WS2 Nanotubes, 2D Nanomeshes, and 2D In-Plane Films through One Single Chemical Vapor Deposition Route

We demonstrate a versatile, catalyst free chemical vapor deposition process on insulating substrates capable of producing in one single stream one-dimensional (1D) WO3–x suboxides leading to a wide range of substrate-supported 2H-WS2 polymorphs: a tunable class of out-of-plane (of the substrate) nanophases, with 1D nanotubes and a pure WS2, two-dimensional (2D) nanomesh (defined as a network of webbed, micron-size, few-layer 2D sheets) at its extremes; and in-plane (parallel to the substrate) mono- and few-layer 2D domains. This entails a two-stage approach in which the 2WO3 + 7S → 2WS2 + 3SO2 reaction is intentionally decoupled. First, various morphologies of nanowires or nanorods of high stoichiometry, WO2.92/WO2.9 suboxides (belonging to the class of Magnéli phases) were formed, followed by their sulfurization to undergo reduction to the aforementioned WS2 polymorphs. The continuous transition of WS2 from nanotubes to the out-of-plane 2D nanomesh, via intermediary, mixed 1D-2D phases, delivers tunable functional properties, for example, linear and nonlinear optical properties, such as reflectivity (linked to optical excitations in the material), and second harmonic generation (SHG) and onset of saturable absorption. The SHG effect is very strong across the entire tunable class of WS2 nanomaterials, weakest in nanotubes, and strongest in the 2D nanomesh. Furthermore, a mechanism via suboxide (WO3–x) intermediate as a possible path to 2D domain growth is demonstrated. 2D, in-plane WS2 domains grow via “self-seeding and feeding” where short WO2.92/WO2.9 nanorods provide both the nucleation sites and the precursor feedstock. Understanding the reaction path (here, in the W–O–S space) is an emerging approach toward controlling the nucleation, growth, and morphology of 2D domains and films of transition-metal dichalcogenides.


CVD set-up for nanostructure synthesis and phase diagram for WO 3-x phases.
Figure S1. CVD set-up for the LPCVD/APCVD growth of nanostructures shown in this work. (a) Typical positioning of boats containing the WO 3 and S precursors, as well as the SiO 2 /Si substrates, all located inside the furnace. A typical temperature profile inside the furnace is shown, highlighting a central zone of constant temperature, and a zone with a temperature gradient where the SiO 2 /Si substrates were placed. An additional heating belt was used for sublimating the S powder placed at the entry of furnace. A magnetically-driven manipulator connected to the quartz rod on which the boats with materials and samples are mounted allowed the boats to be easily moved and positioned in different parts of the furnace; or removed from the high temperature zones at the end of the growth. (b) Phase diagram showing how the distance to the WO 3 precursor and temperature at the substrate location determines the nanophase and morphology (i.e. out-of-plane long nanowires, short in-plane nanorods, or short nanorod bundles) of the grown WO 3-x nanostructures. This relates to the positioning of the boats with substrates in the temperature gradient region of the furnace.  peaks were also present in both XRD spectra (b) and (f).
The peaks identified in the XRD spectra from Figures S2 (b) and (f) are listed in Tables S1 and S2 below. d is the periodic spacing corresponding to the identified (hkl) plane, while I, the peak intensity relative to the most intense peak.

Raman characterization of WS 2 nanophases obtained in this work.
The Raman spectrum of 2H-WS 2 taken at 532 nm laser excitation, as here, is very rich, revealing many second-order peaks that are stronger than those observed at shorter wavelength excitation, e.g. λ exc = 488 nm. 1 Hence, up to fifteen Raman modes of WS 2 could be present, 2 requiring peak deconvolution, as shown in Figure S4. However, usually, for 2H-WS 2 identification, the E 2 1 (in-plane) and A 1g (out-ofplane) vibrational modes are used as fingerprints. 3 Furthermore, a strong 2LA mode, caused by double resonance scattering, 4 can also be seen close to E 2 1 ( Figure S3).  Moreover, the spectral difference (∆ω) between the E 2 1 and A 1g modes can be used for identification of 2H-WS 2 film thickness. [3][4][5] Figure S4 shows such an analysis that complements Figure 4(g), main text. In Figure S4 (a), the in-plane E 2 1 mode is around 355 cm −1 , while the out-of-plane A 1g mode is around 417 cm −1 , the spectral distance ∆ω between being ~ 62 cm −1 ; this is reported as the fingerprint of monolayer 2H-WS 2 , 3 showing that the domains obtained in Figure 4(e-f) are monolayers. With Intensity (a.u.)

