Probing the Optical Properties and Strain-Tuning of Ultrathin Mo_1-xW_xTe₂

Abstract

Ultrathin transition metal dichalcogenides (TMDCs) have recently been extensively investigated to understand their electronic and optical properties. Here we study ultrathin Mo_0.91W_0.09Te₂, a semiconducting alloy of MoTe₂, using Raman, photoluminescence (PL), and optical absorption measurements. Mo_0.91W_0.09Te₂ transitions from an indirect to a direct optical band gap in the limit of monolayer thickness, exhibiting an optical gap of 1.10 eV, very close to its MoTe₂ counterpart. We apply tensile strain, for the first time, to monolayer MoTe₂ and Mo_0.91W_0.09Te₂ to tune the band structure of these materials; we observe that their optical band gaps decrease by 70 meV at 2.3% uniaxial strain. The spectral widths of the PL peaks decrease with increasing strain, which we attribute to weaker exciton-phonon intervalley scattering. Strained MoTe₂ and Mo_0.91W_0.09Te₂ extend the range of band gaps of TMDC monolayers further into the near infrared, an important attribute for potential applications in optoelectronics.

Graphical Abstract

Atomically thin layers of transition metal dichalcogenides (TMDCs), such as MoTe₂, have been extensively studied for fundamental physics and applications.^1-2 Tuning their optical properties, which can be achieved through doping, alloying, strain, and heating, is crucial for understanding their light-matter interactions and for various applications in electronics and optoelectronics. The optical properties of atomically thin MoTe₂ have recently been characterized experimentally,^3-7 and theoretical investigations have indicated that the aforementioned factors can significantly alter the material’s band structure.^8-12 In particular, theory has predicted and measurements have demonstrated that alloying with tungsten or electrical gating of MoTe₂ can induce a phase change from a semiconducting to metallic state.^8-9,13-16 To date, however, few studies have experimentally examined the effects of strain on MoTe₂ or on MoWTe₂ alloys.¹⁷

Here we investigate Mo_1-xW_xTe₂ with x = 0.09 and compare the properties of this alloy with the MoTe₂ compound. For alloys with higher W content, i.e., x > 0.09, all reported growth processes have produced crystals in the 1T’ phase. Therefore, this composition is close to the 2H phase boundary and should exhibit the lowest threshold for inducing a 2H to 1T’ phase change by an external perturbation.^15,18-19 However, the optical properties of the 2H phase of Mo_0.91W_0.09Te₂ have yet to be investigated in detail and compared to those of MoTe₂. We study atomically thin crystals of the Mo_0.91W_0.09Te₂ alloy by Raman scattering, photoluminescence (PL), and absorption measurements, and compare the results with those for MoTe₂. We also measure PL while applying in-plane uniaxial tensile strain to the monolayer (1L) crystals of both compounds to alter their band structure, which has not been previously examined.

We have grown bulk MoTe₂ and Mo_1-xW_xTe₂ crystals using the chemical vapor transport (CVT) method, as detailed in the Supporting Information (SI) sections 1.1 and 1.2. The 2H crystal structure of Mo_0.91W_0.09Te₂ has been confirmed by powder X-ray diffraction (XRD), as shown in Figure 1a (see Figure S1 in SI section 1.3 for an extended version of the XRD data) and by scanning transmission electron microscopy (STEM) (Figure 1b). The W mole fraction in the Mo_1-xW_xTe₂ alloy, as determined from X-ray energy dispersive spectroscopy (EDS), is x = 0.09 ± 0.01 (see Figure S2). For optical measurements, we have exfoliated atomically thin layers from the grown bulk MoTe₂ and Mo_0.91W_0.09Te₂ crystals (see SI section 1.4, Figure S3, and SI section 2 for details on the sample preparation and optical measurement setup).

Figure 1.

