Introduction

One-dimensional (1D) nanotubes derived from van der Waals (vdW) layered quantum materials have attracted considerable attention in nanomaterial research, owing to their distinct physical and chemical properties compared to their two-dimensional (2D) counterparts1,2,3,4,5,6,7,8. Transition metal dichalcogenide (TMD) nanotubes, as an emerging class of vdW nanotubes, possess non-centrosymmetric lattices and inherent lattice strain, resulting in highly polarized structures characterized by broken inversion symmetry9. These structures exhibit intriguing electrical and optical properties, thereby offering broad application prospects in fields such as superconductivity10, intrinsic photovoltaic effects11, and catalysis12. In comparison to carbon nanotubes, TMD nanotubes possess potentially more intriguing physical and chemical properties due to their richer elemental composition and tunable bandgap. However, research in this area remains limited, likely due to the challenges associated with their synthesis. In contrast to the facile synthesis of planar TMDs, transforming their layered chalcogen-metal-chalcogen structures into tubular forms requires overcoming more elastic energy, and the structural stability of few-layer products needs to be considered13. Owing to the influence of weak vdW forces between layers, vdW layered materials are more prone to forming hollow nanostructures (such as nanotubes or nanoscrolls) rather than tightly bonded solid nanowires when synthesized into 1D structures through chemical reactions. Tubular TMD nanostructures are typically synthesized via a solid–gas method under high-temperature conditions. Similar to the synthesis of carbon nanotubes, using metallic nanoparticles as catalysts is an effective approach for producing TMD nanotubes14,15. However, a significant drawback of the catalyzed chemical vapor deposition (CVD) process is the presence of residual metallic nanoparticle impurities embedded within the nanotubes, which degrade their electronic properties by limiting electron mobility. As an alternative, sulfurization of precursors leveraging the Kirkendall effect has also proven to be an effective method for forming tubular TMD structures16. Although significant progress has been made in developing template-based tubular materials using nanowhisker precursors17, challenges remain due to the complexities and stability issues associated with precursor selection.

Recent studies have reported the synthesis of tubular TMD materials using metal suboxides, particularly tungsten suboxides18,19. However, due to factors such as stability, suitable molybdenum suboxides for use as precursors are relatively scarce. Herein, we report an efficient strategy for the direct synthesis of 1D MoS2 nanotubes with high crystallinity and rhombohedral (3R) interlayer stacking. The method involves the introduction of hydrogen gas during a controlled CVD process. Typically, sulfurization of precursors using the Kirkendall effect requires the precursor to be prepared as 1D whisker-like nanomaterials in advance20. We controlled the participation time of the volatile source in the reaction by placing it in different temperature zones. First, MoO3 was reduced to tetragonal MoO2 (I4/m group) under the fluxing action of Te and the reducing action of hydrogen gas, forming 1D nanowires through a VLS growth mode. Subsequently, sulfur vapor participated in the reaction to sulfurize the MoO2 nanowires, eventually forming few-layer MoS2 nanotubes with a high aspect ratio. We conducted a detailed study of the entire reaction process and growth mechanism. Compared to 2D MoS2, the 1D tubular MoS2 exhibits strong symmetry-breaking effects, leading to promising performance in the bulk photovoltaic effect (BPVE), which has the potential to study non-centrosymmetric quantum phenomena in fundamental physics21,22.

