Tungsten Ditelluride: a layered semimetal.

Tungsten ditelluride (WTe2) is a transition metal dichalcogenide (TMD) with physical and electronic properties that make it attractive for a variety of electronic applications. Although WTe2 has been studied for decades, its structure and electronic properties have only recently been correctly described. We experimentally and theoretically investigate the structure, dynamics and electronic properties of WTe2, and verify that WTe2 has its minimum energy configuration in a distorted 1T structure (Td structure), which results in metallic-like transport. Our findings unambiguously confirm the metallic nature of WTe2, introduce new information about the Raman modes of Td-WTe2, and demonstrate that Td-WTe2 is readily oxidized via environmental exposure. Finally, these findings confirm that, in its thermodynamically favored Td form, the utilization of WTe2 in electronic device architectures such as field effect transistors may need to be reevaluated.

The rise of graphene was a defining point for the discovery and development of stable two-dimensional layered materials (2DLM)12. This breakthrough has stimulated the exploration of 2D materials such as hexagonal boron nitride (hBN)3 and transition-metal dichalcogenides (TMDs)4 of formula MX2, where M is a IVB-VIB transition metal atom (IVB: Ti and Zr; V-B: Nb and Ta; VI-B: Mo and W) and X is a chalcogen (S, Se, or Te). Due to the d-orbitals involved in their electronic structure, the TMDs exhibit a wide range of electronic properties that have led to advances in practical devices, including field effect transistors56789101112, photodetectors1314, chemical15 and biosensors161718, and nano-electromechanical systems (NEMS)1920.

The synthesis, structure, and electronic properties of TMD systems have been recently explored using both experimental and theoretical techniques421222324252627. Among the many attributes, the absence of dangling bonds and interface traps22, as well as the presence of finite and distinctive band gaps52228, make TMDs attractive as components of tunneling field effect transistors (TFETs). The foundation of the TFET is to combine dissimilar TMDs in such a way that electrons in the valence band of layers of VIB-TMDs can easily tunnel into the conduction band of layers of IVB-TMDs. Among these materials, single-layer 2H-WTe2 is expected to have the narrowest band gap of the semiconducting VIB-TMDs at ~0.7 eV5232930. This suggests a high electron mobility that could maximize the efficiency of electron injection in TMD TFETs912.

The physical and electronic properties of WTe2 and other TMDs have been widely modeled for their device performance as semiconductor materials based on a 2H structure112123262930313233, despite the fact that experimental data and non-constrained geometry optimization using density functional theory (DFT) provide strong evidence that WTe2 has its lowest energy in a distorted 1T (Td) structure34, also called 1T’3536. Both 2H3738 and Td structures35394041424344 have been reported in experimental studies of WTe2, and recent reports have focused on the extraordinary magnetoresistance353645 and thermoelectric3245 properties of Td-WTe2. In this paper, we study bulk and exfoliated, crystalline WTe2 synthesized under near equilibrium conditions. WTe2 is found to have a distorted 1T structure and semimetallic electronic characteristics. The results presented here agree with early theory and experiments from the 1960s34394043, unlike several recent theoretical calculations and predictions of WTe2 based on a MoS2-like trigonal prismatic isomorph23262930313246.

Results and Discussion

To verify the stable phase and electronic properties of WTe2, we utilize DFT to model the crystal structures, based on atomic positions calculated from X-ray diffraction patterns of Td-WTe244 and from a hypothetical 2H-WTe2 structure1123262947 (Fig. 1). The 2H-WTe2 structure (Fig. 1a) has hexagonal symmetry. The upper and lower tellurium atoms are bonded to a central W atom, forming a trigonal prismatic arrangement similar to that found in 2H-MoS2 and 2H-WSe22148. The Td-WTe2 structure (Fig. 1b) is similar to that of the 1T polytype, in which the upper tellurium atoms are rotated by 180o with respect to the lower tellurium atoms, forming W-centered octahedra. However, in Td-WTe2 the tungsten atoms are shifted by 0.87 Å in the layer plane and 0.15 Å in the perpendicular direction (along the c-axis) from the center of the octahedron. This shift of tungsten atoms results from a shortened metal-metal distance in transition metal tellurides49 due to strong intermetallic bonding3450. As a consequence, the tungsten atoms are unequally spaced and form a zigzag chain along the a-axis (Fig. 1b). The distances between tungsten atoms in Td-WTe2 alternate along the b-axis at 2.862 and 4.394 Å, in contrast to the 2H phase where they are equally spaced at a distance of 3.6 Å. Additionally, the tellurium atoms are no longer coplanar, but instead exhibit a zigzag structure with 0.6 Å c-axis variation in atomic positions. Finally, the tungsten–tellurium bond lengths are also uneven at 2.719 and 2.815 Å, compared to a uniform 2.769 Å for 2H-WTe2. A detailed comparison of the 2H-WTe2 and Td-WTe2 crystal structures, lattice parameters, and bond angles are given in the Supplementary Information. Adjacent WTe2 layers exhibit AB stacking, where each layer is rotated 180o with respect to each other. These changes in bonding environment result in the lowering of the lattice symmetry from hexagonal to orthorhombic. Since WTe2 layers are bound by weak van der Waals interaction, we also performed a structural optimization using the Grimme method for van der Waals corrections5152, but it was observed that LDA yields a better description of the stacking distance, as summarized in Table S2 of the supplementary information.