(a) Powder XRD pattern of 2H-Mo_0.91W_0.09Te₂. (b) High angle annular dark field scanning electron microscopy (HAADF-STEM) image of a 2H Mo_0.91W_0.09Te₂ sample together with an overlapped structural model; red (large) spheres: Mo/W atoms; green (small) spheres: Te atoms. Inset: Fast Fourier Transform (FFT) emphasizing the [001] zone axis.

We first perform Raman spectroscopy on Mo_0.91W_0.09Te₂ and MoTe₂ crystals exfoliated on polydimethylsiloxane (PDMS) substrates. Here and below all characterization measurements have been performed at room temperature and under ambient conditions. Figure 2 shows the Raman spectra of 1L, bilayer (2L), trilayer (3L), four-layer (4L), and bulk Mo_0.91W_0.09Te₂ crystals for excitation wavelengths of 532 nm, 633 nm, and 785 nm (see Figure S4 for the MoTe₂ spectra). We note that the relatively small band gaps of ultrathin MoTe₂ and Mo_0.91W_0.09Te₂ (see the section on PL) compared to other TMDCs enables the use of longer excitation wavelengths for resonant Raman spectroscopy.^3,20-21 We identify the following first-order Raman modes (zone-center phonons) in both materials: the out-of-plane A_1g ($A_1^{\prime }$ for odd, A_1g for even layers), in-plane $E_{2g}^1$ (E′ for odd, E_g for even layers), and out-of-plane $B_{2g}^1$ ($A_2^{″}$ for 1L, $A_1^{\prime }$ for odd, A_1g for even layers), which is Raman inactive in 1L and bulk crystals.^3,19,21-27 For 1L Mo_0.91W_0.09Te₂, the A_1g mode is at 172 cm⁻¹ and the $E_{2g}^1$ mode appears at 236 cm⁻¹. The weaker features at 200 cm⁻¹ and 345 cm⁻¹ have recently been attributed to second-order Raman processes²¹ and appear to be stronger in Mo_0.91W_0.09Te₂ than in MoTe₂, possibly due to changes in the electronic structure with the addition of W.

Figure 2.

Raman spectra of 1L to 4L and bulk Mo_0.91W_0.09Te₂ (modes labeled according to bulk notation) for excitation wavelengths of (a) 532 nm, (b) 633 nm, and (c) 785 nm. The spectra are vertically offset for clarity.

After comparing the Raman spectra of the two materials, we observe that the A_1g and $E_{2g}^1$ modes of Mo_0.91W_0.09Te₂ are slightly blueshifted and the $B_{2g}^1$ mode is slightly redshifted from those of MoTe₂ for 1L and 2L (see Table S1 for the Raman peak positions of both materials). We explain these observations as follows. Consider a simplified linear triatomic molecule model for which interlayer interactions in the 1L and 2L have little effect (e.g., no splitting of the $E_{2g}^1$ mode). We then expect the increased W content in Mo_0.91W_0.09Te₂ to increase the effective mass of the transition metal atoms and to result accordingly in a redshift of modes compared to MoTe₂.²⁸ This was reported for the $E_{2g}^1$ and $B_{2g}^1$ modes in alloys of Mo and W dichalcogenides (e.g., Mo_1-xW_xS₂ and Mo_1-xW_xSe₂).^29-33 We observe this redshift in the $B_{2g}^1$ mode and the second-order Raman mode at 200 cm⁻¹. However, the frequency of the A_1g mode depends more on bond strength than on mass of the transition metal, since this atom remains stationary in this vibrational mode.²² Since W-Te bonds are stronger than Mo-Te bonds,³⁴ adding W to MoTe₂ stiffens the bonds, leading, as in other TMDC alloys,^31-33 to a blueshift of the A_1g mode. Interestingly, in contrast to other TMDC alloys,^30-33 we observe a blueshift of the $E_{2g}^1$ mode for 1L and 2L of the alloy compared to the MoTe₂ crystal. This blueshift also results from the increased bond strength, which apparently outweighs the influence of the increase in mass.

