Changes in structure and chemical composition of α-MoTe₂ and β-MoTe₂ during heating in vacuum conditions

Transition metal dichalcogenides (TMDCs) have recently attracted remarkable attention for a number of applications,^1–8) because TMDCs have a finite energy band gap, which graphene lacks. Among TMDCs, MoTe₂ has a unique property: a thermally induced structural phase transition. The diamagnetic semiconductor α-MoTe₂ is the polytype that is stable below 815 °C, and it has a 2H-type trigonal prismatic unit layer structure [Fig. 1(a)].^9–17) In contrast, the paramagnetic semimetal β-MoTe₂ is stable above 900 °C, and it has a distorted CdI₂-type unit layer structure with one-dimensional zigzag chains of Mo atoms [Fig. 1(b)], the so-called 1T'-type structure.^18–20) α-MoTe₂ has a direct band gap of 1.1 eV in monolayer form, which is smaller than that of other TMDCs such as MoS₂ and WSe₂ (1.8 and 1.6 eV, respectively), and it has been utilized as the channel material of field-effect transistors (FETs).^21–26) On the other hand, β-MoTe₂ is expected to be a stable source material to realize a quantum spin Hall-effect device.²⁷⁾ Usually, artificial MoTe₂ single crystals are grown by a chemical vapor transport (CVT) method.²⁸⁾ MoTe₂ powder and a halogen (Br₂ or I₂) are sealed in a quartz ampoule under vacuum, and the ampoule is heated in a tube furnace containing a temperature gradient. Then, volatile molybdenum halides and tellurium diffuse in the gaseous form from the hot part of the ampoule (T = T_H) and crystallize as single-crystalline MoTe₂ at the cold part of the ampoule (T = T_L). One can obtain single crystals of α-MoTe₂ by slowly cooling the quartz ampoule after the CVT growth.²⁸⁾ In order to obtain β-MoTe₂ single crystals, the ampoule must be quenched from a temperature higher than 900 °C to room temperature as quickly as possible.²⁸⁾

Zoom In Zoom Out Reset image size

In the case of FET application of TMDC, it is crucial to grow a high-quality insulator film on the surface of the TMDC. The simplest approach to growing an insulating oxide on a semiconducting material is thermal oxidation of the surface. Thus, we have much interest in the surface reaction of TMDCs under various environments. Our collaborators recently reported the surface reaction of WSe₂ at elevated temperatures under the exposure to ozone.²⁹⁾ Layer-by-layer oxidation of WSe₂ up to trilayers from the surface was observed at higher temperatures. In the case of MoTe₂, not only the surface reaction itself, but also the difference in the reaction process of α- and β-MoTe₂, which have different crystal and electronic band structures, is also of great interest.

It was recently reported, however, that α-MoTe₂ single crystals, which were grown by a flux method, revealed a "reversible phase transition".³⁰⁾ The authors claimed, based on their X-ray diffraction (XRD) measurements, that α-MoTe₂ transformed to β-MoTe₂ when it was annealed above 500 °C in the XRD sample chamber, and recovered the α-MoTe₂ structure during the slow cooling from the high temperature. Keum et al.³⁰⁾ showed only XRD patterns measured during the heating of α-MoTe₂ to prove the phase transition, and no side reaction during the measurement was reported. However, the XRD patterns were recorded using an "Anton-Paar domed hot stage attachment", which only enables the vacuum condition of about 1 × 10⁻² mbar.³¹⁾ In such a low vacuum, MoTe₂ samples may be degraded during the high-temperature XRD measurement. Furthermore, there were two doubtful points in the XRD data shown in the supporting information of Keum et al.³⁰⁾ First, the 002 diffraction peak of 2H α-MoTe₂ disappeared above 500 °C without the appearance of a 002 peak of 1T' β-MoTe₂. Second, the XRD peak that appeared around 2θ = 26.1° above 500 °C was labelled as "004 of 1T' phase", but the 004 peak should appear at 2θ = 25.74° if the diffraction angle is calculated from the crystallographic parameter of β-MoTe₂, c = 1.386 nm.¹⁸⁾ These two problems suggested to us that the material formed during the high-temperature XRD measurement was not 1T' β-MoTe₂. Here, we report verification of the thermal transition of α- and β-MoTe₂ single crystals in vacuum conditions using our high-temperature XRD system. In order to check the surface composition and chemical states, X-ray photoelectron spectroscopy (XPS) measurements were also performed.

