Introduction

Tungsten disulfide (WS2) nanotubes1 are cylindrical nanotubes composed of WS2 sheets, which are semiconducting transition metal dichalcogenides (TMDCs), and have attracted significant attention over the past few decades owing to their unique electronic, optical, and optoelectronic properties2,3. Unlike carbon nanotubes, WS2 nanotubes exhibit solely semiconducting properties regardless of how they are rolled4 and show advantages for applications such as sensors5, field effect transistors6,7,8, optoelectronic devices9,10, and thermoelectric devices11. Owing to their chiral nanotube structures, WS2 nanotubes exhibit several interesting properties, such as nonreciprocal conduction8,12 and bulk photovoltaic properties12, 13. It has been theoretically predicted that the electronic structure of WS2 nanotubes depends on the nanotube structure, which is characterized by the nanotube diameter14,15,16,17 and wall number18. In particular, WS2 nanotubes with small diameters are expected to exhibit significant band structure modulation due to curvature effects15,19,20, leading to unique properties different from those of WS2 sheets and WS2 nanotubes with large diameters. Thus, an important goal in this field is to clarify the correlation between the properties and nanotube structures experimentally. Most of the studies to date have been performed on relatively large-diameter multiwall nanotubes (~ 100 nm)21,22,23 and have revealed structure-dependent properties, such as size-dependent exciton-polaritons17, structure sorting optical properties16, and light-matter interactions14. In contrast, very few studies24 have investigated the properties of small-diameter WS2 nanotubes because of their challenging synthesis (owing to the high folding energy of WS2 sheets). Although the synthesis of MoS2 nanotubes with small diameters outside carbon nanotubes and inside boron nitride nanotubes has been reported25,26, the structure–property relationship in small-diameter WS2 nanotubes has not been experimentally investigated to date.

One approach to synthesize TMDC nanotubes is the chalcogenization of transition metal oxide nanowires, which has been used for large-scale synthesis22. In this process, the oxide nanowires are synthesized and then converted into nanotubes by sulfurization. Therefore, to synthesize small-diameter nanotubes with few walls, it is necessary to synthesize small-diameter oxide nanowires. Recently, several studies have attempted to reduce their diameter and narrow their diameter distribution by modifying the synthesis of oxide nanowires; as a result, nanotubes with diameters of 70 nm27 and 20 nm28,29 have been fabricated from oxide nanowires synthesized from chemical vapor-deposition (CVD) and solvothermal methods, respectively. In addition, although the structure distribution is not known, nanotubes with diameters of 8–25 nm have been obtained by this approach24. Therefore, these approaches exhibit high potential for the production of small-diameter TMDC nanotubes.

This study reports the synthesis of small-diameter WS2 nanotubes by the chalcogenization of structure-controlled tungsten oxide nanowires fabricated via CVD. Detailed structure characterization confirmed that the as-prepared multi-walled WS2 nanotubes showed the innermost diameter of 6 ± 3 nm, and the wall number of 5 ± 2. The utilization of single-crystalline small-diameter tungsten oxide nanowires as the starting material enabled the synthesis of high-crystallinity small-diameter WS2 nanotubes. The optical absorption and photoluminescence (PL) characteristics of the samples were used to elucidate the effects of the nanotube diameter on the bandgap. The PL of the as-prepared small-diameter WS2 nanotubes was red-shifted from the PL of WS2 flakes and large-diameter WS2 nanotubes, indicating a bandgap reduction, in agreement with the results of theoretical calculations15,19.

Results and discussion

In this study, WS2 nanotubes were synthesized on Si/ SiO2 (300 nm) or quartz substrates under Ar gas at a sulfurization temperature of 600 °C from tungsten oxide nanowires. The nanowires were first grown on silicon substrates by CVD30,31 and then transferred onto Si/ SiO2 or quartz substrates for sulfurization28,29. Details of the synthesis are provided in Methods section. As mentioned previously, the synthesis of small-diameter tungsten oxide nanowires is a prerequisite for the synthesis of small-diameter WS2 nanotubes. In this study, as described in the Methods section, we used WO2.9 powder for tungsten source and the synthesis of the oxide nanowires was performed in Ar + H2 (4%) atmosphere.