Results

Direct growth and morphology control of MoS2 nanotubes

The hydrogen-assisted synthesis of MoS2 nanotubes is schematically illustrated in Fig. 1a. The growth temperature setting has been divided into four stages (Fig. 1b). As shown in Fig. 1c, in the heating stage, the MoO3 powder begins to melt and volatilize under the fluxing action of Te powder23,24, and nucleates as a Mo–O–Te liquid phase on the SiO2/Si substrate. As the saturated vapor pressure of MoO3 increases, MoO2 nanowires begin to precipitate from the liquid nucleation points under the reducing action of hydrogen gas. This process is a typical VLS growth model25. Furthermore, as the temperature increases, the concentration of sulfur vapor begins to rise, reacting with hydrogen to form H2S. The sulfidation of MoO2 nanowires is initiated, eventually resulting in the formation of hollow MoS2 nanotubes via the Kirkendall effect16. The schematic diagram of the modified CVD equipment for growing MoS2 nanotubes is shown in Supplementary Fig. 1. All reactants are placed in a smaller quartz tube with one end sealed, arranged according to different temperature zones. Our simulation results indicate that the internal quartz ampule plays a crucial role in decreasing the rate of gas flow (Supplementary Fig. 2)26. This reduction ensures that atoms are deposited and undergo reactions mainly through diffusion, thereby maintaining an adequate concentration of reactants. MoO3 powder is placed in the highest temperature zone of the tube furnace, facilitating the volatilization reaction. In our synthesis process, Te plays a pivotal but non-structural role as a fluxing promoter. The introduction of Te significantly reduces the melting point of MoO3 by forming a low-melting eutectic phase or volatile Mo-Te intermediates. This assists the formation of 1D tetragonal MoO2 nanowires via a VLS mechanism under hydrogen reduction. Without Te, MoO3 tends to sublimate unevenly and fails to produce uniform nanowire structures, as confirmed by our control experiments (Supplementary Table 1). During the sulfurization stage, Te has already evaporated or been carried away by carrier gas, and our XPS and EDX analyses (Supplementary Fig. 3) confirm the absence of Te in the final MoS2 nanotubes. Therefore, Te acts exclusively as a reaction mediator and fluxing agent, not as a dopant or structural constituent. This behavior is distinct from systems like MoTe2, where Te is a necessary chalcogen source and integrates into the final lattice27,28. Here, Te enables the initial MoO2 nanowire formation, which is essential for the subsequent symmetry-directed transformation into hollow MoS2 nanotubes via the Kirkendall effect23,29. In our method, the MoS2 nanotubes can be well synthesized under a reaction temperature range of 750 to 800 °C, with an optimal temperature of 770 °C (Supplementary Fig. 4). The other details of the growth conditions are shown in the “Methods” section.

Fig. 1: Hydrogen-assisted synthesis of MoS2 nanotubes and the control of product morphology.
figure 1

a Schematic of the synthesis of MoS2 nanotubes by CVD. b Temperature profile for synthesizing MoS2 nanotubes, including Stage Ι: heating; Stage ΙΙ: growth; Stage ΙΙΙ: sulfurization; and Stage ΙV: cooling. Corresponding schematic diagrams of the products (MoO2 nanowires, MoS2–MoO2 core-shell structures, and MoS2 nanotubes, respectively) at different stages. c Proposed growth mechanism of MoS2 nanotubes. The process begins with nucleation and formation of tetragonal MoO2 via a VLS growth mechanism, which is subsequently transformed into MoS2 nanotubes through sulfidation. d Schematic illustration of the morphological changes of products under different H2 concentrations. e–h Optical microscopy images showing the morphology evolution from 2D MoS2 flakes to MoS2 nanotubes with increasing H2 concentration from 0% to 30%, the inset in (h) shows a TEM image of a single MoS2 nanotube. Scale bar: 40 µm in (e–h), and 50 nm in the inset of (h). Source data are provided as a Source Data file.

Full size image

Furthermore, our studies have revealed that the morphology of the products can be regulated by controlling the concentration of H2 during the reaction process. As illustrated in Fig. 1d, in the absence of H2, under the thermodynamic growth mode, the MoO3 source ultimately undergoes epitaxial growth with sulfur to form 2D MoS230 (Fig. 1e). When a small amount of H2 (5%) is involved in the reaction process, the final product is a composite structure of 2D MoS2 film and MoS2 nanotubes (Fig. 1f). As the concentration of H2 further increases (15%), the underlying MoS2 film of the product begins to contract and forms 2D ribbon-like materials along the MoS2 nanotubes (Fig. 1g). When the proportion of hydrogen gas reaches 30% in the argon–hydrogen gas mixture, the final product is pure MoS2 nanotube (Fig. 1h). Interestingly, as shown in Supplementary Figs. 5 and 6, when we changed the growth substrate from SiO2/Si to f-mica, we found that the samples transformed from curved nanotubes to straight nanotubes, which were arranged at 0°, 60°, and 120° on the substrate. Meanwhile, we examined the curvature behavior of nanotubes with different lengths (Supplementary Fig. 7), the transformation clearly demonstrates the strong influence of substrate choice and nanotube length on their morphology and orientation: while longer nanotubes tend to bend on rigid substrates like SiO2/Si due to accumulated internal strain, they remain straight and well-aligned on atomically flat, lattice-matched substrates such as f-mica. This indicates that the substrate can effectively regulate the alignment of the product31.