Our XRD (Bruker D8 Advance) is equipped with a high-vacuum-type hot stage (Anton-Paar HTK 1200N), which enables a better vacuum environment (1 × 10⁻⁴ mbar)³²⁾ than the dome-type attachment. α- and β-MoTe₂ single crystals were grown by CVT using Br₂ as the transport agent. MoTe₂ source powder was synthesized from Mo powder (99.9+%, Nilaco) and Te shots (99.9999%, Nilaco) by heating them at 900 °C for 1 week in a vacuum-sealed quartz ampoule. α-MoTe₂ was grown under the temperature gradient of T_H = 900 °C and T_L = 700 °C for 1 week, and the ampoule was slowly cooled after the growth. β-MoTe₂ was grown under T_H = 1000 °C and T_L = 900 °C for 1 week, and the ampoule was quickly quenched in water. For the XRD measurement, each single-crystal flake (∼1 mm wide) was placed on a small SiO₂ (285 nm)/Si (500 µm) wafer and set onto the alumina hot stage of the HTK 1200N. The SiO₂/Si wafer blocks diffraction from the alumina stage. During the XRD measurement, the HTK 1200N was evacuated by a rotary pump (RP), or a combination of a RP and a turbo molecular pump (TMP). The vacuum pressures during the measurement were around 1 × 10⁰ Pa and 1 × 10⁻² Pa using the RP and RP–TMP combination, respectively. Under these vacuum conditions, oxygen partial pressures were estimated to be 2 × 10⁻¹ Pa and 2 × 10⁻³ Pa, respectively. Hereafter, these vacuum conditions are abbreviated as "∼10⁰ Pa" and "∼10⁻² Pa", respectively. During the high-temperature XRD measurement, the sample temperature setting was increased by 10 °C/min and decreased by 2 °C/min between 25 and 650 °C.

Figure 2 shows XRD patterns of α- and β-MoTe₂ single crystals before and after the annealing at 650 °C in vacuum conditions of ∼10⁻² and ∼10⁰ Pa. Before the annealing, XRD patterns of both crystals [patterns (A), (C), (E), and (G)] agreed well with those in the literature.^9,18) In the vacuum of ∼10⁻² Pa, the intensity of the 002 and 004 diffraction peaks of both crystals [in patterns (B) and (F)] became smaller after the annealing, which suggests the decomposition of crystals. On the other hand, in the vacuum of ∼10⁰ Pa, the 002 peak of α-MoTe₂ in the pattern (D) became very weak and that of β-MoTe₂ in the pattern (H) almost disappeared. In the pattern (D) of α-MoTe₂, a new peak appeared at the right side of the weakened 004 peak. In the pattern (H) of β-MoTe₂, in contrast, only one peak existed around 2θ = 26°, and its position was almost the same as that of the above-mentioned new peak of α-MoTe₂ [see pattern (H) and copied pattern (D) in Fig. 2]. Here, it must be noted that the 2θ angles of these peaks were slightly larger than that of the 004 peak of β-MoTe₂, as indicated by a vertical red line from (H) to (G) in the figure.