Figure 1

Comparison of the tungsten-tellurium coordination (side and plane views) of (a) 2H-WTe2 and (b) distorted “1T”, or Td-WTe2. The theoretical 2H structure exhibits trigonal prismatic coordination with uniformly displaced atoms, whereas tungsten atoms in the Td structure are octahedrally coordinated by Te with alternating long and short distances between W atoms due to strong intermetallic bonding. The electronic band structures (c) indicate that bulk WTe2 in the 2H structure has an indirect 0.702 eV bandgap. Bulk WTe2 in the Td structure (d) has a 0.21 eV band overlap in Γ-X, and the density of states (e) reaches a minimum, but never goes to zero near Fermi level.

The crystal structure plays a significant role in the characteristic electronic properties of WTe2. Based on the optimized structures of 2H- and Td-WTe2, we have calculated the electronic band structures and summarized the results in Fig. 1. Full band structures are displayed in Figure S2 in the supplementary information. Fig. 1c shows the band structure for 2H-WTe2. The d-orbitals of tungsten split into three different bands and the 2H-WTe2 trigonal prismatic coordination gives rise to a calculated 0.702 eV bandgap. In contrast, the band structure and low density of states (DOS) at the Fermi energy of bulk Td-WTe2 (Fig. 1d) shows that it is a semimetal, with few bands crossing the Fermi energy in the three main axes of the Brillouin zone. The highest valence band bends upward while the lowest conduction band bends downward to form a 0.21 eV overlap, confirming the findings of Augustin et al.34 A detailed calculation of the band structure around the crossing point in the ΓX segment shows an indirect band overlap of 0.3 eV, with a separation of 11 meV among the bands at their closest point (see inset in Fig. 1d), which is well below the thermal energy at room temperature (25 meV).

Bulk WTe2 crystals were grown by chemical vapor transport (CVT) as in previous reports (Fig. 2a),344041424450 using bromine (Br) as the transport agent. Following the synthesis, the powder and bulk crystals were characterized by X-ray powder diffraction (XRD), scanning electron microscopy (SEM), X-ray photoelectron spectroscopy (XPS), Raman spectroscopy, and temperature-dependent current-voltage (I-V) measurements. SEM shows that bulk WTe2 crystals (Fig. 2b) exhibit a platelet morphology (Fig. 2c) with no apparent angles that would be expected for hexagonal crystals. XRD patterns (Fig. 2d) indicates an atomic arrangement based on the primitive orthorhombic space group Pmn21, consistent with the formation of the Td structure. The experimental XRD pattern collected from bulk WTe2 crystal was compared to simulated XRD patterns based on the Td-WTe2 and 2H-WTe2 structures, and shown in Figure S3 in supplementary information. It can be observed that distinctive reflections of the Td structure are present in the experimental diffractogram. Moreover, W:Te ratio calculated from high-resolution elemental XPS spectra from W 4d and Te 3d regions in Fig. 2e confirms a W:Te ratio of 1:1.9 for bulk WTe2 crystals, suggesting a slight Te deficiency. The full range XPS in Figure S4 also confirms the W:Te ratio and verifies that the transport agent is not incorporated in significant quantities into the WTe2. To understand why the Td structure is favored, the enthalpies of formation of 2H and Td-WTe2 were calculated within a pressure range near equilibrium, which is representative of the chemical vapor transport (CVT) method (Fig. 2f). As is evident from Fig. 2f, the enthalpy of formation for the Td structure is lower at equilibrium (zero pressure) by 0.58 eV per WTe2 formula unit. This is also the case for non-equilibrium synthesis conditions up to at least 0.6 GPa.