Typical low and high-magnification transmission electron microscopy (TEM) images (Supplementary Fig. S1(a,b) in the Supplementary information (S.I)) and the diameter distribution shown in Supplementary Fig. S1(c) in S.I. indicate that the tungsten oxide nanowires used as precursors for nanotube synthesis are single crystalline and exhibit an average diameter (with standard deviation) of 13 ± 5 nm. The average diameter of these tungsten oxide nanowires is slightly smaller than that of previously reported solvothermally synthesized nanowires (~ 17 nm)28. Field-emission scanning electron microscopy (FESEM) images of the nanowires (Supplementary Fig. S2(a) in S.I.) indicate high-density nanowire growth on the substrate, while X-ray diffraction (XRD) patterns (Supplementary Fig. S2(b) in S.I.) indicate that the nanowires comprise the monoclinic W18O49 phase, consistent with literatures30,31,32,33,34,35.

TEM images of the as-synthesized WS2 nanotubes indicate hollow, multi-walled structures (Fig. 1a). The outermost and innermost diameters of the nanotubes were statistically analyzed to evaluate the nanotube structure distribution (Fig. 1b). The average outermost and innermost diameters (with standard deviations) of the as-prepared nanotubes are 13 ± 4 nm and 6 ± 3 nm, respectively, and the average wall number (with standard deviation) is 5 ± 2 (Supplementary Fig. S3 in S.I.). Figure 1c,d show high-magnification TEM images of the WS2 nanotubes. The average diameter of the as-prepared WS2 nanotubes is significantly smaller than that of commercially available WS2 nanotubes (average diameter ~ 100 nm, purchased from Nanomaterials Co., Supplementary Fig. S4 in S.I.)21,22,23 and slightly smaller than that of WS2 nanotubes derived from solvothermally synthesized tungsten oxide nanowires (average diameter: ~ 20 nm)28,29. The diameters of the as-prepared WS2 nanotubes are compared with those of previously reported WS2 nanotubes28,29 in Supplementary Table S1. Raman spectroscopy and energy-dispersive X-ray spectroscopy (EDS) (Supplementary Fig. S5 in S.I.) were used to confirm the elemental composition and quality of the WS2 nanotubes. X-ray photoelectron spectroscopy (XPS) (Supplementary Fig. S6 in S.I.) indicates that the WS2 nanotubes comprise the 2H–WS2 phase36, while FESEM images (Supplementary Fig. S7(a) in S.I.) indicate that the presence of impurities like nano-flakes is negligible. Additionally, XRD analysis (Supplementary Fig. S7(b) in S.I.) confirms that the WS2 nanotubes comprise the 2H–WS2 phase. Moreover, the absence of XRD signals corresponding to tungsten oxide nanowires in the spectrum of the as-prepared WS2 nanotubes indicates almost complete conversion of oxide nanowires to WS2 nanotubes during chalcogenization. Fast Fourier transform (FFT) patterns of the TEM images were used to investigate the chirality of the WS2 nanotubes (with three walls) (Supplementary Fig. S8 in S.I.). FFT patterns indicate three different chiral angles; therefore, the different walls of the nanotube exhibit different chirality.

Figure 1
figure 1

TEM observation of the as-synthesized WS2 nanotubes. (a) Low-magnification TEM image of the WS2 nanotubes. (b) Histogram of the outermost and innermost tube-diameter distributions of the WS2 nanotubes. (c), (d) High-magnification TEM images of the WS2 nanotubes.

