Direct Synthesis of Large-Scale Multilayer TaSe2 on SiO2/Si Using Ion Beam Technology.

The multilayer 1T-TaSe2 is successfully synthesized by annealing a Se-implanted Ta thin film on the SiO2/Si substrate. Material analyses confirm the 1T (octahedral) structure and the quasi-2D nature of the prepared TaSe2. Temperature-dependent resistivity reveals that the multilayer 1T-TaSe2 obtained by our method undergoes a commensurate charge-density wave (CCDW) transition at around 500 K. This synthesis process has been applied to synthesize MoSe2 and HfSe2 and expanded for synthesis of one more transition-metal dichalcogenide (TMD) material. In addition, the main issue of the process, that is, the excess metal capping on the TMD layers, is solved by the reduction of thickness of the as-deposited metal thin film in this work.

Introduction

Monolayer transition-metal dichalcogenides (TMDs) with unique physical properties are two-dimensional (2D) materials of a new generation following the advent of graphene. TMDs can be mainly categorized into semiconducting and metallic materials, resulting from different elemental compositions and crystal structures. Initially, the semiconducting TMDs such as 2H (trigonal prismatic)-MoS2, MoSe2, WS2, and WSe2 were widely investigated and developed for various kinds of applications since their band gaps do not change significantly under the transition from indirect to direct as the monolayer limit is reached. For example, the states of the conduction band at the K point in MoS2 are composed of d orbitals on Mo atoms and localized between S atomic layers, inferring that they are almost not influenced by the interlayer coupling. On the contrary, the states close to the Γ point in MoS2 primarily arise from the hybridized d orbitals on Mo atoms and the antibonding p orbitals on S atoms, implying that they are more sensitive to the thickness due to the strong interlayer coupling. Therefore, the indirect band gap along the K−Γ direction increases with decreasing number of layers, while the direct band gap along the K–K direction is nearly unchanged. Initially, Radisavljevic et al.[1] successfully fabricated single-layer MoS2 transistors, and Larentis et al.[2] studied back-gated MoSe2 field-effect transistors (FETs). Afterward, the WS2-based FET on a SiO2/Si substrate was announced by Georgiou et al.,[3] and the in-plane p–n diode of ambipolar WSe2 by electrostatic gating was accomplished by Baugher et al.[4] Apart from the semiconducting TMDs, the metallic ones including 1T (octahedral)-TaS2, TiSe2, and VS2 were also studied in order to realize their physical properties for particular applications.[5−9] The charge-density wave (CDW) transition, a structural distortion similar to the Peierls distortion in one-dimensional (1D) nanomaterials, has been discovered in most of metallic TMDs, especially TaSe2.[10,11]

Metallic TaSe2, naturally existing in various allotropes such as 1T, 2H, and 3R (trigonal prismatic) structures,[12] possesses complex behaviors in the CDW phase transitions.[13] Initially, it was demonstrated that the bulk TaSe2 stays at the normal state under higher temperature and goes through incommensurate CDW (ICDW) and commensurate CDW (CCDW) states in sequence with decreasing temperature.[10,11] Thus, there are two phase transition temperatures, which are different between the bulk 1T and 2H-TaSe2.[14] In recent years, the CDW phase transitions of monolayer and multilayer TaSe2, which would be distinct from those of bulks, have been extensively investigated for potential applications in the future. The current switching device driven by a voltage-controlled phase transition between CDW states of TMDs is a popular application in this field.[15] The CDW states in monolayer tantalum dichalcogenides were simulated by using density function theory (DFT) calculations.[16] The fundamental properties of monolayer and multilayer TaSe2 including band structures, crystal structures, and lattice vibration modes were theoretically predicted by Yan et al.[13] A small spin–orbit scattering length (17 nm) and a large breakdown current density (3.7 × 107 A/cm2) of exfoliated 2H-TaSe2 flakes were determined by Neal et al., indicating its high potential for the application of spintronic devices.[17] Moreover, the CDW states and superconductivity of TaSe2 crystals doped by different elements such as S, Te, Ti, and Cu have also been experimentally researched and discussed.[18−22] Interestingly, a 2H-TaSe2 thin film contacted with Ti/Au as the electrodes exhibits a current–voltage characteristic of an abrupt transition from highly resistive state to conductive one, which may be useful for the radiation-hard logic circuits.[23]