Characterization of MoO2 nanowires in the sulfidation process

In our experimental design, we isolated tetragonal-phase MoO2, a key intermediate in the formation of MoS2 nanotubes. As shown in Supplementary Table 1, the sulfur source was removed during CVD synthesis. Optical and scanning electron microscope (SEM) images of the resulting products (Fig. 2a, b) reveal 1D nanomaterials with high aspect ratios, whose ends are decorated with nanoparticles. The growth of MoO2 nanowires follows a hydrogen- and Te-assisted VLS mechanism32,33. Under a reducing atmosphere, Te lowers the melting point of MoO3 and facilitates the formation of a liquid droplet that mediates axial nanowire elongation. As the local MoO3 vapor pressure increases, molybdenum species dissolve into the liquid droplet until supersaturation is reached. Upon supersaturation, MoO2 begins to precipitate from the droplet, and 1D nanowires emerge as the droplet rolls forward along the substrate. After growth terminates, the droplet solidifies as a residual nanoparticle at the nanowire tip. This behavior is consistent with a typical VLS process. SEM-EDX analysis of MoO2 nanowires with terminal nanoparticles (Supplementary Fig. 8) confirms that Te plays a catalytic yet non-reactive role during nanowire formation. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image and energy-dispersive X-ray spectroscopy (EDS) mappings (Fig. 2c) confirm that the 1D structures are solid, uniform nanowires with evenly distributed molybdenum and oxygen elements. The atomic structure in the (001) and (100) directions is shown in Fig. 2d, e, These clearly demonstrate that the material is a tetragonal phase of MoO2 with the space group I4/m. Cross-sectional atomic-resolution STEM imaging of a MoO2 nanowire (Fig. 2f) reveals a high degree of symmetry, consistent with a tetragonal-phase single crystal viewed along the (001) direction. Figure 2g is an atomic-resolution STEM image along the (111) direction, revealing two distinct sets of lattice fringes with average spacings of 0.75 nm and 0.27 nm, corresponding to the (\(\bar{1}\)10) and (\(\bar{1}\)2\(\bar{1}\)) planes of the tetragonal phase MoO2. Furthermore, Raman spectroscopy was performed on MoO2 nanowires (Supplementary Fig. 9), and DFT calculations were conducted to analyze their vibrational modes in the idealized bulk state. Comparison between theory and experiment revealed three matching vibrational peaks at 152 cm−1, 185 cm−1, and 295 cm−1. The remaining discrepancies are likely due to the rod-like morphology of the sample, which deviates from the ideal bulk crystal and alters its vibrational response under laser excitation.

Fig. 2: Microscopy analysis of MoO2 nanowires.
figure 2

a, b Optical and SEM images showing MoO2 nanowires grown on SiO2/Si substrate. Terminal particles attached to the nanowires are highlighted with circles. c HAADF-STEM image and corresponding EDS elemental mappings of a single MoO2 nanowire. d, e Atomic structure of tetragonal MoO2 (viewed from c-axis and a-axis). f Cross-sectional STEM image of Tetragonal MoO2 nanowires. The inset in (f) displays the corresponding fast Fourier transform pattern. g STEM image of tetragonal MoO2 from the (111) direction. The inset shows the corresponding FFT pattern. h Cross-sectional STEM image of the initial stage of the sulfidation process of MoO2 nanowires. i, j HAADF-STEM image of the initial stage of the sulfidation process from zone axis [113]. And a magnified view of the region denoted by a white dashed square shows the synthesized MoS2 stacked in a rhombohedral stacking manner. Scale bar: 10 μm in (a), 5 μm in (b), 50 nm in (c), 1 nm in (f, g), 5 nm−1 in insets of (f, g), 20 nm in (h), 10 nm in (i), 1 nm in (j).

Full size image

The Kirkendall effect enables a solid cylindrical structure to transform into a hollow morphology via diffusion and lattice rearrangement16,19. As the temperature in the sulfur-source zone increases, the corresponding rise in sulfur vapor pressure initiates its participation in the reaction. Assisted by H2, the MoO2 nanowires undergo sulfidation. This transformative phase culminates in the formation of MoS2 nanotubes. By controlling the reaction time, we obtained samples in the initial stage of sulfidation (1 min). Figure 2h presents a representative cross-section STEM image of MoO2 nanowire in the initial stage of sulfidation, showing a MoS2 shell enveloping the MoO2 core. Observations along the (113) crystallographic direction (Fig. 2i, j) show that the MoS2 layers on the MoO2 surface exhibit 3R stacking.

