Introduction

Molybdenum disulphide (MoS2) is a layered two-dimensional (2D) material, which draws intensive attention because of its unique electrical, optical and mechanical properties. It has been considered as a promising candidate for various future nanoelectronic and spintronic applications. It is a semiconductor with an indirect bandgap (1.2 eV) for its bulk material and a direct bandgap (1.8 eV) for monolayer MoS21,2,3,4,5,6. The presence of the direct bandgap in monolayer MoS2 allows room-temperature FETs with an on/off current ratio exceeding 1087. Studies also point out that MoS2 could be used in sensors8 or for photoluminescence9 applications. The spin transport properties were theoretically predicted with a spin diffusion length of 400 nm at room temperature and it becomes longer than 800 nm at low temperature (T < 77 K)10. Because of the semiconducting nature, the contacts between MoS2 and metal electrodes naturally forms Schottky barriers, which plays a crucial role in spin transport. To cope with this problem, theorists proposed that titanium is a suitable electrode material because of favorable geometry and large electronic density of state at the interface11. And experimentalists found that devices with scandium electrodes have very good performance because of low metal work function12. However, until now there is not much research on the ferromagnetic contact with MoS213,14, which may lead to the reduced conductance mismatch and the enhanced spin polarization.

In this letter, we study the contact between MoS2 and ferromagnetic permalloy (Py) electrodes. The Schottky barrier height (SBH) is extracted by 2D thermionic emission theory13,15. For multilayer MoS2, we observed a barrier height of −5.7 mV. This negative Schottky barrier leads to a perfect Ohmic contact between MoS2 and Py electrodes which dramatically enhances the field effect transistor (FET) performance of MoS2: the mobility of multilayer MoS2 with a single back gate reaches 55 cm2 V−1s−1 at low temperature. For monolayer MoS2, the SBH has a positive value, which can be readily tuned by both back gate voltage and by the insertion of an Al2O3 tunneling layer. When inserting 2.5 nm Al2O3, the SBH is reduced from 80.2 to 2.7 mV; by applying a back gate voltage of 30 V on the samples without Al2O3, the SBH is reduced to −8.3 mV. These two approaches of tuning the barrier height are important for realizing spin injection into monolayer MoS2.

Results

Structure of the FET devices and method for extracting SBH

A schematic diagram of our devices is shown in Figure 1a. Electrode on Si substrate is used as a back gate to tune Fermi level and carrier density. Figure 1b shows a scanning electron microscopy (SEM) picture of the device. The channel length of the device is 2 μm and the width of exfoliated MoS2 is about 4 μm. The channel material is verified to be tri-layer MoS2 by a Raman spectrum16 as shown in Figure 1c. The devices were characterized by two-probe I-V measurement for Schottky barrier height and by ID-VG measurement for mobility.

Figure 1
figure 1

(a) A schematic diagram of the device. Py electrodes are covered by Au to prevent oxidation. (b) A SEM image of the multilayer MoS2 device. All the data are measured from electrode 1 and 2. (c) A Raman spectrum of the multilayer MoS2. It is estimated to be a tri-layer MoS2 through the distance of ~22.5 cm−1 between two vibrating modes (in-plane mode E12g and out-of-plane mode A1g)16. (d) A band diagram of Py and MoS2. EF of Py is close to EC of the tri-layer MoS2.

Full size image

The forward I-V characteristics of an ideal Schottky diode can be described as follows17:

where

Is is the diode saturation current, A is the Richardson constant, S is the contact area of junction, q is the electron charge, ΦB is the Schottky barrier height and kB is the Boltzmann constant. As the device is thin enough to be treated as a 2D material, the drain-source current IDS can be defined by 2D thermionic emission equation18, which employs the reduced power law T3/2 for a 2D transport channel:

where A*2D is the 2D equivalent Richardson constant, n is the ideality factor and VDS is the drain-source bias voltage. To determine the Schottky barrier height ΦB, temperature dependent I-V measurements were carried out. Figure 2a shows the I-V curves at several temperatures on a logarithmic scale. To investigate the barrier, it is common to use Arrhenius plot, i.e., ln(IDS/T3/2) against 1000/T for various VDS in Figure 2b. By fitting the data to each VDS, we obtained the slopes with . Then by plotting the slopes as a function of VDS, the SBH could be extracted from the y-intercept (Figure 2c).

Figure 2
figure 2

(a) I-V curves of the tri-layer MoS2 FET device with Py directly contacting to MoS2 from T = 100 to 200 K. (b) Arrhenius plot ln(IDS/T3/2) vs 1000/T at different drain-source voltages (VDS). (c) Extraction of ΦB via the y-intercept value. Each data point here represents the slope obtained from the Arrhenius plot in (b) under a specific VDS. (d) A band diagram showing a negative Schottky barrier.

