Low-temperature synthesis of 2D anisotropic MoTe2 using a high-pressure soft sputtering technique

We demonstrate a high-pressure soft sputtering technique that can grow large area 1T′ phase MoTe2 sheets on HOPG and Al2O3 substrates at temperatures as low as 300 °C. The results show that a single Mo/Te co-sputtering step on heated substrates produces highly defected films as a result of the low Te sticking coefficient. The stoichiometry is significantly improved when a 2-step technique is used, which first co-sputters Mo and Te onto an unheated substrate and then anneals the deposited material to crystalize it into 1T′ phase MoTe2. A MoTe2−x 1T′ film with the lowest Te vacancy content (x = 0.14) was synthesized using a 300 °C annealing step, but a higher processing temperature was prohibited due to MoTe2 decomposition with an activation energy of 80.7 kJ mol−1. However, additional ex situ thermal processing at ∼1 torr tellurium pressure can further reduce the Te-vacancy (VTe) concentration, resulting in an improvement in the composition from MoTe1.86 to MoTe1.9. Hall measurements indicate that the films produced with the 2-step in situ process are n-type with a carrier concentration of 4.6 × 1014 cm−2 per layer, presumably from the large VTe concentration stabilizing the 1T′ over the 2H phase. Our findings (a) demonstrate that large scale synthesis of tellurium based vdW materials is possible using industrial growth and processing techniques and (b) accentuate the challenges in producing stoichiometric MoTe2 thin films.


Introduction
Two-dimensional (2D) transition metal dichalcogenides (TMDs) have attracted much attention owing to their unique quantum properties ranging from excitonics [3][4][5] to Weyl fermions 6,7 and to charge density waves. [7][8][9] For example, TMDs made from group VIB Mo and W metals, such as MoS 2 and WSe 2 are direct gap excitonic semiconductors with strong exciton binding energies as large as 0.5 eV. This allows for the stabilization of excitons even at room temperature. 3,5,[10][11][12] The K-valley in their band structure is degenerated for different spin due to broken inversion symmetry. This leads to exotic spin-valley coupling and generates exotic valleytronic properties. 13 While the S and Se containing group-VIB 2D dichalcogenides all naturally crystallize in the hexagonal (2H) phase, when metal cations bind to tellurium, new phases emerge and 1T 0 anisotropic and 2H isotropic phases start to compete. 14-17 WTe 2 is a 2D Weyl semimetal crystallizing only in the 1T 0 phase. 14,18 In contrast, 2D MoTe 2 is naturally stable both in the 1T 0 type-II Weyl fermion 17 phase as well as 2H excitonic semiconductor phase. The latter has a 1.1 eV infrared band gap and the largest exciton binding energy ($580 meV) reported to date. 15,16 This indicates that the crystalline phases of 2D dichalcogenides have similar formation energies, which result in both new, exciting properties and manufacturing challenges.
Despite their enticing properties, there has been limited progress in producing high quality, large-area MoTe 2 lms. Powder-chemical vapor deposition (CVD) is oen adopted for the large scale synthesis of sulfur and selenium based 2D TMDs. 19 In this method, target substrates are placed close to powder precursors such as MoO 3 , MoCl 5 , WO 3 , etc. and heated up to a high temperature in the presence of H 2 S or S 2 gas. 20 This method enables the synthesis of MoS 2 , WS 2 , and others, 19,21,22 but lacks precise control of gas precursor concentrations and ow, as well as lm thickness, and defect concentrations, and is thus unsuitable for producing device-quality lms needed in large-scale manufacturing. The powder-CVD synthesis of MoTe 2 requires much higher processing temperatures ($800 C) for crystallization to occur. This introduces a large number of tellurium vacancies to a level that the synthesized MoTe 2 becomes highly unstable in air. This is the reason that CVD MoTe 2 sheets are not commercially available. Considering these challenges, MoTe 2 lms have been synthesized by transforming MoO 3 or Mo thin lms into MoTe 2 at temperatures above 700 C and at high Te vapor pressures. This is commonly known as the chalcogenation process. 23,24 This process relies on the diffusion and reaction of elemental Te, but typically yields poor crystallinity materials. Since the chalcogenation occurs at a high temperature, it is not compatible with conventional manufacturing methods and cannot produce vdW superlattices, which require low processing temperatures to retain the layer's structural quality.