Taken together, we propose that the formation of MoS2 nanotubes in our system likely proceeds via a hybrid mechanism, summarized as follows: Following the growth of 1D tetragonal-phase MoO2 nanowires through a VLS mechanism, the introduction of sulfur vapor at elevated concentrations, combined with the presence of hydrogen, initiates the sulfidation process. In the early stages, the outer shell of the MoO2 nanowire undergoes sulfidation first. This transformation induces lattice distortions and local defect formation, which in turn trigger the onset of the hollowing process. These structural defects, particularly near the tube tip (Supplementary Fig. 10), facilitate the inward diffusion of reactive gases (S/H2), enabling continued sulfidation of the inner MoO2 core. Concurrently, the presence of these tip defects also allows for partial Mo loss through evaporation. In addition, volatilized MoO2 can migrate into the nanotube cavity, where it re-deposits onto the inner surface and reacts with sulfur to form additional MoS2 layers, contributing to inner wall growth. As the internal MoO2 core becomes fully consumed, this process results in the formation of multi-walled, hollow MoS2 nanotubes.

Characterizations of MoS2 nanotubes

We characterized the fully sulfurized MoS2 nanotubes, the nanotubes synthesized on SiO2/Si substrate exhibit an average density of ≈51 nanotubes per 104 μm2 (Supplementary Fig. 11). As shown in the characterization diagrams (Fig. 3a–c) of the MoS2 nanotubes, the structures exhibit a high aspect ratio exceeding 1000 and an average diameter of ≈51 nm. STEM analysis of 50 samples indicates that the inner diameters are centered around 37 nm, with an average wall number of ≈13 layers (Supplementary Fig. 12). While precise control over the geometric structure is still limited, we are exploring sulfurization of pre-selected MoO2 nanowires to improve uniformity, which shows promise for future optimization. HAADF-STEM and EDX mapping of fully sulfurized MoS2 nanotubes (Supplementary Fig. 13) show Mo and S signals, although accurate quantification is limited by Mo-L and S–K peak overlap. Oxygen is confined to the terminal nanoparticle, likely due to incomplete sulfurization caused by their larger size. Trace Te (0.37%) is detected, indicating minimal incorporation. These results confirm the transformation of MoO2 into MoS2 nanotubes. Raman intensity mapping of \({E}_{2g}^{1}\) (380 cm−1) and \({A}_{1g}\) (405 cm−1) with a mapping size of 2 µm × 8 µm area (Fig. 3d, e) indicates high spatial uniformity of the MoS2 nanotubes. To probe the presence of sulfur vacancies and curvature-induced effects in the MoS2 nanotubes, we also performed resonance Raman spectroscopy using a 633 nm excitation wavelength and compared the results with monolayer MoS2 (Supplementary Fig. 14). The results suggest that while the nanotube geometry modifies phonon selection rules through structural anisotropy and strain, the intrinsic defect density remains low. Atomic force microscopy (AFM) testing indicated that the sample diameter is around 50 nm (Fig. 3f). Figure 3g shows a schematic diagram of a multilayer MoS2 nanotube. The TEM image (Fig. 3h) shows that the MoS2 nanotubes are uniform and multi-walled, a feature attributed to the high-temperature synthesis method employed. The HAADF–STEM image (Fig. 3i) reveals that the MoS2 nanotube exhibits high crystallinity. We conducted a magnified analysis of the atomic structure in the wall region (the white dashed box marked as 1 in Fig. 3i) and the axial central region (the white dashed box marked as 2 in Fig. 3i). Figure 3j displays the arrangement of atomic columns in S–Mo–S chevrons located at the edge of a MoS2 nanotube. This arrangement corresponds to the 3R phase of MoS234, as demonstrated by its alignment with the overlaid atomic structure model. The atomic arrangement in the middle of the nanotube (Fig. 3k) presents a periodic honeycomb structure along the axial direction of the nanotube, as overlaid and aligned with the atomic structure model. Additionally, a typical 3R-phase stacking pattern is observed on the tube wall (Supplementary Fig. 15). We conducted FFT analysis of atomic-resolution STEM images along [001] and [010] directions (Supplementary Fig. 16). The resulting patterns match well with simulated 3R-MoS2 diffraction, showing strong (003) reflections and absence of 2H spots, confirming the 3R stacking order, consistent with ref. 35. These results indicate that MoO2 nanowires will form 3R phase MoS2 nanotubes after sulfidation in a hydrogen-assisted environment.