Full size image

Here, the influence of the nanotube structure on the optical properties of the as-prepared nanotubes was investigated. The absorption spectra of WS2 flakes (with a thickness of 20–40 nm), commercially available WS2 nanotubes with an average diameter of 100 nm (L-WS2 NTs; purchased from Nanomaterials Co.), and WS2 nanotubes with an innermost diameter of 6 ± 3 nm (S-WS2 NTs; synthesized in this study) were measured and analyzed. For optical measurements, the S-WS2 NTs and L-WS2 NTs were directly dispersed in methanol, while the WS2 flakes were dispersed in methanol after liquid-phase exfoliation (see S.I.)37,38.

The optical extinction (absorption) spectra of the WS2 flakes, L-WS2 NTs, and S-WS2 NTs, recorded by a conventional UV–VIS spectrometer (UV-3600, Shimadzu), are shown in Fig. 2. The A, B, and C excitons of the WS2 flakes are observed at 1.97 eV, 2.33 eV, and 2.65 eV, respectively, in good agreement with the literature39,40. The spectrum of the L-WS2 NTs, unlike that of the WS2 flakes, exhibits dips close to the position of the A and B excitons, owing to the polaritonic nature of the nanotubes (due to the tube diameter of the L-WS2 NTs)14. This phenomenon is caused by strong exciton–photon coupling in the optical-cavity structure of nanotubes with diameters of ~ 100 nm. We examined the optical absorption spectra of L-WS2 NTs by using an integrating sphere in the UV‒VIS spectrometer and confirmed that the A exciton peak of L-WS2 NTs is located at 1.97 eV (Supplementary Fig. S9 in S.I.), which is similar to that of WS2 flakes. Theoretical calculations indicate that this polaritonic effect disappears when the nanotube diameter decreases below 40 nm, owing to difficulties in nanotube-cavity optical confinement14. Therefore, polariton-like dip structures are absent in the absorption spectrum of S-WS2 NTs, which exhibit outermost and innermost diameters of 13 nm and 6 nm, respectively. A similar phenomenon is observed in sorted small-diameter WS2 nanotubes16. Due to the suppression of the polaritonic effect in small-diameter nanotubes, the absorption peak at 1.92 eV in S-WS2 NTs can be attributed to the A exciton, showing a slight redshift compared to the A exciton peak in the WS2 flakes.

Figure 2
figure 2

Extinction spectra of WS2 flakes, large-diameter (L-WS2 NTs) and small-diameter (S-WS2 NTs) nanotubes. The black (upper panel), blue (middle panel), and red (lower panel) curves show the extinction (absorption) spectra of WS2 flakes, L-WS2 NTs, and S-WS2 NTs, respectively.

Full size image

To further investigate the effects of the WS2 nanotube structure on their optical properties, PL spectra were investigated. We placed the samples on quartz substrates by drop-casting the dispersions in methanol. The PL measurements for all samples were carried out with a 532-nm excitation laser. Figure 3 shows the PL spectra of the WS2 flakes, L-WS2 NTs, and S-WS2 NTs; all the background signals are removed to highlight the PL signals from the samples. All the samples exhibit PL peaks; however, the PL intensities of the multi-walled WS2 nanotubes and flakes are weak owing to indirect bandgap structures. The PL peak of the WS2 flakes is observed at 1.97 eV, which is consistent with the A-peak position of its absorption but is blue-shifted to the reported PL peak energy, 1.91 eV, of the mechanically exfoliated WS2 bulk sample41. This blue-shift behavior is commonly reported in liquid phase exfoliated WS242. The PL peak of L-WS2 NTs (1.88 eV) is slightly red-shifted from that of the WS2 flakes. Remarkably, the PL peak position of the S-WS2 nanotubes is located at the smallest photon energy (1.80 eV).

Figure 3
figure 3

Photoluminescence (PL) spectra of WS2 flakes, large-diameter commercial nanotubes (L-WS2 NTs), and the as-synthesized small-diameter nanotubes (S-WS2 NTs). The black (upper panel), blue (middle panel), and red (lower panel) curves show the PL spectra of WS2 flakes, L-WS2 NTs, and S-WS2 NTs, respectively.

