Reducing the Schottky barrier between few-layer MoTe₂ and gold

Two-dimensional (2D) atomic crystals have potential applications in electronics [1–5], photonics [6–11] and energy conversion [12–14]. Compared with graphene, transition metal dichalcogenides (TMDs) are more attractive for future optoelectronics applications because of their intrinsic band gaps (1–2 eV) and high photosensitivity [15]. The realization of low-resistance contacts between the channel and metallic electrodes is a key issue for optimizing the performance of TMD-based devices since Schottky barriers at the interface limit carrier injection, resulting in the reduction of carrier mobility, current, and so on.

Detailed studies have demonstrated that metallic electrodes with the Fermi level close to the edge of the conduction band (CB) for n-type or valence band (VB) for p-type semiconducting TMDs, respectively, can mitigate the Schottky barrier and decrease contact resistance [16, 17]. Chemical doping can also modulate the Fermi level of TMDs close to that of metallic electrodes, thus reducing the contact resistance [18]. The contact resistance has been lowered by introducing other 2D materials into the TMD/metal interface, such as graphene [19] or hexagonal boron nitride [20], but the preparation process is more complex. Alternatively, inducing a metallic interlayer between semiconducting TMDs and metal electrodes by phase engineering provides a more direct approach to achieve ohmic contact. So far, it has been demonstrated that phase engineering can be realized by e-beam irradiation [21], laser irradiation [22], plasmonic hot electron injection [23] and organolithium treatment [24, 25], but these methods are complicated or contamination could be introduced in the treatment process.

Molybdenum ditelluride (MoTe₂) can exhibit stable 2H and 1T' phases, corresponding to the semiconducting and metallic phases, respectively. The 2H phase can transform to 1T' phase at temperatures above 880 °C [26, 27], however, the temperature of the phase transition can be lowered to ~650 °C with the help of Te vacancies [27]. Theoretical calculations predict that few-layer MoTe₂ can be transformed from semiconducting 2H phase to metallic 1T' phase by applying 0.2–3% strain even at room temperature [28]. Thus, it is an interesting challenge to optimize the phase transition of this 2D material. With a band gap of ~1.0 eV, 2H-MoTe₂ is sensitive to near-infrared light and can be realized as broadband photodetectors, making it attractive for 2D optoelectronics.

Here we demonstrate a simple method to significantly reduce the Schottky barrier at the MoTe₂/Au interface. After thermal annealing at 400 °C in nitrogen atmosphere, both the field effect mobilities and ON-state currents of the field effect transistors based on MoTe₂ flakes increase by 3–4 orders of magitude. Raman mapping images show that a new Raman peak appears, suggesting that the improved performance is due to MoTe₂ phase transition from the semiconducting phase to metallic phase at low temperatures (200–400 °C). The extracted Schottky barrier height of the treated MoTe₂/Au interface reduces to ~23 meV. We find that the morphology of the Au film plays an important role in the phase transition process. The rougher the Au surface, the higher the phase transformation rate of the 2H-MoTe₂ flakes. With this method, the phase transition is also realized in vertically stacked graphene-MoTe₂-Au devices. The photocurrent of the vertical transport photodetector increases by ~300%, and the response time decreases by ~20%.

