, along with negligible contributions from the Cu d orbital (right of Fig. 3a and Supplementary Fig. 23), show a significantly steep dispersion with an energy difference of ~ 0.71 eV (0.63 eV) along the Γ→A (K→H and M→L) directions in reciprocal space, corresponding to the tubular direction (or c axis) in real space, while for other paths with high symmetry points, the dispersion widths are small (maximum energy difference < 0.1 eV). The relatively large dispersion strength of the energy bands indicates the facile transport of charge carriers along the c direction,38, 44 which is of paramount importance to photoconductive devices. Moreover, the narrow band width and flat band lines near the VBM were attributed to the relatively larger localization of the Cu d orbital compared to the Ge s orbital. The other elements (K, O and H atoms) do not contribute to the electronic band edges. In addition, according to the charge density distributions of the VBM and the CBM (Fig. 3b-e), the CBM was determined tobe mainly localized on the Se atoms and the Ge(2) and Ge(3) atoms, while Ge(1) atoms make no contribution. Therefore, combined with the above analysis and the features of the crystal structure, we deduced that the excellent conductivity of 1 may be attributed to the more facile oriented transport of electrons in the tubular direction (or along the c axis), in addition to obstructed carrier transport in the ab plane perpendicular to the tubular direction.
In summary, we report an unprecedented supertetrahedral chalcogenide cluster-based crystalline inorganic nanotube array, representing an important first step toward novel nanotube materials. The fine electrical conductivity, oriented photoconductive property, and well-defined structure of 1 render it a fascinating structural model in the optoelectronic and electronic fields. In addition, the precise potassium ions located around or within the nanotubes introduce a platform for the further study of ion transport. Finally, research exploring the syntheses of supertetrahedral chalcogenide cluster-based nanotubes with attractive functions and properties, such as ion-exchange and sensing, are currently underway, and the results will be presented in future work.