Figure 2

Synthesis of WTe2 by (a) chemical vapor transport results in the formation of (b) large bulk crystals, accompanied by (c) smaller crystallites. X-ray powder diffraction (d) confirms that WTe2 crystallizes in space group Pmn21 in the Td structure. (e) XPS shows that Td-WTe2 is stoichiometric. Density functional theory (f) indicates that the enthalpy of formation of the Td-WTe2 phase is lower than that of the 2H phase, regardless of pressure, indicating it is the most stable form of bulk Td-WTe2.

To date, there are no reports on the vibrational properties of 2H- or Td-WTe2. We have explored the vibrational properties as a function of incident photon energy via Raman spectroscopy, and the results are shown in Fig. 3. Flakes of Td-WTe2 were exfoliated onto SiO2/Si substrates (each>10 layers thick), and Raman spectra were acquired using 647 and 488 nm laser excitations. With 488 nm excitation, the vibrational modes are dominated by peaks at 112, 133, 163, 165, and 212 cm−1. The same vibrational modes are evident with 647 nm excitation, but with slight frequency shifts, and an additional peak appears at 118 cm−1. To understand the origin of the Raman peaks, we used density functional perturbation theory (DFPT)53 to calculate the vibrational modes in Td-WTe2. The calculated modes, as well as the symmetry analysis and their infrared and Raman activity are listed in Table S3 (Supplementary Information), and the phonon modes that correlate to the experimentally observed vibrations (Fig. 3a) are shown in Fig. 3b. Because of the structural distortion induced by metal-metal bonding, the out-of-plane vibrational modes of Td-WTe2 are not oriented perpendicular to the WTe2 sheets. The out-of-plane Raman-active modes for Td-WTe2 are vibrating either along the W-Te bond or at an angle to the vertical line, in contrast to the A1g mode of 2H-WTe2 that involves atomic motion perpendicular to the layer plane. The 118, 133, and 212 cm−1 peaks in the Td-WTe2 Raman spectrum are identified as “tilted” out-of-plane A1 modes at 119, 133, 218 cm−1 and a tilted B1 vibrational mode at 216 cm−1. The two deconvoluted peaks near 163 cm−1 and 166 cm−1 in the 647 nm Raman spectrum in Fig. 3(a) are identified as the in-plane B2 and A1 vibrational modes in different directions; the observed frequencies are close to the calculated 159 and 167 cm−1 frequencies in the model. The other calculated Raman-active modes in Table S3 may be too low in intensity (relative to the background signal) to be observed in the Raman scattering experiments.

Figure 3

(a) Raman spectra of Td-WTe2 using 647 (red) and 488 nm (blue) laser excitation. The peaks in the spectra can be assigned to in-plane and out-of-plane (b) Raman-active vibrational modes. The major peaks were processed with Lorentzian peak fitting in both spectra, and the two in-plane vibrational modes at 162 and 167 cm−1 were deconvoluted in the spectrum obtained with 647 nm excitation.

Temperature dependent resistance measurements confirm the metallic nature of synthetic Td-WTe2. This is verified for thick (9 - 130 layers, measured by AFM) exfoliated Td-WTe2 flakes. Two-terminal device structures were fabricated using titanium-gold electrodes (Fig. 4a). The series contact resistance was found to be 6.76 × 10−5 Ω cm using a transmission line measurement (TLM),54 and was subtracted from the total measured resistance. Figure 4c shows the temperature-dependent resistivity, which varies between 1 × 10−3 and 7 × 10−3 Ω cm at 300 K, depending on the layer thickness. The different values obtained at different layer thicknesses suggest that the layer structure may affect carrier transport though Td-WTe2. Importantly, the resistivity of WTe2 is strongly correlated to temperature, increasing with increasing temperature over most of the range measured. The positive temperature dependence of the resistivity and the bulk resistivity values, which are ~2 orders of magnitude higher than those of ordinary metals at 300 K55, are consistent with the calculation that Td-WTe2 is metallic in nature. We note that while two-terminal measurements do not provide direct access to the carrier concentration, and therefore confirmation of semi-metallic WTe2, they are sufficient to verify that Td-WTe2 is not semiconducting – a critical point for the device community when considering this material in electronic device architectures.