Next, we examine the relative intensities of the Raman modes for different layer thicknesses in both materials. For 532 nm excitation, for both MoTe₂ and Mo_0.91W_0.09Te₂ crystals of all thicknesses, the $E_{2g}^1$ mode is stronger than all other modes. The A_1g mode is strongest for 1L and becomes much weaker with increasing thickness (Figure 2a, Figure S4a). For 633 nm excitation, the A_1g mode is enhanced compared to the $E_{2g}^1$ mode for 1L samples of both materials. We attribute the strength of this mode to a resonance effect associated with the excitation photon energy being close to that of the A′ or B′ excitons in the material (see Figure 3 for the A′ and B′ peak positions).^25-26 For 2L crystals, the A_1g and $B_{2g}^1$ modes are comparable in intensity, but weaker than the $E_{2g}^1$ mode. For 3L and 4L crystals, the A_1g mode splits into 2 peaks [A_1g(R1) and A_1g(R2)], known as Davydov splitting,^25,35 which are stronger than the $B_{2g}^1$ mode (see Figure 2b). The ratio of the $B_{2g}^1$ to $E_{2g}^1$ mode can also be used to identify thickness in few-layer crystals using both 532 and 633 nm excitation.^4,23

Figure 3.

(a) PL spectra of 1L to 4L and bulk Mo_0.91W_0.09Te₂. 1L to 4L spectra are vertically offset for clarity. Black arrows indicate the width of 48 meV of the 1L spectrum. (b) Reflection contrast (ΔR/R) spectra for 1L to 3L Mo_0.91W_0.09Te₂. (c) Comparison of the A exciton and PL peak positions as a function of crystal thickness. (d) Absorption spectrum for the 1L in the near-IR, shown in terms of absorption for a free-standing layer. The green curve shows the contributions of the A and B excitons to the spectrum based on a fit (dotted line) to the experimental data.

Using 785 nm excitation, we observe that the $E_{2g}^1$ mode is much weaker than the A_1g mode for all layer thicknesses. For the thicknesses we have measured, the A_1g mode is weakest for 2L and strongest for 4L crystals in both Mo_0.91W_0.09Te₂ and MoTe₂ (see Figure 2c and Figure S4c). The reason for this non-monotonic change of the intensity of the A_1g mode with increasing thickness is unclear. The $B_{2g}^1$ mode does not appear in the 785 nm Raman spectra for any crystal thickness, as in the case of MoS₂ and MoSe₂ for excitation energies far from the C electronic resonance (see Figure 3 for the C peak position).^24,36 Calculations have shown that the A and B excitons have wave functions that are mainly confined to the individual layers,^24,36 and it has been predicted that for excitation energies near the A and B excitons, the active $B_{2g}^1$ mode of the few-layer crystals is inactive, as it is for 1L crystals. However, the wavefunction associated with the C transition is calculated to be not confined to individual layers.^37-38 Hence, excitation nearer to the C feature, in this case 532 and 633 nm, yields Raman spectra with stronger $B_{2g}^1$ modes. The comparisons of the Raman spectra with three different excitation wavelengths presented here provides a rapid means of determining the thickness of ultrathin layers of Mo_0.91W_0.09Te₂ and MoTe₂.

We have performed PL measurements in order to study the band structure of Mo_0.91W_0.09Te₂. Figure 3a displays the PL spectra of 1L to 4L and bulk Mo_0.91W_0.09Te₂ supported on PDMS substrates. We see that the PL intensity decreases, the peak position redshifts, and the spectral width (full width at half maximum) increases with increasing layer thickness. The 1L spectrum exhibits a single emission peak, with the maximum located at 1.10 eV and a width of 48 meV. The width of the 2L is 20 meV greater than that of the 1L. For the bulk, we find a much broader and weaker PL feature peaked at 0.98 eV.

