The Electronic and Physical Structure Evaluation of MoS_2(1−x)Te₂x Alloy Fabricated with Co-Sputtering and Post-Deposition Annealing in Chalcogen Ambient

Silicon or glass was used as substrates. There was no pre-deposition treatment for the substrates. For the sputtering, RF magnetron sputtering was performed and the MoS₂ and MoTe₂ targets were simultaneously sputtered (co-sputtering). The sputtering power was fixed to 100 W for MoS₂ and varied between 0 and 150 W for MoTe₂. The substrate temperature was varied between 300 °C and 450 °C. The base pressure was 5 × 10⁻⁵ Pa and the Ar pressure during the deposition was 0.3 Pa. The sputtering duration was set such that the film thickness was either 6.5 or 10 nm. The thickness was calculated in advance from the individual deposition rate of MoS₂ and MoTe₂. The sample holder was rotated in order to assure uniform thickness and composition within each sample. For some samples, post-deposition thermal treatment in either S or Te ambient was carried out. The thermal treatment was performed in a quartz tube reactor at atmospheric pressure. Organic precursors (t-C₄H₉)₂S₂ and (i-C₃H₇)₂Te were used as S and Te sources, respectively. The precursors were stored in stainless bottle and supplied to the reactor by bubbling method. The reactor temperature was set between 300 °C and 600 °C. The supply rate of the S and Te source was adjusted by controlling the precursor bottle temperature and carrier gas flow rate. The supply rate was set either to 4.5 or 6.7 × 10⁻⁶ mol min.⁻¹ for sulfurization and set to either 1.0, 1.7, or 2.0 × 10⁻⁵ mol min.⁻¹ for tellurization. N₂ or H₂ carrier gas was used as the carrier gas to supply the S or Te precursor. The as-sputter samples are referred to as "sputter samples" and samples on which post-deposition thermal treatment was performed are referred to as "anneal samples." For more detail of the fabrication processes for different samples, please refer to section 1 of the supplemental material (Figs. S1 to S4 are available online at stacks.iop.org/JSS/9/093018/mmedia).

The composition and the valence spectra were obtained by X-ray photoelectron spectroscopy (XPS). K-Alpha from Thermo Fisher Scientific equipped with Al-Kα X-ray source (E = 1486.6 eV) was used. The detection angle was 90°. The composition was calculated by taking the spectral area ratio of molybdenum (Mo), S, and Te with the relative sensitivity factor taken into account. The bandgap was evaluated with spectroscopic ellipsometry (SE). GES5E from SemiLab was used. The spectroscopic parameter Ψ and Δ were evaluated first. Then, Tauc-Lorentz optical model was used to fit Ψ and Δ to obtain the extinction coefficient k. From k, the absorption was calculated and subsequently Tauc-plot method was used to finally obtain the bandgap. Since the films were multi-layer (6.5 or 10 nm), indirect bandgap transition was assumed. The physical structure was evaluated with X-ray diffraction (XRD) using SmartLab from Rigaku. The apparatus is equipped with Cu-Kα X-ray (λ = 1.5406 Å) source and out-of-plane measurements were carried out.

The relative valence edge energy were obtained by XPS. Figure 1a shows the representative valence spectra of the anneal samples for different Te to Mo ratio. The valence spectra gradually changes according to the Te/Mo ratio. The greater the Te/Mo ratio, the larger the intensity of the low binding energy peak becomes. This shift corresponds to the shift in the valence edge energy. Figures 1b and 1c show the valence edge energy of both the sputter and anneal samples for different S or Te to Mo ratio. In order to show the pure alloying effect, samples that formed uniform alloy are only shown and samples that exhibited phase separation, which is discussed below, are not shown (supplementary material, Figs. S5 to S7). The general tendency that can be seen is, the smaller the S ratio or the larger the Te ratio is, the higher the valence edge energy is. It can also be seen that the valence edge energy and its shift aligns for the sputter and anneal samples. The valence edge energy dependence on chalcogen to Mo ratio is also shown in Fig. 1d. Figure 2a shows the valence edge energy for different Te concentration x (Te/(S + Te)). It appears that the valence edge energy does not shift linearly with x, the same trend that was seen with bandgap value in our previous studies. ²⁸ The valence edge energy distribution was calculated by fitting with second order polynomial. The valence band maximum (VBM) value of MoS₂ (x = 0) was set to 0 as a reference point. The resulting curve is shown in Fig. 2b. The bowing parameter for the valence energy curve shown in Fig. 2b was calculated to be 1.12 eV.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 3a shows example XRD profiles. For a single peak profile as shown in (1) of Fig. 3a, the film is considered to form a uniform alloy where the chalcogen atoms are mixed uniformly throughout the film. For a peak that can be deconvoluted into 2 peaks as shown in (2) of Fig. 3a, the film is considered to have phase separation where part of the film is purely MoTe₂ and the rest is the alloy. XRD profile with no peak ((3) of Fig. 3a) indicates that there is no layer structure since the XRD measurement was carried out in out-of-plane configurations. Figure 3b shows an example of peak deconvolution. As mentioned above, one peak corresponds to the alloy and the other peak corresponds to pure MoTe₂. Figure 3c shows the d₀₀₂ spacing distribution for samples with different Te concentration. For the sputter samples, the phase separated ones are omitted (see Supplemental material Fig. S8 for the plot including the phase separation data). For the anneal samples, the phase separated samples showed two different d₀₀₂ values. The fact that one of the d₀₀₂ aligns with the d₀₀₂ spacing of MoTe₂ confirms that the phase separation is occurring with part of the film turning into pure MoTe₂. From the d₀₀₂ values of the annealed alloy samples, the spacing distribution curve was obtained as shown in Figure 3d. The sputter sample was not taken into account for the following reasons. For the sputter-deposited films, the film suffers from chalcogen loss during the deposition. Deficiency in chalcogen, which may be readily replaced by oxygen once the sample is exposed to atmosphere, is reported to cause expansion in the c-axis direction. ^31–33 Taking the d₀₀₂ distribution curve as the "true" distribution, the deviation of the sputter samples from the true distribution was calculated (Fig. S9). The deficient chalcogen amount was calculated by 2—(S + Te)/Mo. It can be seen that the d₀₀₂ spacing deviation from the true distribution increases as the deficient chalcogen amount increases.