Fig. 3: Characterization of MoS2 nanotubes.
figure 3

a Bright optical image of MoS2 nanotubes grown on SiO2/Si substrate. b SEM image shows that the MoS2 nanotubes have a large aspect ratio. c Violin plot of nanotube diameters (n = 80), with mean (54.5 ± 15.4 nm) and median (51.8 nm) indicated. d, e Raman intensity mapping for \({{{ E}}}_{{{2}}{\boldsymbol{g}}}^{{{1}}}\) (380 cm−1) and \({{{A}}}_{{{1}}{{g}}}\) (405 cm−1), respectively. f AFM image illustrating that pure MoS2 nanotubes can be obtained under appropriate H2 concentration ratios. g Structural characterization of multiwall MoS2 nanotubes. h Bright field TEM image of completely sulfurized MoS2 nanotube. i HAADF-STEM image of multiwall MoS2 nanotube. Inset: fast Fourier transform pattern in the central region of the nanotube. j, k Atomic resolution HAADF-STEM image of the area denoted by white dashed squares 1 and 2 in panel e, revealing the characteristic arrangement of the rhombohedral symmetry (3R) phase, this pattern is superimposed and aligned with the atomic structure model for detailed comparison. Scale bar: 40 μm in (a), 10 μm in (b), and 100 nm in the inset of (b), 1 μm in (d, e), 100 nm in (f), 20 nm in (h), 5 nm in (i), and 1 nm−1 in the inset of (i), 0.5 nm in (j, k). Source data are provided as a Source Data file.

Full size image

Optical properties and BPVE

The optical response of low-dimensional materials often reflects their structural characteristics. We compared the Raman spectra, photoluminescence (PL) signals, and polarization-dependent second harmonic signals (SHG) of 2D monolayer MoS2 and MoS2 nanotubes. As shown in Fig. 4a, the \({E}_{2g}^{1}\) vibrational mode of tubular MoS2 exhibits a significant red shift from 384.5 cm−1 in monolayer MoS2 to 375.8 cm−1 in the largest diameter nanotube, indicating a total shift of ≈8.7 cm−1. This redshift, plotted against nanotube diameter in Fig. 4b, increases with curvature and wall number, and is attributed to in-plane tensile strain induced by the tubular geometry. In the rolled-up structure, tensile stress dominates due to the larger volume of outer layers, leading to Mo–S bond elongation and softening of the in-plane phonon mode36,37. This further implies the existence of intrinsic in-plane polarization in MoS2 nanotubes. PL test results (Supplementary Fig. 17) indicate that the PL signal is strictly distributed along the axial direction of the nanotubes, with no significant PL signals observed in the surrounding areas. This confirms that the PL originates from the nanotubes themselves. The PL spectrum shows a peak at 687 nm, corresponding to the A exciton transition. Compared with the PL peak of monolayer MoS2 (≈670 nm), a significant redshift is observed. Under the same testing conditions, the PL intensity of the nanotubes is significantly lower compared to that of monolayer MoS234. We used DFT to calculate the change in the bandgap of MoS2 nanotubes with varying curvature and number of layers (Supplementary Fig. 18). The results show that with the increase in the curvature of the nanotubes and the number of layers, the valence band top and the conduction band bottom gradually converge38. We also verified this in the transfer characteristic tests of the MoS2 nanotube field effect transistor (Supplementary Fig. 19), and the result is consistent with the electrical properties of similarly structured TMD materials reported in refs. 34,39. Nonlinear optics can effectively analyze the symmetry of materials. The polarization-dependent SHG signal of 2D monolayer MoS2 exhibits a classic six-fold symmetry pattern40 (Supplementary Fig. 20), while the polarization-dependent SHG signal (Fig. 4c) of 1D tubular MoS2 shows an anisotropic nonlinear optical response41. These significant differences in optical responses indicate a greater symmetry breaking in the MoS2 nanotube compared to planar MoS2.

