Report

Phase patterning for ohmic homojunction contact in MoTe2

See allHide authors and affiliations

Science  07 Aug 2015:
Vol. 349, Issue 6248, pp. 625-628
DOI: 10.1126/science.aab3175

Making better contacts

A key issue in fabricating transistors is making a good electrical contact to the semiconductor gate material. For two-dimensional materials, one route is through a phase transition that converts a hexagonally packed semiconductor phase into a distorted octahedrally packed metallic phase. Cho et al. show that laser heating of molybdenum telluride (MoTe2) achieves this conversion through the creation of Te vacancies. The phase transition improves charge carrier mobility while maintaining the low resistance necessary for improved transistor function.

Science, this issue p. 625

Abstract

Artificial van der Waals heterostructures with two-dimensional (2D) atomic crystals are promising as an active channel or as a buffer contact layer for next-generation devices. However, genuine 2D heterostructure devices remain limited because of impurity-involved transfer process and metastable and inhomogeneous heterostructure formation. We used laser-induced phase patterning, a polymorph engineering, to fabricate an ohmic heterophase homojunction between semiconducting hexagonal (2H) and metallic monoclinic (1T’) molybdenum ditelluride (MoTe2) that is stable up to 300°C and increases the carrier mobility of the MoTe2 transistor by a factor of about 50, while retaining a high on/off current ratio of 106. In situ scanning transmission electron microscopy results combined with theoretical calculations reveal that the Te vacancy triggers the local phase transition in MoTe2, achieving a true 2D device with an ohmic contact.

Despite the promise of using two-dimensional (2D) atomic crystals in device applications (1, 2), the issue of Schottky contact between the semiconducting 2D channel material and the metal electrode has been studied extensively without being resolved. The solution to the Schottky contact issue for silicon is local doping near silicon-metal junctions to reduce the depletion region width and to decrease the contact resistance (3). The 3D doping method used in silicon technology is unavailable in 2D devices, and previous efforts, such as aligning the metal work function with the conduction/valence band edge of 2D semiconductors (4), inserting graphene in the junction by the transfer method (5, 6), and MoS2 phase-engineering by chemical doping (7), showed inherent limitations in the processes. These limitations include a lack of true ohmic contact caused by a large interface resistance between the 2D semiconductor and the metal, and the formation of metastable and inhomogeneous junctions between 1T-MoS2 and the metal.

Global structural phase transition in transition metal dichalcogenides (TMDs) has been related to diverse areas of physics, such as metal/band insulators (2), superconductors (810), charge density waves (11), spintronics, and valley optoelectronics (12). Recently, considerable academic interest has focused on atomic-scale observations of local phase transitions in TMDs by using transmission electron microscopy (TEM), highlighting the potential engineering of local material properties (13, 14). However, the electrical and chemical properties of a phase-modified area, or a phase interface between intrinsic and modified areas in TMDs, have not yet been clearly verified because of instrumental limitations (14) or inhomogeneous metastable junctions (7).

Local phase transitions in 2D TMDs enable the fabrication of high-quality heterophase structures that are promising as a buffer contact layer to construct an ohmic contact. If the work function match between semiconducting and metallic phases in a TMD material guarantees the band alignment of ohmic contact (4), then the metal-semiconductor homojunction would be an ideal contact for 2D devices. Previous studies revealed that the energy difference between semiconducting hexagonal (2H) and metallic distorted octahedral (1T') structures in MoTe2 is substantially smaller (~35 meV) than the comparable energy differences in other TMD materials (15, 16). In this respect, MoTe2 offers the opportunity for creating stable homojunction contacts by phase transition, in contrast to its MoS2 counterpart, which has a metastable metallic 1T phase (7). In addition to the feasibility of phase transition, a band gap of ~1.0 eV in 2H-MoTe2, which is similar to that of bulk silicon, is appealing for 2D electronics (17, 18).