Zoom In Zoom Out Reset image size

The bowing behavior for the VBM is plausible since the bandgap exhibits bowing behavior shown both theoretically and experimentally. ^21,28 Since the bandgap shows bowing, either the valence band maximum or the conduction band minimum, or both, should show non-linear trend. In previous studies for alloys involving chalcogens (alloy of CdSe and CdTe), it has been discussed that the origin of bowing nature of electronic band structure is the mismatch in electronegativity of the chalcogen atoms. ³⁴ Such difference in electronegativity causes compromise between Mo–S and Mo–Te bonds in terms of electron confinement which results in the bowing characteristics. It is also discussed that the valence band density of states (DOS) is predominantly determined by the chalcogens' partial density of states (PDOS). In this study, a trend where the VBM energy is determined predominantly by the chalcogen ratio is observed as shown in Fig. 2a. It was also shown that the VBM takes the same value for different chalcogen to metal ratio (Fig. 2d). With the discussions made in the literature mentioned above, along with the trends seen in this study, it is strongly suggested that the VBM is determined by the chalcogen composition of the alloy. In contrary to valence band, the conduction band minimum remains almost unchanged regardless of the Te concentration. It has been discussed in Ref. 34 that the conduction band DOS is heavily dependent on the metal PDOS. Since it is only the chalcogen atoms that are changing in number, it is expected that the conduction band edge, mainly determined by Mo PDOS, remains unaffected or the effect of chalcogen ratio change is only subtle on the conduction band edge values. The bandgap distribution that has been previously obtained is shown in Fig. 4a. The bowing parameter was calculated in Ref. 28 and was obtained to be 1.21 eV. The bowing parameter seems much larger compared to the value of 0.41 eV obtained through first principle calculations. ²⁴ This may be attributed to the fact that the theoretical calculation is for the single-layer and the samples in this study is multi-layer (approximately 10 to 15 layers). It is well known that vertical stacking of single layers to form multi-layer changes the electronic structure of TMDs. Hence it is plausible that the electronic structure change according to Te concentration is different between single-layer and multi-layer. In addition, as discussed below, there is a bowing feature observed in the c-axis lattice parameter for the multi-layer alloy. This bowing behavior in the vertical direction is one possible cause of the larger bandgap bowing in multi-layer. Combining the VBM and the bandgap distribution, the band structure shift can be obtained as shown in Fig. 4b. Such tunability of the band structure opens the opportunities to wide variety of applications. For example, controlling the band alignment and combining with different material allows more efficient tunneling of the carriers for tunnel FETs ³⁵ and carrier selective contact for solar cells ³⁶ and so on.

Zoom In Zoom Out Reset image size

Next, we will discuss the reason why there is phase separation. One clear reason is the thermal instability of the alloy. As mentioned in the introduction, previous studies show that the alloy of MoS₂ and MoTe₂ poses higher formation energy than the composing compounds which makes it thermally unstable. ^23,24 However, as it can be seen in Fig. 4c, phase separation occurs preferentially on the Te-rich side. This may be accounted to the formation of 1 T' MoTe₂ phase. It is reported that MoTe₂ forms the distorted monoclinic phase, or the 1 T' phase, more readily than MoS₂ does. ²⁵ Such formation of 1 T' occurs when the film has Te deficiency. Since the film under consideration is an alloy, it can be considered that chalcogen deficiency causes the 1 T' phase to form. Figure 5 shows the S and Te per Mo ratio for different samples and their physical characteristics (alloy or phase separation). For anneal samples, it is true that less chalcogen deficiency leads to alloy formation. However, in the case of sputter samples, some samples show no phase separation in spite of the fact that the chalcogen deficiency is well in the range where annealed samples show phase separation. This may be explained by the different temperature in each process. The annealed sample underwent higher temperature process compared to as-sputter samples. This temperature difference may have caused the discrepancy between the sputter and anneal samples. Therefore, it is crucial to reduce the chalcogen deficiency. Along with the fact that the band structure shifts only according to the chalcogen composition and not much affected by the chalcogen deficiency (mentioned in the electronic structure evaluation), as long as the film forms uniform alloy, the control of chalcogen composition allows the control of band structure. Another reason why the phase separation occurrence disagrees with the theory, no phase separation on S-rich side, is the dominance of MoS₂. Unlike MoTe₂, MoS₂ are forming 2H phase structure. It is possible that abundant S prevents 1 T' phase formation by MoTe₂. This means that as long as the 1 T' phase formation by MoTe₂ is suppressed, the alloy can be formed without phase separation unlike the prediction by the theory. In the event that phase separation cannot be suppressed in the Te-rich side, since the band structure shift is more rapid on the S-rich side, MoS_2(1−x)Te_2x alloy can be still utilized with reasonable range of bandgap value.

Zoom In Zoom Out Reset image size