In order to comment on the nature of the peaks observed in the PL spectra and understand the changes with increasing material thickness, we have also measured the reflection contrast (ΔR/R) spectra of 1L to 3L Mo_0.91W_0.09Te₂ on PDMS substrates, as displayed in Figure 3b, to probe their absorption spectra (see SI section 2 for details on the reflection contrast measurements). We expect to observe only direct optical transitions in the reflection contrast spectrum since the indirect transitions produce only weak contributions to the absorption spectrum. Several features are observed in the spectra of Figure 3b, and we expect them to arise from mechanisms similar to those in MoTe₂. Thus, we have labeled them according to the bulk assignments of Wilson and Yoffe³⁹ as was previously done for the case of MoTe₂.³ These spectroscopic features are associated with transitions in different parts of the Brillouin zone of Mo_0.91W_0.09Te₂. The A, B and A′, B′ pairs have been identified as excitonic transitions with the A-B splitting arising from spin-orbit interactions.^39-40 As for other TMDC monolayers, the A and B peaks are assigned to excitonic peaks associated with the lowest direct optical transition at the K-point.^41-43 The C and D features have been attributed to regions of parallel bands near the Γ point of the Brillouin zone of bulk MoTe₂⁴⁴ and to similar parallel bands in monolayers of other TMDCs.^37,45

We now compare the PL and reflection contrast response of Mo_0.91W_0.09Te₂ for different thicknesses. Figure 3b shows that the reflection contrast increases and the A exciton redshifts with increasing thickness. The PL spectra in Figure 3a show that the peak position also redshifts; however, the rate of the redshift of the PL peak is faster than that of the A exciton and the intensity of the PL peak decreases as opposed to that of the A exciton. We report the position of the A exciton and the PL peak position as a function of thickness in Figure 3c. The A exciton redshifts by 33 meV to 1.067 eV from 1L to 2L and continues shifting gradually with increasing thickness. However, the PL peaks redshift more than the A exciton peaks; the two peaks are separated by 5 meV for 1L, but by 45 meV for 4L.

We attribute this large difference in the shift between the absorption and emission features to the emergence of an indirect transition at lower energies than the A exciton with increasing thickness, as observed in other semiconducting TMDCs.^46-47 We expect the indirect band gap to contribute to PL, but not significantly to the absorption spectrum. We thus conclude that 1L Mo_0.91W_0.09Te₂ exhibits a direct band gap, unlike the crystals thicker than 2L, in accordance with the behavior of its MoTe₂ counterpart. The PL intensity of the 1L crystal is about three orders of magnitude greater than that of the bulk, due to the direct optical gap transition of the former. Even though the PL intensity is significantly weaker for 2L than it is for 1L, since the PL peak and the A exciton are separated by only 10 meV, the assignment of the 2L material as indirect gap is unclear. However, the 20 meV increase in the width of the PL and A exciton peaks of 2L as compared to 1L strongly indicates the presence of a scattering channel for the A exciton in the 2L. We also note that there have been different opinions on the nature of the 2L MoTe₂ band gap, both at room^3,5 and low temperatures.⁴

The near-IR part of the 1L absorption spectrum is shown in greater detail in Figure 3d. We find that the A and B features are located at 1.10 eV and 1.38 eV, respectively. The corresponding features appear at similar energies of 1.10 eV and 1.35 eV, respectively, for 1L MoTe₂.³ We note that the direct gap, or A exciton, position in Mo_0.91W_0.09Te₂ is not very different from that of MoTe₂, which is compatible with the fact that for a given chalcogen (S, Se), the band gaps of Mo and W dichalcogenides are quite similar.²⁰ The A-B splitting for 1L is found to be 276 meV, which is slightly greater than the MoTe₂ value of 250 meV.³ We attribute this increase to the presence of W, since dichalcogenides of W have significantly higher A-B splitting than those of Mo.^20,48-49 Thus, our PL and reflection measurements have revealed that 2H Mo_0.91W_0.09Te₂ is a semiconducting material like 2H MoTe₂.