Materials

Methods

Characterization

Field effect transistors (FETs) based on few-layer MoTe₂ are fabricated by mechanically exfoliating MoTe₂ flakes on polydimethyl siloxane (PDMS) stamp, and then dry transferring onto gold film electrodes, which are prepared by a standard photo lithography process. Atomic force microscope line scans (figure 1(a)) show that the thickness of the exfoliated MoTe₂ flake is ~5.8 nm, corresponding to 8 layers [33]. After thermal annealing treatment at 200, 300 and 400 °C in nitrogen atmosphere, the electrical transport properties of the few-layer MoTe₂ FET are measured at room temperature in ambient air. As shown in figures 1(c) and S1, the I_ds–V_ds curves change from nonlinear to linear, suggesting that Schottky barrier height at the interfaces between the few-layer MoTe₂ and gold electrodes decreases, and contact resistances consequently decrease. After annealing at 400 °C, the transfer curve of the device displays p-type behaviour (figure 1(d)), which is consistent with previous reports [34], and the drain current in ON state increases to 6  ×  10⁻⁶ A at V_ds  =  1 V, V_g  =  −50 V, which is about 4 orders of magnitude higher than that of the pristine device (figure S1 (stacks.iop.org/TDM/4/045016/mmedia)). This indicates that the resistance of the device decreases dramatically, consistent with the linear behaviour of the I_ds–V_ds curve, and low-resistance contact is formed at the MoTe₂/Au electrode interfaces. Using the equation µ  =  Lg_m/(WC_oxV_ds), where L, W, C_ox, g_m are the channel length, width, gate capacitance, and transconductance at a particular V_ds, respectively, the electric field mobility of the device can be extracted. As shown in figure 1(e), the mobilities increase three orders of magnitude from 0.01 to 15 cm² V⁻¹ s⁻¹ after increasing the annealing temperature from 200 to 400 °C, comparable to few-layer MoTe₂ FETs (~50 cm² V⁻¹ s⁻¹) modified by laser-induced phase patterning [22]. It is reasonable to presume that the high mobility of the device after annealing could be due to low-resistance contact formation at the MoTe₂/Au electrode interface, leading to increased carrier injection efficiency. We fabricated 4 devices, and the mobilities after annealing at 400 °C ranged from 12 to 26 cm² V⁻¹ s⁻¹, indicating that this method is a stable and reproducible method way to improve performance of MoTe₂ transistors. Based on equation (1) and thermionic emission theory [35–37], the effective Schottky barrier height Φ_B of MoTe₂/Au interface at various gate bias (figure 1(g)) can be extracted from the slope of the lines in Arrhenius plot (figure 1(f)).

$$\begin{eqnarray}&&{{I}_{\text{ds}}}=A{{T}^{2}}\exp \left(-\frac{e{{ \Phi }_{\text{B}}}}{{{k}_{\text{B}}}T}\right)\left[\exp \left(-\frac{e{{V}_{\text{ds}}}}{{{k}_{\text{B}}}T}\right)-1\right]\end{eqnarray} \tag{ 1 }$$

where I_ds is the source–drain current, A is the Richardson constant, T is the absolute temperature, k_B is the Boltzmann constant, e is the electron charge, Φ_B is the effective Schottky barrier height, V_ds is the source–drain bias. As discussed in detailed in previous articles, the true Φ_B is extracted when the applied gate bias is equal to the flat band voltage (V_g  =  V_FB) [35]. When V_g is higher than V_FB, the effective barrier height is linear to the applied gate voltage, where the drain current is dominated by thermionic emission at the MoTe₂/Au interface. When V_g is lower than V_FB, the effective barrier height deviates from the linear trend, where the thermally assisted tunneling effect at the MoTe₂/Au interface becomes to affect the source–drain current. Therefore, in the Φ_B versus V_g plot (figure 1(g)), the bias value corresponding to the point deviates from the linear region is V_FB, and the barrier height at this point is the true Φ_B, which is 23 meV in this study. Moreover, the contact resistance at modified MoTe₂/Au interface is estimated to be 6 kΩ µm when V_g  =  −50 V (figure S10), and it is much lower than the reported MoTe₂/Au contact resistance (75 kΩ µm) [38]. Thus, Schottky barrier at MoTe₂/Au interface is reduced, and low-resistance contact is achieved by a simple thermal annealing method.

Zoom In Zoom Out Reset image size

As the FET is fabricated by transferring MoTe₂ flakes onto gold electrodes, the contacts are not so intimate, and the interactions between gold electrodes and MoTe₂ flakes are weak with high Schottky barriers [39]. It is well known that thermal annealing is a good way to improve the quality of contact between metal and TMD materials, mainly because annealing can clean the TMD surface and subsequently induce intimate contact. In the control experiment, the Au electrodes are pre-annealed at 450 °C for 2 h, then transferring the MoTe₂ flake onto the annealed gold electrodes, and finally annealed the device at 400 °C again. As shown in figure 1(h), this control FET exhibits poor output curves after 400 °C annealing, indicating the Schottky barrier height is not reduced in this case, which means the effect of annealing is limited. Furthermore, we prepare bottom-contact MoTe₂ FET with thermal-evaporated gold electrodes on prepared MoTe₂ flakes, which also displays Schottky contact behaviour after 200–400 °C annealing (it will be discussed below in figure 3). From the above results, we can conclude that surface cleaning and intimate contact does not dominate the improvement of MoTe₂ FET performance in this study, and there must be other effects induced at the MoTe₂/Au interface to achieve the ultralow-resistance contact for the top-contact MoTe₂ FETs.