Figure 4

Representative (a) SEM and (b) AFM images of 9-layer (9L) Td-WTe2 Resistivity measurements as a function of temperature (c) confirm that Td-WTe2 is metallic in nature (resistivity proportional to temperature), rather than the semiconducting behavior recently suggested22282930.

Finally, the stability of WTe2 is a critical aspect of robust operation in a variety of applications. In the case of exfoliated flakes, the Raman spectra evolved with time during the data collection process, indicating that environmental sensitivity must be considered. Surface characterization tools such as XPS and Raman spectroscopy were used to understand surface stability and sensitivity to ambient conditions. Figure 5 summarizes the high-resolution XPS and Raman spectra, which compare fresh exfoliated WTe2 with WTe2 that was exposed to ambient (air, 1 atmosphere, room temperature) conditions for extended periods of time. The XPS peak positions of the fresh exfoliated and aged WTe2 surfaces are listed in Table S4. Each XPS spectrum was calibrated with the carbon C 1s binding energy (BE) position and corrected with a relative sensitivity factor (R.S.F.). For the high resolution elemental XPS spectrum, normalization of intensities was used to compare spectra collected from the same exfoliated WTe2 sample with increasing exposure time to air. Elemental XPS analysis reveals the evolution of a secondary chemical bond in the Te 3d peaks corresponding to an increase in Te-O binding. The primary degradation appears to be the formation of Te-O bonds, which is accompanied by an increase in the intensity of the O 1s peak, and formation of a small energy loss peak at the left shoulder of the W 4d region. This indicates that the WTe2 surface is air sensitive, which could affect the stability of few-layer exfoliated WTe2. Table S4 lists the binding energies from peak fitting analysis of the Te 3d, O1s and W 4d regions of the spectra of WTe2 and degraded (or oxidized) WTe2. There are two sets of Te 3d3/2 and Te 3d5/2 binding energies from the peak fitting analysis, which refer to the Te 3d binding energies of the fresh exfoliated WTe2 surface and those of TeO2 from the NIST XPS database6061. Raman spectra in Fig. 5b show that the aged WTe2 surface may have minor changes near the 162-167 cm−1 region of in-plane vibrational modes. However, the Raman spectra may not be sensitive enough to detect the formation of tellurium oxides on the surface. With laser excitation, using the 647 nm laser and three periods of 45 seconds acquisition time, the WTe2 surface is visibly modified and two new vibrational modes at 124 and 142 cm−1 were detected. These peak positions correlate well with those of TeO2, and confirm the formation of Te-O bonds under accelerated aging, suggesting this as the mechanism of degradation for Td-WTe2 when exposed to air or a combination of photons and air52.

Figure 5

Stability testing of WTe2 flakes in ambient air by (a) high-resolution elemental XPS spectra and (b) Raman spectroscopy. XPS reveals an increase in Te-O bonding in the Td 3d and O 1s spectra with minor changes in the W 4d spectra, indicating the formation of TeO2 on the surface. The Raman spectrum after 13 days of air exposure (blue) shows small changes in intensity of the two in-plane modes in the 160-167 cm−1 region. The Raman spectrum of a laser-degraded sample (red) suggests that photon-assisted oxidation can lead to rapid degradation of WTe2 via the formation of TeO2.

Conclusion

The distorted 1T structure (Td) of bulk tungsten ditelluride has been experimentally verified to be thermodynamically stable relative to the 2H polymorph. The calculated band structure of Td-WTe2 shows a 0.21 eV indirect band overlap from the Γ to X direction, indicating that it is a semimetallic TMD material. Raman spectra and DFT simulations provide evidence that the out-of-plane vibrational modes involve atomic motions at angles that are displaced from the c-axis direction because the distorted octahedral bonding in Td-WTe2. We have experimentally verified that Td-WTe2 behaves as a metal, with an as-yet unexplained strong dependence of the resistivity on the thickness of multilayer flakes.

We have also evaluated the stability of thin flakes (9 – 130 layers) and found that care must be taken to ensure that oxidation does not occur, as the surface of Td-WTe2 is sensitive to ambient air. Finally, we note that this work clearly verifies that WTe2, grown via CVT under near equilibrium conditions, is not a semiconductor. This ultimately requires careful reconsideration of the use of WTe2 in a variety of device2228 architectures.