Fig. 4: Optical properties and bulk photovoltaic response of MoS2 nanotubes.
figure 4

a Raman spectrum of MoS2 nanotubes and monolayer MoS2 flake, showing a curvature-induced redshift of the \({{{ E}}}_{{{2}}{{ g}}}^{{{1}}}\) mode. b Evolution of \({{{E}}}_{{{2}}{{ g}}}^{{{1}}}\) peak position and \({{{ E}}}_{{{2}}{{ g}}}^{{{1}}}/{{{ A}}}_{{{1}}{{ g}}}\) intensity ratio across multiple samples, indicating an increase in in-plane strain and the number of walls. c Polarization-dependent SHG signals of MoS2 nanotubes. d DFT-calculated polarization distributions in MoS2 nanotubes with varying curvature and wall numbers. ΔPc = 11.47 eÅ/Mo48S96—representing the polarization enhancement from curvature; ΔPw = 0.04344 eÅ/Mo72S144—the additional contribution from increased wall number. e I–V characteristics of few-layer 3R MoS2 nanofilm, showing typical photoconductive behavior. f I–V characteristics are recorded at different illumination intensities. The laser wavelength is 635 nm. g Schematic diagram of MoS2 nanotube BPVE device. h The dependence of Isc on the position of the laser spot in a MoS2 nanotube device. The main response occurs when the laser spot illuminates the center of the device away from the contacts. i Isc was measured during on/off cycles of a 635 nm laser, with three different power levels. The duration of one cycle is 10 s. j Incident power density dependence of jsc for a BPVE device with a crossover from a linear to a square root relationship. Scale bar: 10 μm in the inset of (e), 5 μm in the inset of (h, j). Source data are provided as a Source Data file.

Full size image

As shown in Fig. 4d, first-principles DFT calculations on MoS2 nanotubes show that curvature-induced lattice distortion is the dominant source of spontaneous polarization (ΔPc ≈ 11.5 eÅ/Mo48S96), which is derived by comparing the polarization of a larger single-walled Mo48S96 nanotube with that of two smaller single-walled Mo24S48 nanotubes. Although the total atomic content is equivalent, the decrease in curvature in the larger tube leads to a net reduction in polarization, revealing the substantial role of geometric curvature in breaking inversion symmetry. In contrast, adding extra walls contributes only a minor incremental term (ΔPw ≈ 0.04 eÅ/Mo72S144). This is obtained by subtracting the combined polarization of a Mo48S96 tube and a Mo24S48 tube (representing the inner and outer walls) from that of the double-walled Mo72S144 structure. This small difference indicates that wall-number-induced polarization enhancement is significantly suppressed due to interlayer screening and saturation effects.

The BPVE of tubular TMD materials has so far only been reported in WS2 nanotubes11. To investigate the BPVE of tubular MoS2 materials, we fabricated devices based on MoS2 nanotubes. Firstly, we prepared a device based on a 5-layer 3R-phase MoS2 film (Supplementary Fig. 21) and tested its response to laser illumination (Fig. 4e). The result shows that the MoS2 flake only exhibits a photoconductive effect with a negligible short-circuit (Isc) under illumination. Figure 4f presents the current-voltage (I–V) characteristics as a function of varying laser power levels (Plaser), illustrating that both the Isc and the open-circuit voltage (Voc) exhibit a monotonic variation in response to Plaser. The BPVE in MoS2 nanotubes demonstrates considerable stability, both qualitatively and quantitatively. Figure 4g illustrates a simple MoS2 nanotube photovoltaic device along with the schematic diagram of the test. Besides, we measured the Isc as the laser (532 nm) spot was scanned from one electrode to another. Figure 4h indicates that the Isc of the MoS2 nanotube device is strongest when the laser is irradiated at the middle of the nanotubes. This demonstrates the BPVE of MoS2 nanotubes, rather than the traditional Schottky-junction photovoltaic effect. As depicted in Fig. 4i, the Isc undergoes alterations across numerous laser on-off cycles. Under 1.88 × 104 W cm−2 of illumination, the photocurrent can reach 510 A cm−2 (the diameter of the nanotube is 50 nm). Isc consistently reverts to a comparable value upon the reactivation of laser illumination. Power-dependent photocurrent was measured to further clarify the physical mechanism of the BPVE observed in MoS2 nanotubes. As shown in Fig. 4j, the photocurrent in nanotubes exhibits a transition from a linear to a square-root dependence on power as the power increases, aligning with the predictions of the theoretical shift current model and corroborating previous experimental findings. In the BPVE tests of other devices (Supplementary Fig. 22), the Isc increases with the diameter of the nanotubes (indicating an increase in total polarization), as larger nanotubes inevitably possess more concentric walls. The cumulative dipolar fields from these walls, together with reduced quantum-confinement resistance, enhance carrier separation and transport, thereby outweighing the loss of curvature. Hence, curvature governs the intrinsic lattice polarization, whereas wall number and inter-wall coupling determine how that polarization translates into efficient bulk photovoltaic shift currents in practical MoS2 nanotube devices42. These results indicate that MoS2 nanotube structures have potential value for fundamental physical research.