Zoom In Zoom Out Reset image size

In order to observe the peak positions in detail, XRD patterns with a narrow 2θ range (25.0–26.2°) including the 004 diffraction peaks of α- and β-MoTe₂ were measured, as shown in Fig. 3. They were observed during the heating and cooling of α- and β-MoTe₂ crystals between 25 and 650 °C in each vacuum condition. Differences in the full width at half maximum (FWHM) of XRD peaks observed before the heating mainly came from the fluctuation in the thickness of each single-crystal flake inside of the measured area; a more uniform thickness of the sample flake makes the FWHM narrower. In the case of α-MoTe₂ in ∼10⁻² Pa [group (A)], shifts of the 004 peak due to thermal expansion were clearly observed, but no new peak appeared during the observation. The amount of the c-axis thermal expansion of α-MoTe₂ during the heating from 25 to 650 °C was calculated to be 0.010 nm using the 004 peak shift. We could find no previously measured high-temperature XRD data for α-MoTe₂, but this expansion value is close to the reported values for 2H-MoS₂, 2H-MoSe₂, and 2H-WSe₂.³³⁾ Therefore, our interpretation is that the observed 004 peak shift during the annealing really comes from the thermal expansion of the α-MoTe₂ single crystal.

Zoom In Zoom Out Reset image size

In the case of β-MoTe₂ in ∼10⁻² Pa [group (C)], small peaks could be seen around 2θ = 25.4° after the annealing. These peaks may come from desorption of Te, or from a partial phase transition from the metastable β- phase to the stable α-phase during the cooling. In the cases of annealing α-MoTe₂ and β-MoTe₂ in ∼10⁰ Pa [group (B) and (D), respectively], new peaks could be observed around 2θ = 25.9–26.0°, and these peaks were distinctly different from the 004 peak of β-MoTe₂ (2θ = 25.74°).¹⁸⁾ Two separate peaks could be recognized in each XRD pattern of β-MoTe₂ during the heating at 550 and 650 °C in ∼10⁰ Pa.

From these XRD observations, it is clearly concluded that the reported phase transition from 2H (α) to 1T' (β) in Keum et al.³⁰⁾ is the misinterpretation of XRD data. Keum et al.³⁰⁾ possibly assumed that the new peak appearing around 2θ = 26° in the XRD of α-MoTe₂ heated above 500 °C is the 004 peak of 1T' β-MoTe₂, but our observations are inconsistent with this assumption. In the literature, it was reported that the Cu Kα XRD pattern of MoO₂ shows an intense peak of 110 and 011 diffractions at 2θ = 26.005° (ICDD 01-074-6246).³⁴⁾ Therefore, it is supposed that the new XRD peak originates from the bulk MoO₂ phase produced by the oxidation of MoTe₂ during the annealing in a poor vacuum, ∼10⁰ Pa. Here, the new "MoO₂" peak observed for β-MoTe₂ annealed in ∼10⁰ Pa had a slightly smaller diffraction angle than that of α-MoTe₂. This shift may be caused by the further oxidation of MoO₂ derived from β-MoTe₂ to Mo₂O₅ or MoO₃. Otherwise, the distorted CdI₂ structure of β-MoTe₂ may have resulted in a somewhat different MoO₂ structure after the oxidation in ∼10⁰ Pa.

In order to further characterize the oxidized molybdenum phase, XPS spectra of α-MoTe₂ crystals were measured before and after the annealing of the XRD measurement at 650 °C in ∼10⁻² and ∼10⁰ Pa. Spectra were measured using a Kratos AXIS Nova equipped with a monochromatic Al Kα X-ray source, and the pass-energy of the spectrometer was set to 160 or 20 eV for the survey or core spectrum measurement, respectively. In Fig. 4, (A) shows the survey spectra, (B) shows the Mo 3d core region spectra, and (C) shows the Te 3d core region spectra. In Fig. 4(A), the three spectra are each normalized by the height of the Mo 3d peak. In Fig. 4(B), the intensities of spectra (b) and (c) are multiplied by 1.5 compared to that of the spectrum (a) to enlarge observed peaks. In Fig. 4(C), similarly, the intensities of spectra (b) and (c) are multiplied by 30 compared to that of the spectrum (a).