Raman Shift (cm -1 )
increasing the thickness of the 2H-WS 2 films, the E 2 1 mode red-shifts while the A 1g mode blue-shifts, 6 e.g. ∆ω =64 and 66 cm −1 were reported for bilayer and trilayer, respectively. 3,4 However, the frequencies of the these fingerprint modes tend to remain unchanged when the WS 2 thickness exceeds five layers. 6 Based on this, the spectral distance ∆ω = 67cm −1 obtained from Figure S4 (b) shows that the domains observed in Figure 4(c), main text, consist of at least five layers.

Electrical characterization of WS 2 nanomesh.
The sheet conductivity of the nanomesh shown in Figure 3(b) was measured using a four-point probe method. The four contacts, spaced by 2 mm each, were formed by electron-beam evaporation: first, 100 nm of chromium were deposited (chromium was reported to form ohmic contacts with WS 2 ), 7 followed by 100 nm of palladium. Large contacting pads were then bonded with 75 µm tinned copper wires using low temperature conductive epoxy (H21D).
A constant potential difference of 40 V was applied across the two outside contacts, leading to a voltage drop of 4.7 V across the two inner contacts, while the current was measured between the two outside contacts. The temperature was varied, leading to a current -temperature dependence from which a conduction mechanism was analyzed ( Figure S5). The four-point probe station used two Keithley 2400 SourceMeter as multimeter and voltage source, respectively. A Lakeshore 325 cryogenic temperature controller was used to control the temperature cooling-down rate of the sample mounted inside a Leybold RDK 10-320 closed-cycle Helium cryostat. Temperature was varied between room temperature and 210 K (at which the measured current reached the noise level), using a coolingdown step of 0.5 K. Ten measurements of current were averaged at every temperature step. thermally activated transport with an activation energy E a ≈ 0.33 eV; in the 245 -216 K range (down to the measurement noise level), no single transport mechanism describes uniquely the behavior. This alternative 4-probe method implementation was chosen due to the high resistance of the sample at low temperatures: using constant current injected through the external leads and measuring the voltage drop across the middle contacts -which is the more conventionally used 4 probe methodled to a large potential drop across the external contacts at low temperatures which could induce heating of the sample and artefacts in the form of non-linear behaviour.
Electrical transport mechanisms were investigated by measuring the nanomesh conductivity as a function of temperature: conductivity decreases with temperature indicating the WS 2 nanomesh is semiconducting ( Figure S5), as expected. Then, by plotting ln(σ) vs. T -1/n , where n is an integer, different characteristic carrier transport regimes can be identified. Figure S5 shows such a plot for n =1; a linear regime in this case indicates thermally activated transport (Arrhenius behavior), for which a thermal activation energy E a ≈ 0.33 eV can be extracted, from room temperature down to about 245 K, according to: Such a value of E a suggests that transport takes place via, most likely, a defect band located below/above conduction/valence band edges. Below this temperature and down to about 216K, where the measurements reach the noise level of the experimental set-up, no single transport mechanism (e.g. two-dimensional or three-dimensional variable range hopping, characterized by n = 3 or n = 4, respectively) could be identified, suggesting that this temperature range corresponds to a transition (mixed) regime. Further transport investigations will be reported elsewhere. W-W = 3.31 Å W-W = 3.15 Å Figure S6. Reaction mechanism describing layer-by-layer evolution from WO 3 to WS 2 via WO 2.92 /WO 2.9 Magnéli phases. Top panels: two unit cells of WO 2.9 (shown at two different orientations, with crystallographic a, b, and c axes indicated), highlighting the region of the crystallographic shear (CS) planes that contain edge-sharing octahedral units. Bottom panels: a transformation path from the CS region of WO 2.9 to 2H-WS 2 , via a 1T-like WS 2 phase (which structurally resembles to the CSregion of WO 2.9 , where S replaces O), is shown. The W-W intra-layer distance in the CS-plane regions of WO 2.9 , of 3.31 Å, is close to that in 2H-WS 2 , of 3.15 Å, and significantly closer than the corresponding W-W intra-layer distance in WO 3 , of 3.8 Å, hence, favoring conversion to WS 2 . Moreover, replacement of O by S leads to vertical unit cell expansion, to achieve the 6.2 Å (002) S-W-S basal plane spacing in 2H-WS 2 . Color coding: W (grey), O (red), and S (yellow).