As Au atoms can bond to most chalcogens including Te [40, 41], thermal annealing may facilitate the formation of Au–Te bonds and charge transfer at the MoTe₂/Au interface, resulting in improved contact. This hypothesis can be also ruled out by our control experiments, since the MoTe₂ FETs with preheated or top gold electrodes do not present low-resistance contact after annealing. In previous studies, surface reshape is observed when evaporated metal (Pd, Au, Ag) film is annealed [42, 43], which means their surface atoms reconstruct after annealing. In this study, we propose that the gold surface reshaping process may induce strains in MoTe₂. As previously demonstrated, the semiconducting 2H to metallic 1T' phase transition can be achieved by inducing strain, and especially for MoTe₂, weak strain (~0.3%) is sufficient for phase transition [26]. Therefore, we postulate that the ohmic contact is caused by the phase transition of MoTe₂.

Raman spectroscopy is an effective way to characterize the phase of MoTe₂ [22], and Raman spectra and mapping scans were performed after a series of annealing treatments (figure 2). As shown in figure 2(a), the pristine MoTe₂ flake exhibits sharp peaks corresponding to the $E_{2\text{g}}^{1}$ mode at ~235 cm⁻¹ and A_1g mode at ~174 cm⁻¹, which are the characteristic signatures of the hexagonal (2H) phase of bulk MoTe₂ [27]. The $B_{2\text{g}}^{1}$ peak at 290 cm⁻¹ is also observed, indicating the exfoliated 2H-MoTe₂ crystal is few-layered [34, 44]. The peaks near 139 and 185 cm⁻¹ are attributed to second-order Raman modes of 2H-MoTe₂ [45]. After thermal annealing at 400 °C, the Raman spectrum of few-layer MoTe₂ on SiO₂/Si substrate hardly changes, indicating the few-layer MoTe₂ in the channel of the device remains in the semiconducting 2H phase. In contrast, the Raman spectra of MoTe₂ on the gold electrodes show a new peak at ~123 cm⁻¹, which is reported to correspond to the characteristic A_g mode of metallic phase of MoTe₂ [22]. Figure 2(c) are the A_g mode Raman intensity mapping images for the pristine MoTe₂, acquired after annealing at 200, 300, and 400 °C, respectively. From these images, we can find that, with the increase of the annealing temperature for the few-layer MoTe₂ on the gold electrode, there are more areas where the A_g peak is observed, and the A_g peak intensities increase with the annealing temperature. This suggests that semiconducting 2H-MoTe₂ has been tranformed to the metallic phase in larger regions on the gold electrode, resulting in low-resistance contact at the MoTe₂/Au interface.

Zoom In Zoom Out Reset image size

In previous studies, the metallic phase of MoTe₂ is induced by strain or high power laser irradiation [22, 26]. To exclude the effect of laser-induced phase transition of few-layer MoTe₂ in the Raman experiments, the laser power was set at ~24 µW, which proved to be unable to induce the phase transition (figure S3). As shown in the Raman peak position mapping image (figure S4), a red-shift of 1 cm⁻¹ is observed on annealed MoTe₂/Au, indicating tensile strains are induced in MoTe₂, consistent with a previous Raman study on TMDs flakes, where a red-shift of 1 cm⁻¹ for the E_2g mode was observed when applying in-plane 1% strains [46, 47]. As mentioned above, phase transition of MoTe₂ at room temperature just requires a strain of 0.3% [26, 28], which has been achieved in this study. Moreover, the highest mobility and highest ON-state drain current of the MoTe₂ FET after annealing at 400 °C are consistent with the phase transition of the few-layer MoTe₂ on the gold electrodes. Thus, we suggest that strain induces a phase transition from semiconducting to metallic phase for the few-layer MoTe₂ on gold film, resulting in lower Schottky barrier and lower contact resistance at the MoTe₂/Au interface.

