One-dimensional van der Waals heterostructures

Structure analysis of SWCNT-BNNT heterostructure

We present the initial growth step: the formation of the SWCNT-BNNT 1D heterostructure. We used SWCNTs as a template and synthesized additional hexagonal boron nitride (BN) layers by chemical vapor deposition (CVD). Figure 1C shows a representative high-resolution transmission electron microscope (HRTEM) image of this coaxial heterostructure (additional images are available in fig. S1A). In a conventional HRTEM image, this nanotube is not distinguishable from a triple-walled pure carbon nanotube. The aberration-corrected HRTEM image of a similar tube revealed a contrast of stacking of two perfect nanotubes (Fig. 1, D and E). However, given that the starting material is purely single-walled before we perform a post-BN coating, we expect that the outer wall or walls are BN. This is supported by electron energy-loss spectroscopic (EELS) mapping (Fig. 1F). Because the reaction occurs on the outer surface, unlike previous attempts inside a nanotube (17, 18), we achieve continuous coating and highly crystallized outer BNNTs. The number of outer BNNT walls can be adjusted from a minimum of one to a maximum of five to eight, depending on the duration of BN CVD (fig. S1, B and C). These different layers grow independently, but the first layer is always the longest. The walls of our SWCNT template are very clean, so the nucleation is usually observed at the end of a suspended region, where a SWCNT is connected with another SWCNT or a SWCNT bundle (fig. S2). Occasionally, nucleation also occurs simultaneously at both ends of a suspended SWCNT region. Nucleation from the middle of a SWCNT is rarely observed.

One critical feature of the current structure is that each layer is a single crystal, which distinguishes this study from all previously reported coaxial tubular structures. This perfect crystallization is shown by the ED patterns provided in Fig. 2, A to C, and fig. S1. The perfect SWCNT-BNNT coaxial single crystals could reach a few hundreds of nanometers to ~2 μm and were limited only by the length of individually isolated regions in our starting SWCNTs. The other important feature of the current structure is its small diameter. Many of our SWCNT-BNNT structures are thinner than 2 nm, and the ternary SWCNT-BNNT-MoS₂ are 3 to 5 nm (described in a later section). This small dimension is essential for accessing the distinctive properties of 1D materials, such as confinement of excitons in the 1D crystals.

• Download high-res image
• Open in new tab
• Download Powerpoint

We characterized these coaxial crystals with several other techniques. X-ray photoelectron spectroscopy (XPS) revealed B–N and C–C bonds in this sample but no apparent peaks of C–N and B–C, which confirmed that BN and carbon moieties were chemically isolated (fig. S3). Optical absorption spectra (fig. S4A) revealed a peak near 205 nm, and the peak intensity increased with growth time (number of outside BNNT layers). Cathode luminance spectra (fig. S4B) of the sample confirmed the existence of outer BN by emissions in the ultraviolet range. Also, the growth strategy was shown to apply to SWCNTs with different morphologies, including vertical arrays, horizontal arrays, random networks, and suspended SWCNTs between silicon pillars (fig. S5).

The formation of this SWCNT-BNNT heterostructure follows an open-end growth mechanism (shown schematically in movie S1). In this scenario, the extension of the second layer happened only at the open edge of the BNNT. The third (and beyond) layer followed a similar approach but nucleated in a later stage. This pattern is similar to the growth of additional layers in 2D material, but it has very rarely been observed in previous growth of carbon nanotubes (19). The open-edge growth mechanism is supported by the many atomic steps observed in heterostructures with incomplete growth of outer BNNT crystal layers (Fig. 2, A and B). Our starting SWCNTs have an ultraclean surface and our CVD chamber is a clean, low-pressure system, which avoids any atomic impurities on the SWCNT surface or contamination from the chamber that could cause imperfection in the outer nanotube.