Discussion

In this study, we have developed an efficient hydrogen-assisted CVD method for the direct synthesis of MoS2 nanotubes with controlled crystallinity and structural uniformity. By leveraging the VLS growth process and carefully controlling the hydrogen concentration, we achieved significant control over the morphology of the final products, ranging from 2D MoS2 films to pure MoS2 nanotubes. The transformation from MoO2 to MoS2 nanotubes was driven by the Kirkendall effect, and the resulting nanotubes exhibited a high aspect ratio and 3R stacking. Our detailed structural and optical characterizations revealed the distinct properties of MoS2 nanotubes, including strong polarization, strain-induced redshifts in Raman and PL spectra, and symmetry breaking compared to planar 2D MoS2. Furthermore, we demonstrated the BPVE in MoS2 nanotubes, providing a platform for investigating symmetry-breaking phenomena in fundamental physics. These findings offer insights into the growth mechanisms of TMD nanotubes and indicate their relevance for advanced energy and optoelectronic technologies. Future research should focus on optimizing the synthesis process and exploring additional functionalities of MoS2 nanotubes in practical device architectures.

Methods

Growth of MoS2 nanotubes

The MoS2 nanotubes were synthesized by atmospheric pressure chemical vapor deposition (APCVD). A sufficient amount of sublimed sulfur powder (Aladdin, 99.95%) was placed at the sealed end of an inner quartz tube with a diameter of 12 mm and a length of 320 mm. At 60 mm from the sulfur source, 0.05 mg of Te powder was placed in a quartz boat. A piece of SiO2/Si substrate with a 285 nm oxidation layer is placed over a quartz boat measuring 20 mm in length, 8 mm in width, and 4 mm in height, which contains 0.2 mg of MoO3 powder (Aladdin, 99.999%). The inner quartz tube was placed into a 2-inch quartz tube. Prior to heating, the system was purged with argon gas for 5 min. Argon and hydrogen gases were introduced as carrier and reducing gases at flow rates of 70 cm3 STP min−1 and 30 cm3 STP min−1, respectively. The quartz tubes were heated evenly to 770 °C in 25 min, held for 20 min, and then naturally cooled to room temperature. Finally, the MoS2 nanotubes are obtained on SiO2/Si and mica substrates.

Synthesis of MoO2 nanowires

Like the preparation of MoS2 nanotubes, MoO2 nanowires were synthesized as an intermediate material. The sole difference is that the sulfur source has been removed from the quartz tube within the furnace, resulting in the formation of tetragonal-phase MoO2 nanowires on SiO2/Si substrates.

Obtaining samples at the initial stage of sulfurization

During the synthesis of MoS2 nanotubes, the reaction was deliberately interrupted as soon as the temperature of the tube furnace reached 770 °C, followed by rapid cooling to room temperature. The resulting sample, representing the initial stage of sulfurization, consists of a MoO2–MoS2 composite. This intermediate product enables the investigation of the sulfurization transformation pathway of the material.