Full size image

Schottky barrier height between Py and tri-layer MoS2

For our tri-layer MoS2 transistor device, ΦB is found to be −5.7 mV in the temperature regime of 100 ~ 200 K. Such a negative Schottky barrier produces a good Ohmic contact with a perfect linear I-V curve between MoS2 and Py electrodes (Figure S1). Similar results were also reported in p-type MoS2 transistors with MoOx electrodes19. According to the Schottky-Mott model20, we could roughly estimate the SBH based on the work function of the metal Φmetal relative to the electron affinity (or vacuum ionization energy) of the semiconductor χsemi:

The negative Schottky barrier suggests that the work function of Py is slightly smaller than the affinity of MoS2, as shown in Figure 1d.

To investigate the FET performance, we have performed temperature-dependent ID-VG measurements. Figure 3a shows typical gate-dependent conductance curves at different temperatures and Figure 3b displays the temperature-dependent conductance under different gate voltage. It is noted that when applying a small back gate voltage (VG < 35 V), the channel tri-layer MoS2 shows an insulating behavior that the conductance decreases as temperature decreases. While VG > 35 V the conductance increases as temperature decreases. This is a hallmark of metallic behavior which suggests that the tri-layer MoS2 has entered a metallic state.

Figure 3
figure 3

(a) Conductance G as a function of VG at different temperatures for the tri-layer MoS2 device. (b) Temperature dependence of G at different VG. When VG < 35 V the G-T curve shows an insulating behavior while VG > 35 V it becomes metallic. (c) Mobility as function of temperature under different VDS. μFET is independent of T when T < 10 K, while above ~30 K μFET decreases following a T−γ dependence with γ = 0.56 ~ 0.66.

Full size image

The field effect mobility can be calculated from the linear regime (40 ~ 50 V) of conductance curves using the following expression21:

where dG/dVG is the slope of the conductance curve in the linear regime, L = 2 μm is the channel length, W = 4 μm is channel width and Cox = 1.3 × 10−4 F m−2 is the capacitance between the channel and the back gate per unit area (Cox = ε0εr/d; εr = 3.9; d = 270 nm). Figure 3c shows the extracted mobility as a function of temperature under different drain-source voltage. The mobility is nearly independent of temperature when T < 20 K, indicating that the scattering of charged impurities is reduced by drain-source voltage22,23. At higher temperatures (T > 100 K), the mobility of the tri-layer MoS2 is mainly influenced by the phonon scattering23,24. Fitting to the expression μ ~ T−r, the range of exponent part γ can be obtained between 0.47 and 0.66. This value is much smaller than the theoretical prediction, i.e., γ = 1.69 for single-layer MoS225 or γ = 2.6 for bulk crystals26, indicative of a weak electron-phonon interaction20. The mobility of our tri-layer MoS2 exceeds 55 cm2V−1s−1, which is a comparatively high value for single back gate MoS2 transistor devices. As the Schottky barrier can significantly impact the electron mobility12, the high mobility of our device also provides a strong evidence of a low SBH between Py and tri-layer MoS2.

Schottky barrier height between Py and monolayer MoS2

Different from the tri-layer, the SBH between Py and the monolayer MoS2 turns out to be 80.2 mV (Figure 4d). This can be understood by the fact that the monolayer MoS2 has a large bandgap of 1.8 eV and consequently it has a smaller electron affinity4. Thus, the increase of Schottky barrier height in the monolayer system is consistent with the Schottky-Mott model (Equation 4), i.e., for the same metal work function, when the vacuum affinity decreases, the Schottky barrier height increases.

Figure 4
figure 4

(a–c) Arrhenius plot of monolayer MoS2 with Py electrodes in a large temperature range. A back gate voltage of VG = 0 V, 10 and 20 V was applied to the device as shown in (a), (b) and (c), respectively. (d) Schottky barrier height ΦB as a function of VG. The insets show the band diagram of Schottky barrier between Py and monolayer MoS2.

Full size image

However, great caution must be exercised when the Arrhenius plot is used to extract SBH. The data in Figure 4Figure 4a are not completely linear for the entire temperature range. In the high temperature regime (T > 130 K), the data show negative slopes, corresponding to the positive SBH. At low temperatures, however, ln(IDS/T3/2) versus 1000/T has positive correlations, suggesting negative SBH. These fitted results are contradictory to the observation of “S” shape I-V curves at low temperatures (Figure S2), which are strong evidence of positive Schottky barriers. To explain this contradiction, one needs to take the semiconducting nature of MoS2 into consideration. As the temperature decreases, the resistivity of monolayer MoS2 increases and the device reaches an “off” state. It is well known that at the off state the channel resistance is too large that the current does not change much with temperature20. Thus, the thermionic emission equation is no longer suitable to describe the current. With this limitation, the SBH could only be extracted in the high temperature regime (above the turning points) using 2D thermionic emission equation, while the “V” shaped turning point in the Arrhenius plot represents the entrance of device “off” state.