The atomic step in composition along the growing nanotube also allowed us to obtain nano-area ED patterns (20) of the pristine SWCNT and the same tube after BNNT growth, which allowed us to assign chirality of each layer. In Fig. 2C, the inside SWCNT is assigned as (17, 13) and the outside BNNT as (33, 3). In another case, a (34, 0) single-walled BNNT formed on the surface of a (16, 14) SWCNT (fig. S6). Even without these atomic steps, the 2% difference between the lattice constants of SWCNT and BNNT (21) allows us to distinguish them in the ED patterns.

In a collection of 74 SWCNTs and 40 SWCNT-BNNT double-walled nanotubes (details shown in figs. S7 and S8 and table S1), a greater number of SWCNTs were in near-armchair form (Fig. 2D). This near-armchair enrichment was consistent with previous experimental and theoretical analyses (22, 23). However, the outer BNNTs were randomly distributed (with a slight preference for the zigzag conformation). No chiral angle dependence was observed in these SWCNT-BNNT heterostructures, so symmetry and lattice matching did not limit our ability to combine materials using our technique (24). This absence of correlation of inner and outer layers differed from what has been observed in CVD fabrication of 2D vdW heterostructures, where the growth layer is usually aligned with the base material. We attribute this difference to the symmetry breaking in 1D materials compared with that in their 2D counterparts.

Optical, thermal, and electronic characterization of SWCNT-BNNT heterostructure

Raman and photoluminescence (PL) spectra of the SWCNT-BNNT showed peaks typical of pristine inner SWCNTs, which showed that their structural integrity was preserved (Fig. 3, A and B). However, a G band downshift of 5 to 10 cm⁻¹ appeared in both isolated SWCNT-BNNT heterostructures and heterostructure networks. That downshift was a fingerprint for BNNT coating in our experiments. We tentatively attributed this downshift to the thermal strain between SWCNT and BNNT. The difference in thermal expansion during BN synthesis caused a slight distortion for SWCNTs along the tube axis. The outer BN layers acted as a protective coating and protected the inside of the SWCNTs against oxidation. Burning of the SWCNT inside BNNT started at 700°C in air, whereas naked SWCNTs burned at 400°C (Fig. 3C), as measured in a Raman cell. After annealing the heterostructure in oxygen, inside SWCNTs were removed, and clean and crystallized single-walled and few-walled BNNTs were obtained (fig. S9).

• Download high-res image
• Open in new tab
• Download Powerpoint

The BNNT coating did not change the intrinsic electronic transport of the inner SWCNT. We fabricated a back-gated field-effect transistor (FET) on an individual SWCNT-BNNT heterostructure (Fig. 3D). This SWCNT-BNNT was synthesized from a suspended SWCNT and then transferred onto a Si substrate. The SWCNT-BNNT FET showed performance similar to that of a SWCNT FET and had an on/off ratio of 10⁵ (Fig. 3, E and F), which means that the high quality of the SWCNT was preserved. To further investigate the isolation and tunneling of outer BNNT layers, we measured the current through different BNNT layers in a TEM equipped with two probes (Fig. 3G). We distinguished the numbers of BNNT layers on SWCNTs directly from TEM and simultaneously measured the electron conduction at local positions separated by ~10 nm (Fig. 3, H to I). Current-voltage (I-V) curves—measured at the positions of the SWCNT-BNNT with zero, one, two, and three layers of BNNTs (Fig. 3I)—show the exponential increase of the zero-bias resistance with the layer number, indicative of direct tunneling through the insulating BN layers (Fig. 3H, bottom). This characteristic of tunnel current is similar to that of 2D BN layers exfoliated from bulk crystals (25, 26), and it was consistent with the layer quality being comparable to exfoliated crystals. The quality of these structures could allow for the experimental exploration of properties that have been theoretically predicted for 1D heterostructures, such as SWCNT-BNNT being a topological insulator if combined in a proper symmetry (12).