Zoom In Zoom Out Reset image size

As shown in Fig. 4(A), the intensities of Te-related peaks were drastically reduced after the vacuum annealing, and in ∼10⁰ Pa Te almost disappeared from the surface. Instead, large peaks of O 1s appeared after the annealing, which suggests partial decomposition and the surface oxidation of α-MoTe₂ crystals. The Mo 3d core spectra, shown in Fig. 4(B), also suggest that α-MoTe₂ close to the crystal surface was decomposed by the annealing, since the Mo 3d_5/2 peak coming from MoTe₂, which had appeared at the binding energy (BE) of 228.5 eV,³⁵⁾ became very weak after the annealing in ∼10⁻² Pa, and completely disappeared after the annealing in ∼10⁰ Pa [see inset in Fig. 4(B)]. After the annealing in ∼10⁻² Pa, a large Mo 3d_5/2 peak originating from MoO₂^35,36) could be seen at BE = 229.6 eV, which suggests the oxidation of surface molybdenum to MoO₂, although no bulk MoO₂ phase was detected by XRD. In the higher BE region, another component of Mo 3d existed, labeled as "MoO_3−x". This component had a little higher binding energy after the annealing in ∼10⁰ Pa than in ∼10⁻² Pa, but this BE in ∼10⁰ Pa was still slightly smaller than that of the Mo 3d_5/2 peak of MoO₃, which should appear at BE = 232.65 eV.^35,36)

From these XRD and XPS results, it is concluded that the annealing in ∼10⁻² Pa decomposed only the surface of α-MoTe₂ and the surface Mo mainly formed MoO₂. Some Mo atoms were further oxidized and a complicated mixture of MoO₂, Mo₂O₅, and MoO₃ was composed on the surface. Nonstoichiometric oxides might be included, too. In ∼10⁰ Pa, almost the whole α-MoTe₂ crystal was oxidized to MoO₂, which was detectable by XRD, and its surface had a little more MoO₃ than in ∼10⁻² Pa. Additionally, Fig. 4(C) also reveals that the remaining Te formed MoTe₂ and TeO₂^35,36) in ∼10⁻² Pa, and almost all Te atoms desorbed from the surface in ∼10⁰ Pa. Here, the two peaks coming from MoTe₂ in the spectrum (b) of Fig. 4(C) look larger than the above-mentioned MoTe₂ peaks in Fig. 4(B). This is because the intensity of spectra (b) and (c) of Fig. 4(C) is multiplied by 30, and the sensitivity of Te 3d is far larger than that of Mo 3d.

These results clearly prove the impossibility of the phase transition of α-MoTe₂ to β-MoTe₂ during the heating in the vacuum condition. As for the surface reaction of the α-MoTe₂ single crystal in ∼10⁻² Pa, the oxidation of Mo and Te observed by XPS might occur not during the high-temperature XRD measurement, but when the sample was exposed to air while transferring samples from XRD to XPS. It is obvious from XRD measurement results that α-MoTe₂ was almost completely oxidized to MoO₂ during the annealing in ∼10⁰ Pa. In order to thoroughly investigate the surface reaction in the high vacuum condition, experiments of in situ vacuum annealing of MoTe₂ samples just before the XPS measurement is under preparation.

In conclusion, we characterized changes in the structure and composition of α- and β-MoTe₂ single crystals during the heating in low and high vacuum conditions by XRD and XPS measurements. Contrary to the results of a recently published paper,³⁰⁾ α-MoTe₂ never transformed to β-MoTe₂ at the elevated temperature; in a low vacuum (∼10⁰ Pa) it was mainly oxidized to MoO₂, and in a high vacuum (∼10⁻² Pa) it partially decomposed and its surface was oxidized to MoO₂. The reported "reversible phase transition"³⁰⁾ of α-MoTe₂ is the misinterpretation of XRD data.

Acknowledgments