Full size image

To understand this trend of PL peaks, the bandgap energies of WS2 monolayers and WS2 nanotubes with different diameters were theoretically calculated. The geometric and electronic structures of planar and tubular WS2 were investigated by density functional theory with generalized gradient approximation43,44. The calculated band gap energies are summarized in Table 1. The bandgap energies of the monolayer, (28,28) nanotube (diameter: 5.4 nm), and (17,17) nanotube (diameter: 3.4 nm) are 1.963, 1.962, and 1.900 eV, respectively. Details of the calculated band structures are provided in Supplementary Fig. S10 in S.I. Theoretical calculations indicate that the bandgap decreases as the nanotube diameter decreases, in agreement with previous publications15,19,20,45. It is known that this band-structure narrowing is caused by strain and curvature effects15,19,20,45.

Table 1 Calculated bandgap energies of the WS2 monolayer and nanotubes.
Full size table

Calculations indicate that bandgap narrowing is negligible when the nanotube diameter is greater than 5 nm. Therefore, for L-WS2 NTs, whose diameter is approximately 100 nm, we neglect the influence of the narrowing of the band gap on the redshift of PL compared with the PL of the WS2 flakes. We assume that the slight red-shift of the PL of L-WS2 NTs could be attributed to the lower polariton state formed by exciton-light coupling by the nanotube cavity, as similar PL has been reported in strong coupling single walled carbon nanotube microcavities46. In contrast, the polariton effect, which did not influence the optical absorption of S-WS2 NTs, was negligible in the small-diameter S-WS2 NTs. Theoretical calculations (of this study and in the literature15,19,20,45) predict bandgap narrowing for nanotubes with diameters of ~ 3 nm. The as-synthesized S-WS2 NTs, with an innermost diameter of 6 ± 3 nm, contain 3 nm diameter nanotubes on the inner-wall sides. In contrast to single-walled carbon nanotubes47, as the diameter of TMDC-NTs decreases, the bandgap reduces. This indicates that PL from the smallest (innermost-wall) nanotubes in the multi-walled WS2 nanotubes contributed to the observed redshifted PL spectrum of the S-WS2 NTs.

As the optical absorption characteristics of WS2 nanotubes with a diameter of ~ 10 nm predominantly influence the absorption spectrum of S-WS2 NTs (with an outermost diameter of 13 ± 4 nm), the optical absorption of S-WS2 NTs is similar to that of WS2 flakes. Contrarily, as PL occurs due to energy transfer and relaxation processes from the lowest excited state in nanotubes, the smallest-diameter nanotube (i.e., the innermost-wall tube) contributes predominantly towards light emission. Therefore, the PL of S-WS2 NTs is red-shifted from the PL of WS2 flakes.

The red-shift of PL is sometimes observed by formation of trions in WS2 monolayers48. However, this phenomenon does not explain the large PL red-shift (> 100 meV) observed in this study. Therefore, the observed redshift would be majorly caused by the bandgap narrowing effect. In inorganic fullerenes like WS2 with small layer numbers (< 5), the quantum-size effect in the direction perpendicular to the S–W–S layer induces a blue-shift in the optical absorption of the A exciton49. However, the PL spectra of WS2 flakes with a series of different layer numbers41,50 indicate that the PL-peak position remains almost unchanged on decreasing the number of layers from 10 to 3. The nanotubes synthesized in this study contained an average of five layers; therefore, such quantum effect was assumed to negligibly influence the PL red-shift of the as-synthesized samples.

In this study, small-diameter WS2 nanotubes were synthesized by the sulfurization of tungsten oxide nanowires through the CVD method. The as-prepared WS2 nanotubes showed an innermost diameter of 6 ± 3 nm and a mean wall number of 5. The relationships between the optical properties and nanotube diameter were investigated and a clear red-shift of the PL in the small-diameter WS2 nanotubes was observed. The red-shift of the PL was attributed to bandgap narrowing, in agreement with the results of theoretical calculations. The synthesis of small-diameter TMDC nanotubes reported in this study could guide future studies to clarify the basic properties of WS2 nanotubes, facilitating their optoelectronic applications.