So far, most of the studies have utilized exfoliated TMDs for device fabrication, implying that the thickness of TMDs cannot be precisely controlled and the area of them is limited. The chemical vapor deposition (CVD) method was also developed for the growth of TMDs on SiO2/Si substrates.[24−26] Unfortunately, optical microscope (OM) and/or atomic force microscope (AFM) images of TMD triangular flakes obtained by CVD growth indicate that those TMD films are noncontinuous even though it was claimed that their dimensions are larger than 100 μm.[27−29] The noncontinuous film seriously prevents the real applications in industry. On the other hand, SiO2 would be corroded by S or Se, resulting in losing its insulating capability for further device fabrication; hence, a transfer step seriously deteriorating the quality of TMD films is still inevitable. Another synthesis method based on sulfurization or selenization of transition metal or transition-metal oxide films can achieve wafer-scale continuous TMD films. Actually, the quality of TMD films synthesized by this method is controversial since the crystallinity of deposited metal or metal oxide thin films is not good enough and they could not be uniformly sulfurized or selenized owing to a diffusion-limited reaction. As a result, the inhomogeneity might be the main cause leading to the TMD FETs with the carrier mobility as low as 10 cm2/(V·s).[30] Previously, we have developed an innovative process using Se ion beams to directly synthesize a uniform large-scale MoSe2 thin film on sapphire.[31] All of the analytical results confirm that the Mo film is selenized by the implanted Se ions during the thermal treatment to form MoSe2 layers on a sapphire substrate.[31] Furthermore, we have successfully expanded this technique for the synthesis of multilayer HfSe2 to investigate its band structure by using photoluminescence (PL) spectroscopy and comparing with the calculation results.[32] However, the excess metal at the surface after the process cannot be accurately removed without etching of TMDs because the etching agent with high selectivity cannot be found. In this report, the ion beam technique is applied to synthesize a large-scale 1T-TaSe2 thin film on SiO2/Si. It is noted that the thickness of the Ta thin film is reduced to overcome the issue of excess metal and the sapphire substrate is replaced with SiO2/Si in order to approach industrial applications.

Results and Discussion

The Raman spectrum of the Se-implanted Ta thin film on SiO2/Si without any peaks is shown in Figure a. After annealing, there are two Raman peaks obviously appearing in the spectrum as shown in Figure b. The Eg and A1g peaks corresponding to the in-plane and out-of-plane lattice vibrations of 1T-TaSe2 are located at ∼162 and ∼230 cm–1, respectively, which is in agreement with the computational results.[13] Herein, the positions of Raman peaks of 1T and 2H-TaSe2 should be discussed and realized. According to the literature, the Raman spectrum of 2H-TaSe2 may be consistent with that of theoretical prediction.[13,17,23,33] Nevertheless, it seems that there are discrepancies between the experimental and computational results of Raman analysis of 1T-TaSe2. Samnakay et al.[12] claimed that the E2g and A1g Raman peaks of 1T-TaSe2 emerge at positions of ∼177 and ∼187 cm–1, respectively; agreeing with the previous results.[34,35] In fact, the results from the old references cited by them only exhibit a quite weak signal close to ∼190 cm–1, which might be the A1g Raman peak indicated by them, and the E2g Raman peak at ∼177 cm–1 assigned by them is absent.[34,35] In addition, there are theoretically not any active modes in the Raman spectra of 1T and 2H-TaSe2 within the range of 170–200 cm–1.[13] As mentioned above, there were some issues concerning the Raman analysis of 1T-TaSe2 in the past. Therefore, it is reasonable that the spectrum of Figure b initially confirms the formation of 1T-TaSe2 layers on SiO2/Si via our process.

Figure 1

(a) Raman spectrum of the as-implanted sample. Inset: the sample structure. (b) Raman spectrum of the as-annealed sample. Inset: the sample structure and the illustration of lattice vibration modes of 1T-TaSe2.