Characterizations

Optical images were acquired with a ZEISS Primotech microscope. X-ray photoelectron spectroscopy (XPS, Kratos AXIS Ultra DLD) is carried out using an Al Kα (h\(\upsilon\)=1486.6 eV) source with an operating power of 150 W, and the analysis area is 700 μm × 300 μm. SEM images were obtained by a Field Emission Scanning Electron Microscope (FESEM, Gemini SEM 300). Raman and PL spectra were carried out using an excitation laser of 532 nm from a WITEC alpha 300R Raman system. Resonance Raman measurements- in which the excitation wavelength is tuned to match an electronic transition of MoS2, thereby significantly enhancing defect-sensitive vibrational modes- were performed using a Raman spectroscopic system (Renishaw, InVia) equipped with a 633 nm wavelength excitation laser. AFM topography measurements were conducted using an Asylum Research Cypher S system. The PMMA-assisted wet-transfer method was used to prepare STEM samples, and the cross-section samples were prepared using Zeiss Crossbeam 540 focused ion beam (FIB) with a gallium ion source. TEM was performed using FEI Themis Z transmission electron microscope operated at 300 kV. Atomic-resolution HAADF-STEM images were conducted on an aberration-corrected JEOL ARM-200CF (S)TEM equipped with a cold field-emission gun, operating at 200 kV, and an Advanced STEM Corrector (ASCOR) probe corrector. The beam convergence semi-angle is 28 mrad, and the HAADF collection angle is 65-280 mrad. The EDX spectra were collected using an Oxford X-Max TLE 100 mm2 windowless SDD detector. The SHG characteristics of the MoS2 nanotubes were evaluated using a home-built vertical microscope setup with the reflection geometry. A fiber pulse laser with a central wavelength of 1558 nm (PriTel Inc., with a pulse width of 8.8 ps and repetition rate of 18.8 MHz) was used as the fundamental pump source. The pulse laser was focused using the 50× objective (with a numerical aperture of 0.75), and the focused laser spot is about 2 μm. The SHG signal scattered from the MoS2 nanotubes was collected using the same objective lens and examined using a spectrometer (Princeton Instruments, SP2558 and 100BRX) equipped with a cooled silicon CCD camera. To investigate the polarization dependence of the SH radiation, a circularly polarized pump laser was passed through a rotating polarizer to achieve linear polarizations along different directions. Another polarizer was then placed in the signal collection path and rotated to match the pump polarization to gather the parallel components of the SHG signal.

Device fabrication and measurement

The SiO2/Si substrate bearing the as-grown MoS2 nanotubes was first spin-coated with LOR 5 A (Nippon Kayaku, USA), followed by S1805 (Merck, Germany). Micro-Writer (ML3 Serial) was then used to define the electrode patterns. After developing the patterns in the developer (Micro-posit MF-319), electron-beam evaporation and the lift-off process were adopted to deposit the metal contacts. The electrical measurements were carried out in a probe station connected to a semiconductor parameter analyzer (PDA FS380 Pro, Platform Design Automation).

Simulation methods

Computational details for Raman-mode calculations of tetragonal MoO2 nanowires: The calculations of crystal structure optimization of tetragonal MoO2 (I4/m) and vibrational mode analysis were performed using the VASP43 software package with the PBE functional44 and an energy cutoff of 400 eV. The lattice parameters were a = 10.623 Å, b = 10.623 Å, c = 3.264 Å. A Γ-centered k-point mesh of 1 × 1 × 4 was employed. The vibrational eigenvectors were projected onto the atomic coordinates and visualized with VESTA45.

Computational details for electronic structure and curvature-induced polarization of MoS2 nanotubes: First-principles DFT calculations were performed using the Vienna Ab initio Simulation Package (VASP)43. The exchange-correlation energy was treated within the Perdew–Burke–Ernzerhof (PBE) generalized gradient approximation44. The plane-wave basis set with a kinetic energy cutoff of 500 eV was employed. For the geometry optimization and static self-consistent calculations, the Γ-center scheme was sampled with a 12 × 1 × 1 grid within the Brillouin zone. The structure relaxation considering the atomic positions was performed by the conjugate gradient (CG) scheme until the maximum force on each atom was less than 10−3 eV Å−1, and the total energy was converged to 10−6 eV using the Gaussian smearing method.

Photovoltaic response measurements

Electrical measurements, including I–V characterizations and photocurrent responses, were performed using a semiconductor device analyzer (PDA FS380 Pro, Platform Design Automation) under vacuum conditions. MoS2 nanotubes were illuminated by an excitation laser (wavelength of 635 nm) with a spot size of ≈2.5 μm. The line-scan mapping of photocurrent was carried out using a laser of 532 nm by the 100× objective lens from a WITEC alpha 300R Raman system under ambient conditions. The laser spot size was ≈1 μm.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.