We report laser-driven phase patterning, which is a type of local polymorph engineering, in MoTe2 to realize a heterophase homojunction that shows ohmic contact in MoTe2 transistors. The phase transition from the 2H to 1T' phase in MoTe2 is driven by laser irradiation in a desired area. We named this robust design of heterophase structures “phase patterning.” We verified that the lattice vibration modes and chemical states of the phase-patterned areas correspond to those of single-crystalline 1T'-MoTe2. The heterophase homojunction between 2H- and 1T'-MoTe2 is stable up to 300°C, indicating that this phase patterning should be compatible with most semiconductor manufacturing processes. We also identified that the formation of the Te vacancy is the key origin of the unidirectional phase transition from 2H to 1T' phase via laser irradiation.

We synthesized 2H- and 1T'-MoTe2 single crystals by the flux method, as previously reported (16). The two structures (2H and 1T') of MoTe2 are schematically described in Fig. 1, A and B, in which six Te atoms constitute two triangles in the same orientation (2H) and in the opposite orientation (1T') around the centered Mo atom. Mechanically exfoliated flakes of multilayer (~30 layers in Fig. 1C) 2H-MoTe2 were used for the laser-irradiation–driven phase patterning in the desired area [figs. S1 and S2 (19)]. The optical images in Fig. 1C and the Raman spectra in Fig. 1D revealed the effect of the laser irradiation with a spatial resolution of less than 1 μm. The laser irradiation decreased the flake thickness and led to a new structural phase evolution in the irradiated region. The initial 2H-MoTe2 flakes exhibited two distinct Raman modes, an in-plane E2g mode near 235 cm−1 and an out-of-plane Ag mode near 174 cm−1 (20), whereas the laser-irradiated areas showed new peaks near 124, 138, and 272 cm−1, corresponding to the Ag mode of 1T'-MoTe2 (16). The evolution of Raman modes at the representative stages shown in Fig. 1D, from the 2H phase (bottom) to the 1T' phase (top), manifests the local phase transition in MoTe2. Other possible origins for the Raman mode changes were excluded by a comparison experiment with hexagonal boron nitride (h-BN)-covered MoTe2 [figs. S3 and S4 (19)].

Fig. 1 Laser-driven phase patterning.

Lattice structures of (A) 2H and (B) 1T'-MoTe2. Green spheres are Mo atoms, and orange spheres are Te atoms. (C) Optical microscope images of a mechanically exfoliated 2H-MoTe2 flake before (top) and after (bottom) laser irradiation. Patterned and nonpatterned areas are marked by 1T' and 2H, respectively. (D) The evolution of Raman spectroscopy from 2H phase (bottom) to 1T' phase (top). (E) Schematic representation of the laser-irradiation process.

Figure 1E shows a schematic of the laser-irradiation process. Although the confocal Raman spectroscopy cannot provide information about the vertical location of the 1T' phase in the MoTe2 flake, we verified from surface-sensitive x-ray photoemission spectroscopy that the phase transition initially occurs at the top layer, and the thinning effect follows the phase transition in the flake, as represented in Fig. 1E. The phase transition begins around a temperature of 400°C, estimated by the 2H-MoTe2 Raman peak shift [fig. S5 (19)]. The irradiated region in the 2H-MoTe2 flake continued to thin during the laser irradiation until the flake thickness reached a few layers of the transformed 1T' phase. We deduce that these few layers remained because of the heat-sink effect toward the bottom of the SiO2 substrate during the irradiation process (19, 21).

The phase patterning can resolve the Schottky contact issue in 2D transistors (Fig. 2). Before depositing a metal electrode in the fabrication process of the 2H-MoTe2 transistor, the 1T' phase was patterned only near the area where metal electrodes would be located (Fig. 2A). To demonstrate the effect of the heterophase homojunction in Fig. 2A, adjacent 2H-MoTe2 flakes on the same SiO2 substrate were used to fabricate transistors without the 1T' buffer layer. Figure 2, B and C, show the geometry of a device in which we conducted phase patterning and the evidence of the phase change from 2H to 1T' in the desired areas by 2D confocal Raman mapping. The two circled areas in the Raman 1T' map are located at the deposition sites of the metal electrodes.

Fig. 2 Transport through a heterophase homojunction structure in MoTe2.