Figure a shows the Kelvin probe force microscopy (KPFM) mapping of the sample after our process with the optimized parameters. In order to first obtain the work function of the probe (φp), Au was used as the reference to measure the average difference of work function (φd) between Au and the probe. Thus, the work function of the probe (φp) could be calculated for deriving that of the as-synthesized materials since the work function of Au (φAu) has been known as 5.1 eV. As can be seen in Figure a, the work function of TaSe2, deduced from the average difference of work function (φm) between TaSe2 and the probe (Supporting Information), is around 5.1 eV, close to the value extracted from the results of ultraviolet photoelectron spectroscopy (UPS).[36] Besides, the KPFM mapping also displays the good uniformity of the TaSe2 thin film synthesized by our process. The Ta 4f X-ray photoelectron spectroscopy (XPS) spectrum of the surface of TaSe2 on SiO2/Si shown in Figure b apparently reveals that there are four main peaks related to the Ta–Se and Ta–O bonding.[22] The weaker Ta–O peaks of 4f7/2 (∼27 eV) and 4f5/2 (∼29.2 eV) are caused by the inevitable oxidation under atmosphere. The intense Ta–Se peaks of 4f7/2 and 4f5/2 are located at ∼22.4 and ∼24.4 eV, respectively, implying that the TaSe2 synthesized in this work belongs to 1T rather than the 2H phase.[37] The Ta 4f photoemission spectrum of 1T-TaSe2, fitted by using the six Doniach–Sunjic line shapes, was obtained by Horiba et al.[37] Their fitting results indicate that each Ta–Se peak is composed of three subpeaks (A, B, and C subpeaks) corresponding to three Ta sites (A, B, and C sites) in the lattice. Among these subpeaks, the binding energy of A is close to that of B, while the binding energy of C is averagely ∼0.75 eV higher than those of A and B. Actually, the positions of Ta–Se peaks in Figure b conform to those in ref (37). It looks like each Ta–Se peak in ref (37) splits into two bands containing A, B, and C subpeaks; however, it seems that the three subpeaks merge together in Figure b. This dissimilarity should result from the instruments with different resolutions since the photoemission analysis in ref (37) was carried out by using an angle-resolved photoemission spectrometer (ARPES). The KPFM and XPS results further make sure that the TaSe2 synthesized in this study possesses the 1T structure.

Figure 2

(a) KPFM image of the surface of 1T-TaSe2 on SiO2/Si. (b) Ta 4f XPS spectrum of the surface of 1T-TaSe2 on SiO2/Si.

The cross-sectional transmission electron microscopy (TEM) image in Figure a clearly shows a uniform bilayer on the SiO2/Si substrate. The top layer should be the multilayer 1T-TaSe2 with a thickness of ∼10 nm, while the bottom one may be the damaged SiO2 layer, caused by the ion beam irradiation, with a thickness of less than 10 nm. The higher magnification image exhibits that multilayer 1T-TaSe2 with an interplanar spacing of ∼0.64 nm is polycrystalline. The SRIM (the stopping and range of ions in matter) simulation of the depth distribution of Se in the 10 nm-Ta/SiO2 structure is in agreement with the TEM result. As can be seen in Figure b, almost all of the Se ions are accumulated in the 10 nm-Ta thin film owing to the low implantation energy. Only a few Se ions penetrate into the SiO2, resulting in the damaged layer. Figure c shows the high-resolution (HR) TEM image of 1T-TaSe2 from the region surrounded by the dashed line in Figure a. The diffraction pattern processed by fast Fourier transform (FFT) in the inset of Figure c indicates that the zone axis is [210]. The simulation of the HRTEM image and diffraction pattern of 1T-TaSe2 in Figure d seems to be consistent with that in Figure c. The interplanar distances derived from the diffraction patterns in Figure c,d should be utilized for further structure identification. As a result of calculations (Supporting Information), the real (1̅20) and (001) interplanar distances, which are equal to ∼0.18 and ∼0.64 nm, respectively, are in agreement with the theoretical values (∼0.174 and ∼0.627 nm). The lines representing the (1̅20) and (001) plane groups in the diffraction pattern of Figure c are perpendicular to each other (Supporting Information), corresponding to the angle between the (1̅20) and (001) planes that is certainly 90° as shown in the atomic model of Figure e. Moreover, the interplanar spacing (∼0.64 nm) of multilayer 1T-TaSe2 directly estimated from the TEM image of Figure a implies that these crystal planes belong to the (001) plane groups. Combining the Raman and XPS spectra together with KPFM and TEM results, we can conclude that the polycrystalline thin film synthesized in this study is indeed multilayer 1T-TaSe2.