In this work, we have developed a low-temperature and lowenergy co-sputtering technique (Fig. 1a) to grow MoTe 2 lms over large areas. The sputtering technique allows precise thickness control by adjusting the deposition time and produce thin lms at a low processing temperature. By using high pressure growth, the proposed method reduces the kinetic energy of the impinging Mo, Te and Ar atoms to avoid surface and subsurface damage to the MoTe 2 sheets. To understand what limits the growth conditions of MoTe 2 , 25 relative sticking coefficients and decomposition rates were determined by co-sputtering Mo and Te as a function of temperature and time (Fig. 1b). To produce enhanced stoichiometric lms, Mo and Te atoms were co-deposited onto vdW graphene surfaces at room temperature, followed by in situ thermal crystallization at 350-450 C. We have further improved the stoichiometry from x ¼ 1.86 to 1.90 by ex situ thermal processing at elevated Te pressures ( Fig. 1c and d). The resulting MoTe 2 shows excellent structural and physical properties, as evidenced by Rutherford backscattering spectroscopy (RBS), Raman spectroscopy, AFM, kelvin probe force microscopy (KPFM), and 4-point resistivity measurements. Interestingly, the deposited MoTe 2 lms always crystallize in a 1T 0 type-II Weyl semimetal structure due to large sheet carrier density (originating from V Te ) driven 1T 0 phase stabilization. 1,2 The overall results optimize a new low-energy sputtering technique that can produce high-quality largearea MoTe 2 at low temperatures. Fig. 1a and 3a show the schematic diagram of the deposition chamber and the most-oen used temperature timetable, respectively. The ultra-high vacuum (UHV) chamber is evacuated to a low 10 À7 torr range prior to deposition, followed by sputtering. In our studies, elevated Ar pressures of 150 mTorr were used during the sputtering process to decrease the deposition rate and reduce the kinetic energy of the impinging species on the growth surface. The latter is important to minimize surface and subsurface damage to the growing lm, while potentially enhancing surface diffusion. The deposition rate of each element is dependent on sputter power. With increased sputter power of the Te source, we observed effects of charge build-up on the Te target, which caused rather unpredictable deposition rates. To avoid this charge buildup, the Te sputter power was xed to 5 W while using a DC sputtering source. The effects of the Mo sputter power on the crystal quality and Te/Mo ratio were studied and we optimized the conditions (Fig. S2 †).

Sticking coefficient and decomposition rate
The sticking coefficient is determined by the adsorption energy of precursor atoms (Mo and Te atoms) on the growth surface. The decomposition rate is determined by the energy of the atoms that are part of the fully reacted MoTe 2 phase, and is indistinguishable from the evaporation rate of a bulk solid. These two factors ultimately dictate growth characteristics and are two critical parameters for the establishment of a controllable growth process. 26 To determine these growth parameters and access the Mo and Te sticking coefficients and decomposition rates, the amount of Mo and Te evaporated and deposited was measured during co-sputtering onto graphite/sapphire and bare c-cut sapphire substrates at substrate temperatures ranging from room temperature to 450 C (Fig. 2a).