To gain further insight into the band structure of 1L MoTe₂ and Mo_0.91W_0.09Te₂, we perform strain-dependent PL measurements on these materials. We apply uniaxial in-plane tensile strain to a flexible substrate using a two-point bending apparatus (see SI section 1.4 and Figure S3 for more details). We used PEN (polyethylene naphthalate) or PETG (polyethylene terephthalate glycol-modified) instead of PDMS for the bending platform (Figure S3), due to their higher Young’s modulus. Figure 4a-b shows the PL measurements of 1L Mo_0.91W_0.09Te₂ and MoTe₂ as a function of strain. As the strain increases, the PL peak redshifts, corresponding to a decrease in the band gap. For Mo_0.91W_0.09Te₂, the PL peak shifts from 1.09 eV at 0% strain to 1.02 eV at 2.3% strain or −30 meV/% strain. For MoTe₂, the PL peak shifts from 1.08 eV at 0% strain to 1.01 eV at 2.1% strain or −33 meV/% strain. We note that these values of strain and peak shift with strain are reasonable, as discussed in SI section 4 and shown in Figure S5. The widths of the Mo_0.91W_0.09Te₂ and MoTe₂ PL peaks decrease from 63 meV and 59 meV, respectively, at 0% strain to minimum values of 49 meV and 42 meV, respectively, with strain. Figure 4c shows how the 1L MoTe₂ PL peak energy and spectral widths vary with strain (see SI Figure S6a for the Mo_0.91W_0.09Te₂ results).

Figure 4.

PL spectra of 1L (a) Mo_0.91W_0.09Te₂ and (b) MoTe₂ under different amounts of uniaxial tensile strain. All spectra are vertically offset for clarity. (c) The dependence of the peak energy and spectral width of 1L MoTe₂ on strain. (d) Schematic band structure of strained and unstrained 1L MoTe₂, showing changes in the energy separation between the minima of the K and Q valleys in the conduction band.

To account for the unexpected decrease of the width of the emission peak in the PL spectra with increasing strain for both materials, we can identify three possible factors: 1) the contribution to PL from trions is suppressed, 2) the material conforms better to the flexible substrate (after being transferred), reducing inhomogeneous broadening, or 3) exciton-phonon scattering is partially suppressed.

We first consider the suppression of trion emission due to strain. Trions are not expected to contribute much to absorption unless the sample is heavily doped.^50-51 However, trions can still contribute appreciably to PL since they are more stable even at room temperature due to their high binding energy of about 25 meV.^6-7 Therefore, if suppression of trion PL caused the reduction in the linewidth, we would not expect to see a parallel narrowing in the absorption feature with strain. However, our measurements show a similar decrease in the width of the absorption and PL features, thus ruling out this mechanism (see SI section 6 and Figures S7 and S8).

Second, we consider the possibility that the material conforms better to the substrate under strain, leading to a reduction of inhomogeneous broadening. We first observe that the width of the emission feature of strained 1Ls drops to a value as low as 42 meV (Figure 4c), which is less than that of any as-exfoliated samples (for which the width is around 47-50 meV).^3,5 Further, similar measurements on 1L WSe₂ by Schmidt et al. have demonstrated that the strain-induced narrowing of the spectral widths is reversible upon the release of strain.⁵² This observation suggests that the elimination of inhomogeneities alone cannot account for line narrowing in their case, as the inhomogeneities are not expected to recover fully when the strain is released. Hence, we do not expect this mechanism to explain our observations.