Methods

Synthesis of WS2 nanotubes

To synthesize WS2 nanotubes, first, tungsten oxide nanowires were synthesized by temperature-controlled CVD under vacuum, according to previous studies30,31. Among tungsten oxide compositions, reduced tungsten oxide W18O49 is known to exhibit a tendency of uniaxial growth and contribute to the synthesis of small-diameter nanowires. Thus, in this study, to synthesize oxide nanowires in a more reducing atmosphere than that in the previous studies30,31, the tungsten oxide source was changed from WO3 to WO2.9, and the synthesis atmosphere was changed from air to Ar + H2 (4%). The nanowire synthesis was carried out using a three-zone furnace. A quartz boat with WO2.9 powder (3 g, 99.99%, Alfa Aesar) was placed in a quartz tube and positioned upstream/midstream of the furnace, and a quartz boat with Si substrates was positioned downstream of the furnace. During synthesis, the system pressure was maintained at 0.6–0.9 Pa with a vacuum pump while supplying 0.08 sccm of Ar/H2 (Ar:H2 = 96:4). The temperatures of the upstream/midstream and downstream zones were set at 880 °C and 580 °C, respectively. The temperatures were maintained for 6 h and were cooled naturally to room temperature. For the growth of the tungsten oxide nanowires, several substrates can be put into a furnace for CVD, and the amount of our synthesized tungsten oxide nanowires was approximately 15 ± 3 µg on a substrate (area: 1 cm × 2 cm).

The as-synthesized tungsten oxide nanowires were converted into WS2 nanotubes via typical conversion mechanisms reported in the literature51,52,53. S atoms were trapped at the W18O49-nanowire oxygen vacancies, and the nanowires were continuously converted to amorphous WS3 with a transition to the 2H–WS2 phase53. Details of the experimental setup are provided below. The sulfurization of the obtained nanowire samples was carried out using two tube furnaces, as in previous studies28,29. To prevent undesirable side reactions of Si substrates during sulfurization, the as-prepared nanowire samples were dispersed in ethanol and transferred onto Si/ SiO2 or quartz substrates. A ceramic boat with the nanowire samples on the substrates, preferentially annealed at 400 °C under high vacuum, was placed in a quartz tube and positioned downstream of the furnaces, and a ceramic boat with sulfur powder (99.99%, NewMet) was positioned upstream of the furnace. Under Ar flow (200 sccm), the furnace temperature for the nanowire samples was raised to a set temperature of 500–650 °C. After the furnace temperature reached the set temperature, the furnace temperature for S powder was raised to 250 °C, and S vapor was supplied. The furnace temperature was maintained for 1 h and then rapidly cooled to room temperature.

Structural characterization

Structures and elemental compositions of the tungsten oxide nanowires and the as-prepared small-diameter WS2 nanotubes were evaluated by TEM with EDS (JEM-3200FS, JEM2100F, and JEM2010F, JEOL Ltd.), FESEM (JSM-7100F JEOL Ltd.), XRD (Rigaku SmartLab) with Cu Kα1 (1.5406 Å) radiation, XPS (JPS-9010, JEOL Ltd.) and Raman spectroscopy (532 nm excitation, WItec).

Theoretical calculation

All calculations were conducted using the STATE program package based on density functional theory54,55. The exchange–correlation potential of the interacting electrons was treated by the generalized gradient approximation with the Perdew–Burke–Ernzerhof functional56. An ultrasoft pseudopotential was used to describe the interaction between the valence electrons and ions57. The valence wave functions and deficit charge density were expanded in terms of plane-wave basis sets with cutoff energies of 25 and 225 Ry, respectively. Brillouin-zone integration was performed with 9 × 9 and 9 × 1 k-meshes for planar and tubular WS2, respectively. Atomic coordinates of planar and tubular WS2 were optimized until the force was less than 1.33 × 10−3 Hartree/bohr.