Compared to the phase transition of the MoTe₂ flake on Au electrodes, the Raman A_g mode of the metallic phase is not observed for the MoTe₂ flake on SiO₂/Si substrate (figure 2), which means that the phase transition does not occur for the MoTe₂ flake on SiO₂/Si substrate, suggesting that the Au substrate plays an important role in the phase transition process. To clarify the effect of the Au substrate, the morphology of an Au film (~85 nm thick) without the transferred MoTe₂ flake is measured before and after thermal annealing at 400 °C. As shown in the AFM images of the same area before and after annealing (figure S5(a)), the morphology of the annealed Au film obviously changes, indicating that the thermal annealing can make the Au atoms move and the surface reconstruct. To quantify the change of partial surface roughness, root-mean-square roughness (Rq) values are measured in the two new hillock regions (figure S5(b)), and the corresponding Rq values increase respectively from 1.35 to 1.77 nm and from 1.35 to 2.05 nm after 400 °C annealing. After thermal annealing, we observed the gold surface reshaping in AFM images (figure S5), Raman peak shifts (figure S4), and new phase formation of MoTe₂ (figure 2) in the Raman study. From these results, we propose that the annealing-induced reconstruction of Au surface can induce strains in the MoTe₂ flakes, resulting in the new MoTe₂ phase. In the case of MoTe₂ FET with preheated gold electrodes, reannealing cannot induce gold surface reconstruction again on the preheated gold electrodes (figure S6), so no strain and metallic phase of MoTe₂ can be induced, resulting in the poor contact (figure 1(h)).

We also studied the annealing effect on the bottom-contact MoTe₂ transistor with top gold electrodes (figures 3(a)–(d)). Interestingly, the I_ds–V_ds curves did not become linear with the increase of annealing temperature, and the drain current at the same V_ds and V_g is smaller than that of top-contact MoTe₂ FETs, indicating that the Schottky barrier height is still high and thermal annealing cannot improve the contact in this case. To analyze the phase of the MoTe₂ in this case, the annealed MoTe₂/top-Au is prepared on a transparent quartz substrate so that Raman laser can irradiate the gold-covered MoTe₂ flake from the bottom (as shown in figure 3(e)). From the Raman mapping images (figure 3(f)), there is no signature peak of metallic phase MoTe₂, indicating top-layer gold surface reconstruction cannot induce phase transition of bottom-layer MoTe₂. Therefore, without phase engineering, the I_ds–V_ds curves of the bottom-contact MoTe₂ transistor exhibit nonlinear behavior, the ON state current are smaller than that of the top-contact MoTe₂ FETs.

Zoom In Zoom Out Reset image size

To further confirm the morphology effect on phase transition, Au electrodes with different surface roughness was prepared by controlling the deposition conditions of the thermal evaporator [42, 48]. As shown in figures 4(a)–(c), three categories of surface roughness were prepared, and the measured Rq values are 2.08, 1.08 and 0.43 nm, respectively, corresponding to the rough, medium and smooth Au electrodes. Few-layer MoTe₂ flakes were transferred onto these electrodes, and Raman mapping performed after annealing treatment at 400 °C (figures 4(d)–(f)). As shown in figures 4(d)–(f), the Raman intensity of the A_g mode is the strongest for the MoTe₂ flakes on the rough Au electrode, while it is the weakest for the MoTe₂ flakes on the smooth Au electrode, indicating the roughness of the Au electrode is directly correlated to the phase transition probability of the MoTe₂ flakes. As it has been shown that the phase transition is induced by strains, we presumed that strains can be adjusted by annealing gold with different surface roughness and consequently induce varying degrees of MoTe₂ phase transition. The electrical transport properties are measured for the three categories of devices (figures 4(g)–(i)). It is found that the I_ds–V_ds curves are the most nonlinear for the device with the smooth Au electrode, suggesting that the Schottky barrier at the contact interface is the highest when the phase transformation rate is too low. In contrast, the I_ds–V_ds curves become linear for the device with the rough Au electrode, and the low-resistance contact is achieved with metallic phase formation. In summary, the roughness of the Au electrode can influence the phase transition of the MoTe₂ flakes, and the phase transformation rate of the 2H-MoTe₂ flakes is directly correlated to the surface roughness of the Au electrodes.