The typical thickness of the deposited Mo/Te lm was approximately a few nanometers, as determined by Rutherford backscattering spectroscopy (RBS). The number of Mo and Te atoms deposited on sapphire substrates per unit area is measured by RBS, then divided by the growth time to determine the deposition rates. By taking the slope of an Arrhenius plot of the deposition rate, we extracted the activation energy for Fig. 2b. Closer inspection of this gure reveals two main properties: (1) rst, it can be seen that the Mo deposition rate (red circles in Fig. 2b) is nearly independent of the substrate temperature, while the Te deposition rate signicantly decreases above 300 C. Tellurium metal has a high vapor pressure at low temperatures ($10 À2 torr at 400 C), while Mo's is very small (extrapolated to be $10 À40 torr at 400 C). The normalized Te sticking coefficient, referenced to the room temperature deposition rate, decreases with temperature to a low coefficient of 0.58 at 450 C. This results in highly Te decient MoTe 2 lms. (2) The Arrhenius plot for Te shows that there are two distinct slopes with two different activation energies (Fig. 2b blue circles). The lower activation energy (E a ¼ 4.3 kJ mol À1 ) is well-dened between 20 C and 350 C and is related to the desorption of the precursor forms of tellurium from the growth surface, oen referred to as the sticking coefficient. We attribute the higher activation energy (E a ¼ 80.7 kJ mol À1 ) to the evaporation of Te as a result of the thermal decomposition of MoTe 2Àx . Raman data collected on these lms deposited in a range of temperatures (Fig. 2c) show that crystallization occurs at 300 C and above, as evidenced by the emergent of out-of-plane A (8) g (163 cm À1 ) and A (12) g (259 cm À1 ) phonon modes. In-plane B (6) g (190 cm À1 ) modes of 1T 0 -MoTe 2 were observed at 400 C. The Raman spectra of the lms deposited on HOPG and sapphire showed similar substrate temperature dependence and Raman spectra. These Raman peaks match with the literature values. 27 This absence of Raman peaks in lms deposited below 300 C and above 450 C indicates that the associated lms do not contain signicant levels of crystalline MoTe 2 . For low temperature growth, there is not enough thermal energy available to overcome the kinetic barriers of thin lm growth (e.g. surface diffusion and reaction). And for high temperature growth, Te-poor lms are formed as a result of the high MoTe x decomposition rate.

MoTe 2 lm deposition at room temperature followed by in situ annealing
To produce a lm with better stoichiometry, we co-sputtered Mo and Te at room temperature, where the sticking coefficient is high, followed by an in situ annealing. The temperature prole of this process and corresponding deposition rate and Te/Mo ratio on the sapphire substrate are shown in Fig. 3a and b, respectively. We note that there were no observable differences in the deposition rate or Te/Mo ratio when MoTe 2 was deposited onto sapphire and HOPG substrates. Similar to the previous section, our ndings reveal that there are two activation energies for Te. One low energy desorption process and another higher energy decomposition process that dominates above 300 C. The experiments support the following observations: (1) Mo and Te atoms deposited at room temperature have a sticking coefficient near 1 and (2) annealing of the composite in the range of 300-350 C provides sufficient energy to facilitate a reaction, without experiencing a high Te desorption rate. This allows us to produce MoTe 2Àx thin lms with a high Te/Mo ratio (x < 0.3).
The degree of crystallization of 1T 0 MoTe 2 can best be monitored by observing the temperature-dependence of the Raman spectrum above 300 C (Fig. 3c). Raman spectra of the lms for HOPG and sapphire showed a similar substrate temperature dependence and Raman spectra (Fig. S2 †). The minimum crystallization temperature of $300 C is similar to that found when Mo and Te are sputtered onto heated substrates, as described above. The Raman peak width represents a gauge of the crystallinity of the lms, as defects and grain boundaries create local strain and broaden Raman peaks. The crystallinity of the lms as a function of annealing temperature was analyzed using the gure-of-merit of FWHM of the A (8) g mode, 15,17,28,29 as shown in Fig. 3d. The crystallinity of MoTe 2 lms improves with increasing annealing, and we do not nd any deterioration of MoTe 2 crystal quality up to 450 C. Thus, room temperature co-sputtering followed by an in situ annealing process can be used to grow 1T 0 -MoTe 2 lms. The thickness of the MoTe 2 lms can be precisely controlled from an order of a nanometer to tens of nanometers by adjusting the sputtering time. Fig. 3e shows the temperature dependence of the resistivity of a deposited MoTe 2 lm. Raman spectra in the previous section indicated that the deposited lms form in the 1T 0 metallic phase. However, the r-T plot shows an increase in resistivity at lower temperatures, which is characteristic of carrier concentrations lower than the metal-insulator transition. 30 This can be attributed to doping of the material by the V Te , consistent with the Tepoor stoichiometry measured with RBS.
