High-performance monolayer MoS₂ field-effect transistor with large-scale nitrogen-doped graphene electrodes for Ohmic contact

Two-dimensional (2D) transition metal dichalcogenides (TMDCs) have unique electrical and optical properties. Of them, monolayer molybdenum disulfide (MoS₂) is theoretically expected to be used as material for a replacement channel for postsilicon electronics, such as a field-effect transistors (FETs), owing to its direct bandgap (1.9 eV), high electron mobility, high transconductance, and excellent on/off ratio (>10⁸).^1–3 In addition, the conduction and valence band edges of monolayer MoS₂ can indicate valley degrees of freedom for next-generation optoelectronics.^4,5 Because of the large valley separation in the momentum space, an inversion symmetry breaking with strong spin–orbit coupling leads to spin-valley physics that enables the manipulation of the spin and valley in monolayer MoS₂. Despite these potential advantages, there is a major bottleneck in developing high-performance MoS₂-based devices and studying their exotic spin-valley quantum physics because the charge transport in MoS₂ devices is largely dominated by poor electrical contact. Due to Fermi level pinning at the metal–MoS₂ interface, most metals form the Schottky contact that interferes with efficient charge carrier injection and extraction, and this often limits the study of the intrinsic transport properties of MoS₂.

Past attempts to solve the problem of poor electrical contact have featured the selection of the most suitable metal for MoS₂ to overcome Schottky contact,^6–8 but have achieved limited success. In the relevant approaches, achieving the lowest contact resistance using low-work function scandium contact⁹ has been shown to lower the Schottky barrier height (SBH) to ∼30 meV. The large contact resistance is the result of Fermi level pinning in MoS₂ near the conduction band due to the charge neutral level position¹⁰ as well as sulfur vacancy or the defect level.^11,12 In addition, experimental and computational studies suggest that absorbed contamination and damage due to kinetic energy transfer during metal deposition can result in Fermi level pinning and an increase in contact resistance.^13,14 Other approaches to contact engineering, including the use of edge contact,^15,16 phase engineering,^17,18 thermal annealing,¹⁹ and selective etching,^20,21 have limitations for approaching to Ohmic contact. To prevent Fermi level pinning, field-effect transistors based on graphene–MoS₂ van der Waals heterojunctions exhibit high mobility and on/off ratio.^1,22 The Fermi level of graphene can be easily changed by using the gate voltage to ensure a match with the conduction band of MoS₂.^1,22,23 The use of graphene electrode platforms is the best-known strategy at present for developing high-performance monolayer MoS₂ FETs and observations of quantum transports such as the quantum Hall effect and the Shubnikov-de Haas oscillation.¹ However, reliable Ohmic contact with monolayer MoS₂ FETs at low temperature has been demonstrated only at very high carrier density (n ∼ 1 × 10¹³ cm⁻²) due to the finite difference in the work function between graphene and monolayer MoS₂. In a previous study,²⁴ we used monolayer hexagonal boron nitride (hBN)/Co contact to achieve Ohmic contact under a modest carrier density regime (n < 10¹³ cm⁻²), but this strategy has fundamental limitations in terms of device stability and large-scale fabrication.

In this study, we propose a large-scale highly electron-doped graphene contact platform of monolayer MoS₂ for reliable Ohmic contacts and Fermi level alignment. We demonstrate monolayer MoS₂ FETs using nitrogen-doped graphene (NGr) with barrier-free Ohmic contact, instead of the pristine graphene contact platforms that were previously used. The highly electron-doped graphene leads to low contact resistance, negative threshold voltage, and high performance of the MoS₂ device. Further, we performed electrical transport, Raman spectroscopy and X-ray photoelectron spectroscopy (XPS) before and after thermal annealing to determine the reason for the dominant effects of electron doping in NGr.

Mechanical exfoliated monolayer MoS₂ [atomic structure shown in Fig. 1(a)] was characterized by Raman and photoluminescence (PL) spectroscopies. Figure 1(b) shows the Raman spectrum of monolayer MoS₂, which exhibited two major peaks corresponding to an in-plane (E_2g) mode at 384.72 cm⁻¹ and an out-of-plane (A_1g) mode at 403.62 cm⁻¹ at an excitation laser wavelength of 532 nm. The Raman peak difference (∼19 cm⁻¹) can be used to reliably identify monolayer MoS₂. Furthermore, we observed PL peaks at 1.86 eV (666 nm) and 1.99 eV (623 nm) corresponding to the A1 and B1 excitons of monolayer MoS₂ as shown in Fig. 1(c).