Third, we consider the effect of strain on the exciton-phonon coupling. As we know from the current and earlier studies, 1Ls of both materials have direct optical gaps.^3-5,53 However, the 1L has an indirect transition at a slightly higher energy than that of the A exciton, which is known from band structure calculations and from extrapolating the energy of the indirect gap as compared to that of 2L and thicker crystals.^3-5,53 Moreover, the energy separation between the minima of indirect Q (also known as T)^54-55 and direct K valleys, ΔE_QK, in the conduction band will increase with uniaxial (as well as biaxial) strain in similar material systems.^11-12,56-58 This is illustrated in Figure 4d, which shows a schematic band structure of strained and unstrained 1L MoTe₂. However, scattering of an electron from the K to Q valley requires the absorption of a phonon. This becomes less likely with increasing ΔE_QK. Therefore, we infer from our results that a weakening in the exciton-phonon intervalley scattering is mainly responsible for the decrease in the spectral linewidths. We attribute at least 6 meV of the width of the emission feature in the unstrained material to the exciton-phonon scattering from K to Q (Γ_QK). We also estimate an upper limit of 110 fs for the lifetime of the exciton-phonon scattering from K to Q in 1L MoTe₂ and Mo_0.91W_0.09Te₂ (see SI section 6). The ability to suppress phonon absorption indicates that Q and K states are within a few times the phonon energy and the highest energy phonon to mediate such scattering [the LO(E′) mode] is about 30 meV.^21,54,59 Thus, we predict that ΔE_QK of the unstrained 1Ls is not much larger than 30 meV. We expect the linewidth narrowing effect via tensile strain to be smaller (larger) for MoS₂ (WSe₂ and WS₂), for which ΔE_QK is predicted to be larger (smaller) than that in MoTe₂.^54,59

Here we note an advantage of a strain-dependent study over a temperature-dependent one. Strain, like temperature, can modify the band structure, but in the former case without significantly affecting the population of phonons. We also note that it is typically assumed that the relative energy of different valleys is not affected by temperature. This is not always correct as lowering the temperature has a similar effect to biaxial compressive strain.⁶⁰ Therefore, tuning the bands of nearly direct or indirect gap semiconductors via strain will provide more insight about the band structure and can help interpret the temperature-dependent studies on the spectral linewidths of TMDCs.^6,52,61-62

Finally, we would like to consider the implications of weaker exciton-phonon scattering for electron transport. There have been calculations of the enhancement of transport properties in group VI TMDCs due to the change in the relative energies of K and Q valleys in the conduction band under tensile strain, leading to a reduction in electron-phonon scattering.^11,63 We propose that this effect should be larger for MoTe₂ than MoS₂, as the former has a smaller value for ΔE_QK. A similar phenomenon has been exploited to enhance the electron mobility of silicon transistors, where the intervalley energy separation increases with tensile strain, leading to reduced intervalley electron-phonon scattering.^64-65

In summary, we have characterized single-crystal 2H Mo_0.91W_0.09Te₂ down to monolayer thickness with photoluminescence, absorption, and Raman spectroscopy. We have determined that atomically thin 2H Mo_0.91W_0.09Te₂ is a semiconductor with similar optical properties to MoTe₂; the monolayer possesses a direct optical band gap at 1.10 eV, and the thicker layers become indirect. The presence of W in the alloy alters the band structure, as observed by absorption measurements, and can be studied further by Raman spectroscopy, particularly using resonant excitation.⁶⁶ Given that this alloy is closest in W-content to the 2H to 1T’ phase transition,¹⁹ it may be a promising material for phase-change memory applications. We have manipulated the band structure of monolayer MoTe₂ and Mo_0.91W_0.09Te₂ via tensile strain up to 2.3% and have thereby lowered the optical band gap of these materials to near 1 eV. We have thus extended the optical range of group VI TMDC monolayers further into the near-infrared (NIR) region. We have also observed that the relative energy separation between valleys changes with strain, and that exciton-phonon intervalley scattering can also be manipulated in this fashion. We attribute the reduced spectral linewidth of the A exciton under tensile strain to a decrease in the rate of exciton-phonon scattering. This suggests a corresponding decrease in the electron-phonon scattering rate and a potential improvement in the electrical transport properties in these materials with tensile strain.