The origin of 1T 0 -phase stabilization: charge driven 2H / 1T 0 phase transition Here we discuss why the high-temperature 1T 0 phase forms during the growth reported in this study. As shown by the phase diagram in Fig. 4a, the 2H phase is formed at temperatures lower than that of the 1T 0 -phase. When the substrate is cooled down to room temperature, the materials would thus be expected to form the 2H-phase alone. However, we observe the presence of only the 1T 0 -MoTe 2 phase, presumably as a result of the stabilization of this phase by the presence of the V Te donors. 16,31 This has previously been documented using ionic gating experiments on exfoliated 2H-MoTe 2 akes when the sheet electron density is greater than 2.2 Â 10 14 cm À2 . 14 Hall measurements on our MoTe 2 lms show that our sheets contain n-type carriers with a concentration of 4.6 Â 10 14 cm À2 . Kelvin probe force microscopy (KPFM) data collected from sputtered MoTe 2 and exfoliated 1T 0 -MoTe 2 (Fig. 4d-f) show that the work function of sputtered MoTe 2 is lower than that of exfoliated 1T 0 by 130 meV, indicating that the deposited MoTe 2 lm is decisively n-doped. For a lm with a Te/Mo ratio of 1.76, the estimated density of the double-donor V Te per layer is 2.7 Â 10 14 cm À2 , resulting in a free-electron density of 5.4 Â 10 14 cm À2 . This is also consistent with recent reports on exfoliated MoTe 2 sheets that nd that a $3% V Te concentration is sufficient to stabilize the 1T 0 -phase over the 2H-MoTe 2 phase. 1 Improving the stoichiometry by high pressure Te annealing Lastly, we show that the stoichiometry can be further improved using ex situ thermal processing. A MoTe 2 lm was deposited at room temperature, in situ annealed at 400 C and then sealed in a glass ampoule with elemental Te and heated at high temperatures (Fig. 1c). A $1 torr Te partial pressure in the ampoule was produced by keeping the Te side at 520 C. The side with the MoTe 2 lm was kept at 250 C. A thin Te overlayer was formed on the MoTe 2 lm, which was subsequently removed using H 2 SO 4 (Fig. 4b). During the treatment, excess Mo was removed by H 2 SO 4 and the Te/Mo ratio of the lm increased even though a Te overlayer was removed. Aer the treatment, the lm surface was found to be smoother (Fig. 4c), the MoTe 2 stoichiometry increased to MoTe 1.90 and the structure remained in the 1T 0 phase, but there was no noticeable change in the Raman FWHM values. Despite the improvement, this value still remains above the 3% vacancy threshold which was previously suggested for stabilizing the 1T 0 phase.

Conclusion
The overall results have established a low temperature so sputtering technique that can synthesize MoTe 2 lms. The process can produce 2D anisotropic MoTe 2 1T 0 phase lms using room-temperature co-sputtering of Te and Mo under relatively high pressure, followed by in situ annealing. This industrially compatible sputtering method offers large area growth capability, thickness control, freedom in substrate selection, and the ability to adjust the stoichiometry. Activation energies for lm decomposition were determined by measuring the incorporation rate of Mo and Te into MoTe 2 lms on heated substrates. The crystallization of these MoTe 2 lms required temperatures greater than 300 C and the deposited MoTe 2 has always crystallized in the 1T 0 anisotropic phase as a result of stabilization by using a high concentration of V Te doubledonors. Ex situ thermal annealing in a high-pressure Te environment increased the stoichiometry of the 1T 0 phase from MoTe 1.86 to MoTe 1.90 .

Conflicts of interest
There are no conicts of interest declared.