We demonstrated the controlled growth of highly electron-doped NGr on Cu foil by CVD method at 1000 °C with pyridine (C₅H₅N) liquid precursor, which is a nitrogen-contained organic molecule, instead of the ammonia (NH₃) gas precursor. Because a high concentration of ammonia gas precursor leads to the oxidation of Cu foil during CVD graphene growth, and results in an insufficient nitrogen doping and high defect density in NGr.²⁵ On the contrary, the liquid pyridine (C₅H₅N) precursor as the source of both carbon and nitrogen enables an efficient nitrogen doping and stable growth of monolayer NGr over a large area with reduced defect density.^26,27 The detailed method of NGr synthesis is described in supplementary material 1.

Figure 1(d) shows schematics of three types of possible substitutional nitrogen dopant in NGr: graphitic, pyridinic, and pyrrolic. The incorporation of substitutional nitrogen dopant into the carbon lattice is a direct way to control the electronic structure of graphene.^28,29 The graphitic−nitrogen in NGr induces n-type conductivity because an electron participates in the $π$ bond and the fifth electron forms a partial $π*$ bond in the conduction band,³⁰ whereas the pyridinic- and pyrrolic-nitrogen in graphene form p-type dopants.³¹ 

To characterize the effect of the doping of nitrogen in graphene, we performed Raman spectroscopy on NGr and pristine graphene. In Fig. 1(e), the pristine graphene (red line) shows two intense Raman peaks corresponding to G (∼1588.52 cm⁻¹) and 2D (∼2684.75 cm⁻¹) at an intensity ratio of I_2D/I_G = 2.55.³² In the NGr (black line), the slightly blue-shifted G peak (∼1596.68 cm⁻¹) of NGr was observed compared with the G peak (∼1588.52 cm⁻¹) of pristine graphene. There are several reasons for the Raman peak shift, including the effects of both doping and strain.³³ Upon electron doping by nitrogen, the carrier concentration based on the G peak shift was approximately 3.5 ± 0.2 × 10¹² cm⁻². Moreover, a strong D peak at ∼1351.37 cm⁻¹, a D′ peak at ∼1630.88 cm⁻¹, and a combination mode ∼2955.95 cm⁻¹ (D + D′ peak) appeared in NGr. They were activated by such defects as substitution heteroatom, vacancies, and grain boundaries typically observed in NGr.^33–35 Figure 1(f) clearly distinguishes between G and D′ peaks of highly electron-doped NGr compared with those of pristine graphene. The shifts in the G-peak and D′-peak were due to charge distribution, which is indirect evidence of the doping effect of nitrogen (for additional information, see the supplementary material S2).

To further investigate the effect of the doping of NGr, we measured electrical transport under a constant source–drain voltage of 10 mV and observed a charge neutrality point (CNP) at −49.2 V as shown in Fig. 1(g), which corresponded to a carrier density of 3.72 × 10¹² cm⁻². This carrier density is consistent with the G peak shift of NGr. The NGr FETs on SiO₂ had an electron mobility (μ_e) of ∼420 cm²/V s at V_g = +80 V and hole mobility (μ_h) of ∼550 cm²/V s at V_g = −80 V at room temperature (supplementary material S3). We confirmed the level of nitrogen doping and high mobility of NGr using previous studies.^26,36 This result shows that the properties of NGr were controlled successfully via CVD method.

Figure 2(a) shows the schematic of the fabrication process of MoS₂ FETs with large-scale CVD-grown NGr electrodes. The details are presented in supplementary material S1. The optical microscopic image of the MoS₂ FETs device with NGr is shown in Fig. 2(b), where the NGr electrodes (dashed black line) and monolayer MoS₂ (red dashed line) are marked. Figure 2(c) shows the schematic of the final MoS₂ FETs device with NGr electrodes encapsulating only the top hexagonal boron nitride (hBN). Figure 2(d) shows the cross-sectional view of the structure of the monolayer MoS₂ FETs with NGr electrodes.

We investigated the electrical transport properties of monolayer MoS₂ FETs with pristine graphene electrodes as a reference and compared the results with those of the highly doped NGr electrode. Figure 3(a) shows the linear and semilog plots of the two-probe transfer characteristics of MoS₂ FETs obtained at a drain voltage (V_ds) of 100–400 mV with a 100 mV step. The transfer characteristics of monolayer MoS₂ FETs with pristine graphene electrodes exhibited typical n-type behaviors with a threshold voltage (V_th) of +39 V, where V_th was extracted as the gate voltage axis intercept of the linear extrapolation in the linear region in the range of V_g from +60 V to +79 V. The field-effect mobility of MoS₂ with pristine graphene electrodes was μ = ∼2.21 cm²/V s, which was obtained from $μ=LwCoxVdsdIdsdVg$⁠, where $dIdsdVg$ is transconductance, L (8 μm) and W (9 μm) are the channel length and width, respectively, $Cox$ is the back-gate capacitance per unit area (1.21 × 10⁻⁸ F/cm² for 285-nm-thick SiO₂), I_ds is the drain current, and V_g is the gate voltage. Figure 3(b) shows the two-probe output curves of the monolayer MoS₂ FETs with pristine graphene electrodes depending on the gate voltage from −80 to +80 V, and they indicate the characteristics of the linear output and modest I_ds at a high V_g. A relatively small I_ds for the pristine graphene electrodes for monolayer MoS₂ is due to the high contact resistance and a finite SBH between pristine graphene (4.5 eV) and MoS₂ (4.15 eV),³⁷ as shown in Fig. 3(c).