To investigate the tunability of the Schottky barrier, we have performed the gate-dependent I-V measurements (Figure 4a–c). When VG = 0 V, the turning point is about 130 K. As the voltage is increased to 10 and 20 V, the turning point shifts to 20 and 8 K, respectively. Further increasing the gate voltage (up to 30 V) makes the turning point completely vanished (Figure S3). For traditional field effect transistors, the channel conductance can be tuned by gate voltage. When VG > 0 V, the conduction channel is broadened such that it needs a much lower temperature to turn the device off, which explains the systematic shift of the turning point. Figure 4d summarizes the SBH as a function of VG: the SBH is reduced to 20.4, 1.1 and −8.3 mV when VG changes from 10, 20 to 30 V, respectively. This is attributed to the upwards shift of Py Fermi level by positive gate voltage (Figure 4d inset)27.

For the spin injection into monolayer MoS2, the contact between ferromagnetic metal and semiconductor is crucial28. The spin polarization of injected carriers through Ohmic contact is extremely small due to the conductance mismatch13. In order to alleviate this issue, the resistance-area (RA) product should be designed properly to obtain a significant spin polarization and magnetoresistance29. In conventional 3D semiconductors, the Schottky barrier can provide the required resistance30, but the surface doping is needed to facilitate single step tunneling31. Another approach to obtain the proper RA product is to achieve a pinning-free FM/Oxide/SC interface by inserting a tunneling oxide layer, which could lower the SBH and the resistance could be tuned by the thickness of the oxide layer. Here we choose Al2O3 as the tunneling oxide layer, which is well known for high spin injection efficiency and large tunneling magnetoresistance in magnetic tunnel junctions32,33,34,35,36,37.

Similar to the back gate voltage, the insertion of Al2O3 layer also causes the “V” shaped turning point in the Arrhenius plot to shift towards lower temperatures (Figure 5a–c). When the thickness of Al is 0.8, 1.7 and 2.5 nm, the turning point shifts to 40, 25 and 10 K respectively. This can be explained by the reduction of conductance mismatch38 (or reduced SBH) between the electrodes and monolayer MoS2 as Al2O3 thickness increases. For that reason, the shift of the turning point could be regarded as an evidence of the alleviation of conductance mismatch with the inserted Al2O3 layer. The corresponding SBH is extracted to be 32.1, 15.9 and 2.7 mV (Figure 5d), respectively. Such a dramatic decrease shows an effective control of the SBH via changing the thickness of oxide layer. By the combination of applying a gate voltage and inserting an Al2O3 layer, we could minimize the conductance mismatch and tune the Schottky barrier height down to zero, which may help to achieve the proper RA product.

Figure 5
figure 5

(a–c) Arrhenius plot of monolayer MoS2 with different thickness of Al2O3 inserted. Al2O3 thickness is 0.8, 1.7 and 2.5 nm in (a), (b) and (c), respectively. (d) Schottky barrier height ΦB as a function of the Al2O3 thickness. The inset shows the band diagram of Schottky barrier when inserting the Al2O3 tunneling layer.

Full size image

Discussion

In conclusion, we have investigated the properties of MoS2 FET with ferromagnetic Py electrodes; the Schottky barrier height is extracted using 2D thermionic emission analysis of I-V curves. For the tri-layer MoS2, there is a negative Schottky barrier between Py and MoS2 and this Ohmic contact yields a high mobility due to low contact resistance. For the monolayer MoS2, there is a positive Schottky barrier, which is dramatically reduced either by applying a gate voltage or inserting a tunneling Al2O3 layer. To some extent, the insertion of Al2O3 layer also alleviates the conductance mismatch. Such control of Schottky barrier paves the way of proper design of the RA product, which sheds light on the future research of spin transport and spin injections in MoS2.

Experimental Section

Multilayer MoS2 is obtained through mechanical exfoliation from bulk MoS2 crystals onto pre-patterned SiO2/Si substrate (the thickness of SiO2 is 270 nm). FET devices were fabricated by e-beam lithography (EBL) using PMMA/MMA bilayer polymer. Subsequently, Py electrodes are deposited by magnetron sputtering, followed by a deposition of gold layer to protect Py from oxidation. All the data are measured from electrode 1 and 2 in Figure 1b. For the monolayer transistors, MoS2 is obtained via chemical vapor deposition (CVD) using high purity molybdenum and sulphur as the source materials, similar to the previous report39. After growth, they were transferred onto clean SiO2/Si substrate for the following EBL process using PMMA stamping method40. The tunneling Al2O3 layer was produced as follows: first a thin layer of Al was deposited by e-beam evaporation (the SEM picture in Figure S4 shows a high-quality Al layer without visible pinholes). Then the samples were placed in the air overnight for natural oxidation to develop Al2O3. The thickness of Al2O3 can be estimated by 41. Before measurement, the devices were annealed at 360 K for two hours in vacuum to remove polymer residues between the interface of Py and MoS242,43.