MoS₂-based binary and ternary heterostructures

MoS₂ 2D sheets have been studied intensively as the representative transmission metal dichalcogenide material in recent years (27). Multiwalled MoS₂ nanotubes, with diameters usually >20 nm, and their hybrid materials are well known (28–30), but single-walled, single-crystal MoS₂ nanotubes have not been convincingly demonstrated in previous studies. Thus, we explored the growth of MoS₂ on SWCNTs. Figure 4, A to C, shows the atomic structure, TEM, and scanning TEM (STEM) images of SWCNT-MoS₂ coaxial nanotubes obtained after applying our growth strategy. The MoS₂ nanotube has much stronger image contrast than carbon in both TEM and STEM images (additional images and EELS mapping are shown in fig. S10). Single-walled MoS₂ nanotubes were predicted to have a direct band gap, in contrast to multiwalled nanotubes that have an indirect band gap (31), and to exhibit strong quantum confinement effects (32).

• Download high-res image
• Open in new tab
• Download Powerpoint

We observed a strong diameter dependence for the formation of single-walled MoS₂ nanotubes. Unlike BNNT wrapping, the yield of MoS₂ nanotubes was very low (less than 1%), and seamless wrapping was only observed on large-diameter (>3 nm) SWCNTs. We performed simulations of the strain energy of single-walled MoS₂ nanotubes with different diameters, D. A 1/D² relation was obtained (Fig. 4D), which suggests that, in small diameter range, strain energy was much higher than for SWCNTs and BNNTs (31, 33, 34). This difference can be simply attributed to the thickness of a single layer of MoS₂, containing three atomic planes and becoming unstable when rolled into a tubular structure. This also explains why MoS₂ nanotubes were only seen previously as multiwalled or on multiwalled CNTs (14, 28). Thus, the minimum diameter of a single-walled MoS₂ nanotube should be much larger than that of SWCNTs. The MoS₂ nanotubes (fig. S10, C to F) have diameters ranging from 3.9 to 6.8 nm and formed only on SWCNTs with diameters of at least 3 nm. Because most of our starting SWCNTs were thinner than 3 nm, the yield of SWCNT-MoS₂ heterostructures was low.

Finally, we grew a ternary, SWCNT-BN-MoS₂ coaxial nanotube (Fig. 5, A to D). This 5-nm–diameter structure consisted of an inner SWCNT, a middle three layers of BNNT, and a single outer layer of MoS₂. The elemental information was visualized by EELS mapping (Fig. 5D). Typical ED patterns in Fig. 5E reveal diffractions from all three crystal layers, with MoS₂ having the strongest intensity and SWCNT and BNNT showing weaker but distinguishable contrast. The unidirectional distribution of the patterns also supports the coaxial feature. The four sets of hexagonal pairs with different colors indicate the orientation of atomic arrangement in the upper and lower surfaces to the electron beam.

• Download high-res image
• Open in new tab
• Download Powerpoint

Because the nanotube diameter increased from 2 to >3 nm after BNNT coating, the synthesis of an additional MoS₂ nanotube became much easier, and the yield of the SWCNT-BNNT-MoS₂ heterostructure reached ~10% after 20 min of MoS₂ CVD. A rough count of the yield of heterostructure versus CVD time is provided in fig. S11. The coating can be produced on the centimeter-length scale, and the difference can be observed between the original SWCNT, SWCNT-BNNT, and SWCNT-BNNT-MoS₂ films even with the naked eye (Fig. 5F). The optical absorption spectrum (fig. S12) of the sample revealed the photon absorption from three different layers.

We noticed that the innermost SWCNT and outermost MoS₂ in a SWCNT-BNNT-MoS₂ heterostructure are electronically coupled. This is supported by the different PL intensities of the SWCNT-BNNT-MoS₂ and BNNT-MoS₂ heterostructures. In the former case, PL of MoS₂ was markedly quenched by the existence of SWCNT (fig. S13). In the latter case, however, PL of MoS₂ was 10 times as high. We calculated the band alignment of a graphene, BN, and MoS₂, and the Dirac point of graphene was found to be −4.26 eV, only 0.1 eV below the conduction band edge of MoS₂ (fig. S14). This calculation suggests that, if the starting SWCNT could be a small-diameter, semiconducting SWCNT, the SWCNT-BNNT-MoS₂ heterostructure is a type II junction. Interlayer excitons probably exist, even though electrons and holes are spatially separated by few layers of BNNT.