Zoom In Zoom Out Reset image size

Low-resistance contact can also be achieved in vertically stacked graphene/MoTe₂/Au devices by our method. As shown in figure 5(a), a few-layer MoTe₂ flake (red dot line) is transferred onto an as-prepared Au film (blue dot line), and then a CVD-grown monolayer graphene (black dot line) is transferred onto the MoTe₂ flake. To ensure vertical charge transport, the top graphene layer is smaller than the MoTe₂ flake, and located directly above the MoTe₂ flake (figure 5(b)). In the vertical transport device, the Au film and top graphene layer are the bottom and top electrodes, respectively. From the Raman mapping images (figures 5(c) and (d)), the characteristic Raman A_g peak of metallic MoTe₂ is observed, suggesting phase engineering can be achieved on MoTe₂/Au even if other 2D materials (e.g. Graphene) are used as the top (transparent) electrode. Compared with other strategies such as laser or E-beam irradiation, our work shows great potential in the phase engineering of vertical MoTe₂ devices. Laser and E-beam irradiation may cause unintended beam damage or modifications to the device structure [22]. For our graphene/MoTe₂/Au vertical device, the photocurrent (I_ph  =  I_light  −  I_dark) mapping images are taken at V_ds  =  0.1 V and illuminated by a continuous wave (CW) 532 nm laser at ~24 µW, and the photoresponsivity (R  =  I_ph/P_light) is calculated to be 46 mA W⁻¹. As shown in figure 5(e), the photocurrent is higher in the phase transition area, indicating photo-generated electron-hole pairs are more efficiently separated and collected by the electrodes [6, 49]. Compared with the photocurrent of the device before thermal annealing (figure 5(f)), after the phase transition of the MoTe₂ flake, the photocurrent increases about ~300%. Figure 5(g) shows the time-resolved photoresponse of the device, and the rise and decay times are respectively defined by the increase in photocurrent from 20% to 80%, and the corresponding decrease from 80% to 20% of the maximum photocurrent [50]. Clearly, the rise time decreases from 131 to 96 ms, while the decay time decreases from 114 to 95 ms after annealing. Thus, an MoTe₂/Au ohmic contact can also be achieved in a vertically stacked photosensor, resulting in higher photosensitivity and faster photoresponse.

Zoom In Zoom Out Reset image size

In summary, the Schottky barrier between Au and MoTe₂ flakes is significantly reduced by simple thermal annealing at 400 °C in a nitrogen atmosphere. The characteristic Raman A_g peak of metallic phase MoTe₂ is observed at the interface between MoTe₂ flakes and Au electrodes, indicating that the reduction of Schottky barrier results from the phase transition observed in the 2H-MoTe₂ flakes. We find that the transformation rate to the metallic phase is directly correlated to the roughness of the Au electrode substrate. In this work, we have demonstrated a simple phase engineering process for ohmic contacts at the MoTe₂/Au interface, and as no chemicals are introduced, this method could have important applications in future optoelectronics based on 2D materials.

Supporting Information is available from the IOP Online Library or from the author.

This work was supported by the National Science Foundation of China (Grant No. 51472164), the 1000 Talents Program for Young Scientists of China, Shenzhen Peacock Plan (Grant No. KQTD2016053112042971), the Educational Commission of Guangdong Province project (Grant No. 2015KGJHZ006), the Science and Technology Project of Guangdong Province (Grant No. 2016B050501005), the Educational Commission of Guangdong Province (Grant No. 2016KCXTD006), the Natural Science Foundation of SZU (Grant No. 000050 and No. 2017029), and the China Postdoctoral Science Foundation (2015M582415). ATSW acknowledges financial support from ASTAR Pharos Project No. 144-000-359-305. We also acknowledge Junyong Wang (PhD student, National University of Singapore) for his help in electrical characterization.

The authors declare no competing of financial interest.