Methods

Crystal Growth

Tungsten ditelluride (WTe2) bulk crystals were produced by the chemical vapor transport (CVT) method with bromine as the transport agent. WTe2 powder was synthesized by heating a mixture containing stoichiometric amounts of tungsten (Acros Organics 99.9%) and tellurium (Strem Chemicals 99.9%) at 800 °C for 3 days in an evacuated and sealed quartz ampoule (10 mm ID, 12 mm OD, 150 mm length). The mixture was slowly heated from room temperature to 800 °C for 12 h; slow heating was used to minimize the possibility of explosion due to the strong exothermicity of the reaction. Some tellurium sublimed into the cooler zone of the ampoule (~350 oC), so the two ends of the ampoule were kept at 950 °C and 775 °C for another day to ensure that all the tellurium reacted with the tungsten. WTe2 single crystals were grown from the synthesized powder by chemical vapor transport with bromine (Sigma-Aldrich, 99.8 + %) as the transport gas at ~6 mg/cm3. The growth process ran for 4 d in an evacuated and sealed quartz ampoule (10 mm ID, 12 mm OD, 100 mm length), the hot and growth zones of which were kept at 840 °C and 900 °C, respectively. The resulting crystals were pumped under dynamic vacuum at room temperature for 1 d in order to remove any residual bromine.

Mechanical Exfoliation

WTe2 flakes were mechanically exfoliated onto fresh and cleaned Si/SiO2 substrates via the “scotch-tape” method53 and imaged by using an Olympus MX50 optical microscope.

Characterization

WTe2 powder and crystals were analyzed by X-ray powder diffraction (XRD) using a PANalytical XPert Pro MPD theta-theta diffractometer with Cu α x-ray source. Energy dispersive spectroscopy (EDS) on a FEI Nova NanoSEM 630 FESEM as well as Kratos Analytical Axis Ultra X-ray photoelectron spectra (XPS) by Kratos Analytical Axis Ultra were used to confirm the stoichiometry of both WTe2 powders and bulk crystals, and investigate the surface bonding and stability of the WTe2 flakes. Raman spectroscopy of exfoliated thick and few-layer WTe2 flakes was carried out using a Renishaw inVia confocal microscope-based Raman spectrometer with a spectral resolution less than 1 cm−1. Laser power was kept at 0.2mW at all times with 488 and 647 nm laser excitations. Electrical properties of WTe2 samples of different thicknesses were tested using two Ti/Au contacts made by a lift-off process at both edges of the exfoliated WTe2 flakes. Total resistance measurements were collected by using a Lakeshore Cryo Probe Station, which controlled the temperature from liquid nitrogen temperature 77 K to 400 K under vacuum. The size of each flake was measured by using images obtained with a Leo 1530 Field Emission Scanning Electron Microscope (FESEM) operated at 2 kV. The thickness and number of layers of the WTe2 flakes were determined by atomic force microscopy (AFM) using a Bruker Dimension Icon in tapping mode in air.

Theoretical calculations

A Td-WTe2 crystal structure model was created from the crystallographic data reported by B. Brown49 in 1966, adjusting the axes to match the conventional Pmn21 representation; while the 2H model was constructed from a MoS2 based model with lattice parameters from Kumar et al.29 and Ding et al.26 Geometry optimization of the initial structures was followed by the calculation of their electronic structures, vibrational properties, as well as enthalpies of formation as function of pressure, within Density Functional Theory (DFT), as implemented in CASTEP56 (Materials Studio 6.1, Accelrys, accelrys.com), as well as in Quantum Espresso 5.1.57 The Local Density Approximation (LDA) as parameterized by Perdew and Zunger5859 was selected for exchange and correlation functional, and dispersion corrections were implemented following the semi-empirical Grimme method (LDA + DFT-D)5152. Norm-conserving pseudopotentials were used for all the elements. Convergence analysis for the total energy, band gaps and forces set the cutoff energy of the plane wave basis set at 740 eV (CASTEP) and 680 eV (QE), and a Monkhorst-Pack grid of 10x20 × 5 for sampling of the Brillouin zone. Under these computational conditions the total energy and band gaps were converged to 0.1 meV. Geometrical optimizations were performed for both the LDA and LDA plus DFT-D functions until the structures reached configurations with energy differences of 5 × 10−6 eV/atom, and forces were less than 0.01 eV/A.

Additional Information

How to cite this article: Lee, C.-H. et al. Tungsten Ditelluride: a layered semimetal. Sci. Rep. 5, 10013; doi: 10.1038/srep10013 (2015).