We also performed the transfer- and output-curve characteristics of the monolayer MoS₂ FETs with highly doped NGr electrodes. Figure 3(d) shows the linear and semilog plots of the transfer characteristics of MoS₂ FETs with NGr electrodes obtained at range of drain voltage (V_ds) of 100–400 mV in steps of 100 mV. The V_th of NGr electrodes to the monolayer MoS₂ FETs was −54.2 V while that of the pristine graphene electrode platform was +39 V. The negative shift in V_th of the MoS₂ FETs indicates a higher level of electron doping of NGr for carrier transportation at the NGr/MoS₂ interface than at the pristine graphene/MoS₂ interface as well as a lowering of SBH.

In the MoS₂ FETs with NGr electrodes, the on-current improved to 214% compared with that of pristine graphene contact at the same source/drain bias at V_g = +80 V, where the ratio of the on/off current was approximately 10⁶, similar to previous results.^1,38,39 As shown in Fig. 3(d) (L—9.3 μm and W—11.2 μm), the field-effect mobility of the device in contact with NGr was 8.76 cm²/V s, which was four times that of the device in contact with pristine graphene (2.21 cm²/V s). Figure 3(e) shows the output curve of MoS₂ FETs with varying V_g from −80 V to 80 V. The output curve was linear when both pristine graphene and NGr contact were used for MoS₂ at room temperature. Even in the presence of barrier height, the thermal energy at room temperature (26 meV) provided enough energy for the charge to pass over. However, in the output curve at 77 K, as shown in supplementary material S5, MoS₂ with NGr showed the linear Ohmic contact behavior in the gate range from −80 V to +80 V, and in case of pristine graphene contact showed nonlinear Schottky contact behavior even at large gate voltage (V_g = +100 V). We can estimate the Fermi level of NGr based on the carrier density (n = 3.72 × 10¹² cm⁻², NGr at V_CNP = −49.2 V) under zero gate voltage, $EF=ℏvFπn$⁠, where, $vF$ and n are Fermi velocity and carrier density, respectively. Consequently, the Fermi energy level of NGr can be inferred to be 239 meV higher than that of pristine graphene (4.5 eV).³⁷ Figure 3(f) shows the corresponding band alignment between NGr and MoS₂.

The fourfold improvement in the mobility of monolayer MoS₂ FETs with NGr electrodes can be attributed to the barrier-free Ohmic contact for efficient carrier injection and extraction, as shown in Fig. 3(f). Regarding the threshold voltage and SBH, we previously reported a correlation between V_th and SBH by adjusting the Fermi level of graphene contact to monolayer MoS₂ using dual top and bottom gates.⁴⁰ When the Femi level of graphene approached the conduction band of monolayer MoS₂, we observed that a reduction in SBH led to a higher conductance as well as a negative shift in V_th. Based on previous reports, the large negative shift of NGr electrodes contact platform is due to the high Fermi level of NGr compared with the conduction band of monolayer MoS₂, which led to Ohmic contact at room temperature and low temperature. The value of V_th of the monolayer MoS₂ FETs with NGr electrodes was −59.2 V while that of the pristine graphene electrode platform was +39 V, which was in good agreement with barrier-free Ohmic contact of NGr contact at low temperature. We observed a negative shift in V_th and Ohmic contact behaviors from several NGr contact platforms MoS₂ (supplementary material S4).

To determine whether the electrical properties of the MoS₂ device were influenced by substitutional nitrogen dopant in NGr, we investigated the electrical transport properties of monolayer MoS₂ with NGr contact before and after thermal annealing at 360 °C for 30 min in a vacuum (∼10⁻⁶Torr). The thermal annealing process (360 °C) in a vacuum is the typical process for removing the polymer residue of van der Waals heterostructure fabrication, and do not damage MoS₂ with hBN encapsulation.^40,41 Due to vacancy defects in the internal structure of graphene, the substitutional nitrogen dopant in NGr was thermodynamically unstable. The annealing process was carried out to desorb only the substitutional nitrogen dopant in the NGr.