(A) Schematic diagrams of a device with a 1T'/2H phase homojunction. (B) AFM image of a device with the 1T'/2H phase homojunction in MoTe2. (C) Raman mapping images of 1T' (Ag) and 2H (E2g) vibrational modes in the device channel in (B). (D) Source-drain current ISD characteristics for gate voltage VG ranging from –60 V to 60 V. (E) Arrhenius plots of the conductance. (F) Field-effect mobility as a function of temperature.

The consequences of this new homojunction in the MoTe2 transistor are summarized in Fig. 2, D to F. The n-type transistor characteristics (Fig. 2D) showed ohmic contact behavior with current Ion/Ioff of 106 (fig. S6). Furthermore, the robust heterophase homojunction structure in the MoTe2 transistor allows a high-temperature transport up to 300°C, where operation is still stable. The Arrhenius plots reveal the substantially decreased energy barrier height in the phase-patterned device (1T' contact in Fig. 2E). The energy barrier extraction from the Arrhenius plot provides direct evidence of the true ohmic contact, negligible energy barrier height, in phase-patterned devices. Accordingly, the carrier field-effect mobility is enhanced by a factor of ~50 compared with that of the device with a conventional 2H-MoTe2/metal contact [Fig. 2F, fig. S7, and table S3 (19)].

The chemical states of the Mo and Te atoms in the phase-patterned MoTe2 were investigated by scanning photoelectron microscopy (SPEM) with a spatial resolution of 200 nm. The binding energies of the Mo and Te 3d electrons measured in 2H- and 1T'-MoTe2 single crystals (Fig. 3A) revealed an energy shift of 0.6 eV between the 2H- and 1T'-MoTe2 samples that can be explained by the different lattice symmetry of 2H and 1T'-MoTe2. A similar energy shift, 0.4 eV, was seen between patterned and nonpatterned areas (Fig. 3B). Thus, the phase-patterned area has chemical states of Mo and Te atoms similar to the 1T'-MoTe2 single crystal.

Fig. 3 SPEM study of phase-patterned MoTe2.

(A) Photoemission spectra of Mo 3d and Te 3d electrons in single-crystalline 1T'- (red) and 2H-MoTe2 (blue). Green and dark cyan curves are MoTe2 and MoO3 (or TeO2) fitting, respectively. (B) Photoemission spectra of Mo 3d and Te 3d electrons in patterned (1T', red) and nonpatterned (2H, blue) areas. AFM images (C) before and (D) after laser irradiation, with a white rectangle indicating the illuminated region. Binding energy map of Mo 3d electrons for the (E) 2H phase and (F) 1T' phase in the same region as (D).

Atomic force microscopy (AFM) images with a rectangle indicating the laser-irradiated area (Fig. 3, C and D) show a thinning effect after laser irradiation. The sample shown in Fig. 3D was used for the SPEM study. The SPEM mapping images in Fig. 3, E and F, were obtained with photoelectron counts in a selective window of binding energies for each phase, which reveal the spatial distribution of the 2H (Fig. 3E) and 1T' (Fig. 3F) phases, respectively. The surface-sensitive SPEM shows only the 1T' phase in the patterned area. By noting that Raman spectroscopy shows a mixed phase of 2H and 1T', we support the argument that the 1T' phase is located in the upper layer (Fig. 1E). Furthermore, no MoO3, TeO2, or other elements were observed in the SPEM results [figs. S8 and S9 (19)], which is consistent with the Raman spectroscopy results in Fig. 1. This demonstrates a clean heterophase homojunction structure without oxides or other elements by our phase patterning, which is an advantage compared with the precedent of metastable phase engineering with MoS2 by a chemical method (7).

The structural transformation process during phase patterning was clarified by in situ scanning transmission electron microscopy (STEM) with a monolayer of 2H-MoTe2 (Fig. 4). Of the two major physical processes, local heating and valence electron excitation related to laser irradiation, we explored the former on the atomic scale. In Fig. 4A, the atomic images obtained at T = 400°C, similar to the estimated temperature during laser irradiation, primarily show Te atoms of 2H-MoTe2. Although Te atoms can be sublimated at T = ~400°C (16), the monolayer 2H phase region remained robust without generating Te vacancies (Fig. 4A). To slightly stimulate the temperature effect, we used a scanning electron beam irradiation (with a beam size of 1 Å and a beam current of 20 pA). Then, low-density atomic Te defects were created (Fig. 4B), and, more important, a clear sign of the structural phase transition from the 2H to 1T' phases was observed in Fig. 4C, which is a fast Fourier-transform (FFT) image of Fig. 4B. Although the hexagonal symmetry remains, as shown by the lattice symmetry of 2H-MoTe2, extra periodic spots with a rectangular symmetry appear in Fig. 4C. The rectangular lattice symmetry is a feature of distorted octahedral (1T') MoTe2 (16, 22).