Figure 3

(a) Cross-sectional TEM image of 1T-TaSe2 on SiO2/Si. (b) Depth profile of Se concentration in the as-implanted Ta thin film on SiO2 simulated by the SRIM. (c) Cross-sectional HRTEM image of 1T-TaSe2 from the region surrounded by dash line in panel (a). Inset: the diffraction pattern of 1T-TaSe2. (d) Simulation of cross-sectional HRTEM image of 1T-TaSe2. Inset: the simulation of diffraction pattern of 1T-TaSe2. (e) Atomic model of 1T-TaSe2.

Beyond the material identification, the relationship between resistivity (ρ) of multilayer 1T-TaSe2 and temperature (T) was measured in order to comprehend the electronic phase transition of 1T-TaSe2 obtained in this study. Figure a shows the temperature dependence of resistivity of 1T-TaSe2 on SiO2/Si below room temperature. The resistivity of multilayer 1T-TaSe2 at 5 K, which is ∼2.2 × 10–1 mΩ·cm, initially remains nearly the same with temperature until ∼30 K. Then it linearly increases with temperature to ∼2.55 × 10–1 mΩ·cm as the room temperature is reached. For the measurements above room temperature (Figure b), another sample synthesized by our process under the same conditions was measured in another system. Although the resistivity value at 300 K in Figure b is a little different from that in Figure a, the inconsistency between them is reasonable and acceptable for different samples. In Figure b, the resistivity of multilayer 1T-TaSe2 slightly increases with temperature from 300 to 450 K. It is kept at a maximum value within the range of 450–500 K and then decreases as the temperature is higher than 500 K, suggesting that the transition temperature of normal metal-CCDW of the multilayer 1T-TaSe2 is around 500 K. Roughly, the tendency of the ρ–T curve in Figure is similar to that in ref (20). Consequently, the multilayer 1T-TaSe2 obtained in this work similarly behaves as the bulk one in the electronic phase transition.[20]

Figure 4

(a) Temperature dependence of resistivity of 1T-TaSe2 on SiO2/Si below room temperature. (b) Temperature dependence of resistivity of 1T-TaSe2 on SiO2/Si above room temperature.

Conclusions

In conclusion, the multilayer TaSe2 has been successfully synthesized on a SiO2/Si substrate by using the ion beam technique for the first time. The results of material analysis demonstrate that the TaSe2 obtained by our method possesses the 1T structure indeed. The temperature dependence of resistivity indicates the behavior in the electronic phase transition in which the normal metal-CCDW transition temperature is around 500 K. In this study, it is worth noting that the process based on ion implantation for the synthesis of TMDs has been improved by reducing the thickness of the as-deposited Ta film to prevent the excess Ta at the surface after annealing.

Experimental Section

A 4 in. Si (100) wafer capped by 100 nm SiO2 was chosen as the substrate on which an ∼10 nm Ta thin film was deposited by radio frequency (RF) sputtering at room temperature. Then the Se ions (10 keV, 1.5 × 1016 ions/cm2) created from Se pellets were implanted into the Ta/SiO2/Si wafer by a medium current ion implanter (ULVAC, IMX3500). The dose and energy of implantation were estimated by the SRIM simulation in advance. Eventually, the cut samples were annealed at 900 °C with a temperature increasing rate of 20 °C/min in the N2/H2 (6/1, v/v) ambiance for 1 h. A confocal micro-Raman spectrometer (HORIBA, LabRAM, HR800 UV) with a He-Ne laser of 632.8 nm wavelength and 35 mW power was employed to initially check the formation of TaSe2 layers. The Kelvin probe force microscopy (KPFM) of the peak force mode in an atomic force microscope (AFM, Bruker, Dimension Icon) was used to measure the work function of the as-synthesized TaSe2. An X-ray photoelectron spectrometer (XPS, Ulvac-PHI, VersaProbe II) was utilized to analyze the surface composition. The nanoscale layer structure was observed by using a spherical-aberration corrected transmission electron microscope (TEM, JEOL, JEM-ARM200FTH) with 0.1 nm resolution of the lattice image and 200 kV accelerated voltage. The resistivities of the multilayer TaSe2 were measured by using the van der Pauw method in the two systems of Hall measurement. The first system (Lake Shore, HMS-9709A) was applied to the measurement at temperatures below 300 K, while the other (Ecopia, HMS-3000) was used for those higher than room temperature. Indium balls were used as the electrodes for the Hall measurements.