To prove the possibility of degradation of MoS₂ (oxidation or sulfur vacancy), the Raman and PL spectroscopes of the monolayer MoS₂ were obtained before and after annealing (supplementary material S6). The shift and intensity of the Raman peak of MoS₂ were almost negligibly small. The PL of MoS₂ showed a red shift of 0.02 eV at the peak after annealing due to the effect of strain. The full-width at half maximum (FWHM) of sharp PL peaks of 93 meV (before annealing) and 86 meV (after annealing) of A-exciton also indicate that the hBN-encapsulated MoS₂ retained its crystal structure without degradation (Fig. S6).

For accurate analysis of the composition of NGr, we analyzed its binding state and chemical composition before and after annealing by X-ray photoelectron spectroscopy (XPS). Figure 4(a) shows the C 1s XPS spectra of NGr before and after annealing. Before annealing, we observed the strongest peak at ∼285 eV (C-C) and the weak peak at ∼288 eV (C-N),⁴² which originated from the substitution of nitrogen. Following annealing, the only sharp peak at around 285 eV corresponds to the sp² carbon with C-C bonds. The N 1s XPS spectrum [Fig. 4(b)] featured the formation of pyridinic (397.3 eV), pyrrolic (400.2 eV), and graphitic (401.7 eV) N structures before annealing. This shows that graphitic N atoms were dominant, which can lead to a strong n-type doping effect. However, after annealing, N-C bonding in the nitrogen binding states (N 1s) was not observed. This is clear evidence that the substituted nitrogen was desorbed from NGr by thermal annealing.

To determine the effect of annealing on the reduction in the substitutional nitrogen dopant in NGr, we measured the electrical transport of NGr [Fig. 4(c)] and monolayer MoS₂ FETs with NGr electrodes [Fig. 4(d)]. We observed significant changes in the CNP of NGr (−49.2 V to −9 V) and the value of V_th of the monolayer MoS₂ FETs (−54.2 V to −18.8 V) before and after thermal annealing in vacuum. This shows that electron doping due to nitrogen desorption in NGr changes the V_th of MoS₂ with NGr after annealing. The field-effect mobility of the device after thermal annealing decreased from 8.76 cm²/V s to 1.02 cm²/V s, and the value of V_th of the monolayer MoS₂ FETs with NGr electrodes after annealing shifted by 35.4 V in the positive direction. The mobility of the MoS₂ FETs appeared to have been influenced by changes in the electrical properties of NGr, rather than from the degradation of MoS₂ after thermal annealing.

In this letter, we reported high-performance monolayer MoS₂ FETs, where large-scale CVD-grown NGr graphene with an intrinsic high electron carrier density was used to establish Ohmic contact with monolayer MoS₂ at a modest gate voltage. The fabricated MoS₂ FETs exhibited a large negative shift in threshold voltage and barrier-free Ohmic contact at zero gate voltage owing to the Fermi level alignment of the monolayer MoS₂ and NGr. Moreover, the use of NGr in case of contact led to remarkable enhancements in the on-current and electron mobility of the monolayer MoS₂ FETs compared with when the pristine graphene electrode platform was used. Further, the atomic C-N bonding in NGr was responsible for the dominant effects of electron doping. We proposed a strategy that allows for the large-scale production of NGr electrodes as a promising device platform for the development of high-performance electronic devices, and for the examination of unique spin-valley physics for future electronics based on 2D materials.

See the supplementary material for fabrication, the mobility of NGr, and the Raman and PL spectroscopies of the MoS₂ and NGr studied here after annealing.

D.S. was supported by the Graduate School of YONSEI University's research scholarship grants in 2017 (No. NRF-2018M3D1A1058924). D.S. and H.J.C. were supported by the Creative Materials Discovery Program through the National Research Foundation of Korea (NRF), funded by the Ministry of Science and ICT (No. 2018M3D1A1058536). K.S.K. was supported by the Priority Research Center Program (No. 2010-0020207) of the National Research Foundation (NRF) of Korea, funded by the Ministry of Education, and the Global Research & Development Center Program (No. 2018K1A4A3A01064272) of the NRF funded by the Ministry of Science and ICT. J.H. was supported by the NSF MRSEC program through Columbia in the center of Precision Assembly of Superstratic and Superatomic Solid (No. DMR-1420634). G.-H. L. was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (No. NRF-2017R1A2B2006568) and the Korea Institute of Energy Technology Evaluation and Planning (KETEP) and the Ministry of Trade, Industry & Energy (MOTIE) of the Republic of Korea (No. 20173010013340). Y.D.K. and J.J.L. were supported by Samsung Research & Incubation Funding Center of Samsung Electronics under Project Number SRFC-TB1803-04. This work was supported by a grant from Kyung Hee University in 2017 (No. KHU-20171743).