Fig. 4 In situ STEM observation of Te-defect–driven structural phase transition in a monolayer of MoTe2 (T = 400°C).

(A) Atomic image of a monolayer of 2H-MoTe2. Bright spheres are Te atoms with hexagonal symmetry. (B) Atomic Te vacancies created artificially. Te single vacancy and divacancy are visible and marked by 1 and 2, respectively. (C) FFT image of (B). (D) Filtered high-resolution image near a Te vacancy showing the splitting of the Te atoms. (E) Atomic resolution image of a Te divacancy defect. (F) The energy differences between the 2H and 1T' phases as a function of the Te vacancy concentration from the DFT calculation.

In contrast to 2H-MoTe2, in which two Te atoms completely overlap in the top view of the crystal in the STEM images, 1T'-MoTe2 should show split Te positions in the top view STEM image (Fig. 1A). The initial stage of splitting the Te atom positions is captured in Fig. 4D, which is a filtered STEM image to clearly show the Te atoms. As marked by two circles and an arrow, the Te atoms start splitting but do not yet completely reach the 1T' phase over the entire area. No such split or rectangular symmetry in FFT was observed without Te vacancies in the monolayer of 2H-MoTe2. This result indicates that the phase transition originates from the Te vacancies at an elevated temperature, either by laser irradiation or by a heating stage. An atomic vacancy is shown in Fig. 4E.

Our density functional theory (DFT) calculations that explain the phase transition by the Te vacancy are shown in Fig. 4F. To select the most stable phase at a low temperature, the relative binding energy per unit formula between the 2H and 1T' phases is plotted as a function of Te vacancy concentration in Fig. 4F. It is clear that a Te monovacancy concentration exceeding 3% causes the 1T' phase to be more stable than the 2H phase, which is qualitatively consistent with our experimental results. Furthermore, the band alignment of ohmic contact at the homojunction of the 2H and the 1T' phases of MoTe2 was verified by our DFT calculations (19). Compared with the Schottky contact between 2H-MoTe2 and the Au electrode, the interlayer charge transfer across the homojunction in MoTe2 causes only a slight energy difference between the electron affinity of 2H-MoTe2 and the work function of atomically thin 1T'-MoTe2.

The reverse phase transition from the 1T' to the 2H phase was not observed by laser irradiation, even with a higher energy or intensity of the laser. In previous studies (23, 24), temperature has been considered as the primary origin of the phase transition, but the irreversible phase change from the 2H to 1T' phase cannot be explained simply by the thermodynamic reaction in the phase diagram. Moreover, the estimated temperature by laser irradiation (~400°C) is markedly lower than the reported temperature (880°C) for the phase transition. A possible scenario for the one-way phase transition is the strain effect originating from thermal expansion, but the expected thermal coefficient (~10−6 K−1) produces substantially less lattice strain (~0.05%) than the required amount of strain (~5%) for the phase transition (15). Thus, the driving force for the phase patterning is the irreversible Te vacancy created by laser irradiation, as verified in the STEM experiment.

Supplementary Materials

www.sciencemag.org/content/349/6248/625/suppl/DC1

Materials and Methods

Figs. S1 to S15

Tables S1 to S3

References (2539)

References and Notes

  1. Supplementary materials are available on Science Online.
  2. Acknowledgments: We acknowledge support from the Institute for Basic Science (IBS-R011-D1), the National Research Foundation of Korea (NRF) under grant NRF-2014R1A1A2056386 (H. Y.), the NRF under grant 2013R1A1A1008025 (S.W.K.), and the NRF under grant NRF-2005-0093845 (D.H.C. and K.J.C.). All data described are presented in the supplementary materials.
View Abstract

Navigate This Article