Introduction

Two-dimensional (2D) materials can be assembled into the so-called Van der Waals (vdW) heterostructures, offering an exotic platform for exploring many fundamental physics and important device applications1,2. One of the crucial structural parameters in a vdW heterostructure is its twist angle, a new degree of freedom for modulating its properties. Indeed, many fascinating twist angle-dependent properties have been investigated including the strong twist angle-dependent resistivity of Gr/graphite contact3,4, the quantum transport in Graphene/BN superlattice5,6,7, resonant tunneling in Gr/BN/Gr8, interlayer excitons in MoSe2/WSe29, and band evolution in MoS2/graphite10, just to mention a few. Among many investigations, unraveling the vertical conductivity of a vdW heterojunction is of fundamental importance; however, only limited knowledge has been established so far.

Here, we report the experimental investigation of vertical conductivities of MoS2/graphene (MoS2/Gr) heterojunctions under various lattice twist configurations. We found that vertical conductivities of these heterojunctions are strongly twist angle-dependent. When varying the twist angle from 0° to 30°, conductivities monotonically decrease and the modulation depth (maximum/minimum) is ∼5. Considering that MoS2/Gr heterojunction has great potential in various devices applications11,12 and even batteries13, graphene can also form excellent contact to MoS214,15,16,17; our results provide a guidance in using the MoS2/Gr heterojunction for electronics.

Results

Growth and characterization of MoS2/Gr heterojunction

In this study, the MoS2/Gr heterojunctions were fabricated by an epitaxial growth technique18,19 (also see Methods for more details). Figure 1a shows an optical image of as-grown MoS2 triangle domains on graphene. These domains have obviously preferred orientations and similar sizes, suggesting the epitaxy nature of MoS2 domains. Figure 1b shows the atomic force microscope (AFM) image of an area zoomed in Fig. 1a. It can be clearly seen that the surfaces of both MoS2 and graphene substrate are clean and free of contaminations. The height of as-grown MoS2 triangle domains is ∼0.855 nm, corresponding a monolayer thickness20,21. Raman and photoluminescence (PL) spectra also indicate that the monolayer MoS2 domains are of high qualities (Supplementary Figure 1). Selected area electron diffraction (SAED) was also used to characterize the lattice alignment of our MoS2/Gr samples. As illustrated in a typical pattern (Fig. 1c), the hexagonal diffraction spots of both MoS2 and graphene have the same orientations, suggesting an either 0° or 60° twist angle, which are geometrically equivalent between the as-grown MoS2 and graphene. To further demonstrate the high quality and the clean surface of these MoS2 domains, we further performed conductive AFM (C-AFM) imaging (Fig. 1d and Supplementary Figure 2). We can clearly see that the moiré superlattice with a period of ∼1.18 nm arises from the lattice mismatch between MoS2 and graphene (∼30%)22. Note that in Fig. 1c, the sample is very thin, leading to quite a large center transmission spot, and diffraction spots from the moiré superlattice are not visible in the SEAD pattern, since they are within the center transmission spot.

Fig. 1
figure 1

Structures of MoS2/Gr heterojunctions. a Optical image of as-grown MoS2 triangle domains on graphene, scale bar, 10 μm. b AFM image, scale bar, 2 μm. c SAED pattern of a typical MoS2/Gr heterojunction, scale bar, 5 nm−1. d C-AFM imaging of the moiré superlattice of MoS2/Gr after Fast Fourier Transform filter, scale bar, 5 nm. The period of superlattice is ∼1.18 nm as shown in the inset

Full size image

Twist-angle-dependent conductivities of MoS2/Gr heterojunction

To achieve different twist angles between the MoS2 domains and the graphene substrates underneath, we rotated these as-grown MoS2 domains via AFM-tip manipulation technique. This manipulation process is like that reported in our previous work18,23. As illustrated in Fig. 2a, during the rotation process, we firstly engaged the AFM tip to the graphene surface with a load of tens of nN and then pushed the MoS2 triangle from its corner to actuate its rotation. By controlling the tip moving direction and length, we thus can rotate the MoS2 domains on graphene with any deterministic twist angles, as illustrated in Fig. 2b–f. The heights of this MoS2 domain under different twist angles are likely to be different if considering the interlayer coupling; however, due to the probing limit (no more than 60 pm) of our AFM, these height variations were not observed in our experiments.

Fig. 2
figure 2

Rotation of as-grown MoS2 domains on graphene substrates. a Schematic of AFM-tip manipulation setup. b–f AFM images of a typical MoS2 domain rotated on graphene to achieve a series of twist angles, scale bar, 1 μm. Blue dash lines indicate the original direction of the MoS2 domain, while white arrows indicate the rotation directions. Green rectangles represent the scan area during C-AFM measurements

Full size image

Since MoS2/Gr heterostructures with deterministic twist angles are available, we thus measured their vertical conductivities to investigate the twist-angle dependence. In our experiments, the conductivity was directly measured under the C-AFM mode. During C-AFM scanning, a metal-coated AFM tip was placed in direct contact with the sample surface under a controlled load of ∼23 nN (cantilevers were calibrated by Sader’s method)24. The bias applied to the measurement circuit was fixed to be 1.5 V, and a 110-MΩ resistor was connected into the circuit to prevent current overload to our C-AFM holder, which has a measuring range of ±20 nA and a noise level of 1.5 pA, as illustrated in the inset of Fig. 3a. Note that, during scanning, we adjusted the fast scan direction of the AFM tip in parallel to one edge of the MoS2 triangle to reduce the feedback noise.

Fig. 3
figure 3

Twist-angle-dependent conductivities of MoS2/Gr heterojunctions. a A typical current mapping of a MoS2/Gr heterojunction with a twist angle of ∼28.03°, scale bar, 100 nm. Bright and dark areas in this mapping reflect the tip-to-sample current through the bare graphene and heterojunction, respectively. Inset is the experimental setup for current mapping. b Statistical chart derived from a. c Normalized current distributions of seven heterojunctions with different twist angles. d Statistic resistances of MoS2/Gr heterojunctions with different twist angles. Error bar is derived from the full width at half maximum of the MoS2 peaks, etch statistics contains at least 20,000 points

Full size image

Figure 3a shows a typical C-AFM current mapping of a MoS2/Gr heterojunction with a twist angle of 28.03°. The mapping can be clearly divided into two different regions: the bright region from the bare graphene surface with higher conductivities and the dark region from the MoS2/Gr heterojunction with lower conductivities. To quantitatively extract the conductivity meanwhile eliminating the metal at the two different regions, we thus inverted this current mapping into a statistical current distribution chart as illustrated in Fig. 3b. The x-axis of the chart is current and the y-axis is the counts, which is from an area of 500 nm × 500 nm, corresponding to 256 × 256 points. From the statistics, we can clearly see two Gaussian peaks whose centers reflect the statistic currents Ig and Ij flowing through the graphene and the junction, respectively. Because the resistance of the graphene is much smaller than the connected 110-MΩ resistor, Rg = V/Ig just reflects the system resistance and barely shifts. In order to facilitate comparison, we used Q = Ij/Ig to normalize the current distribution statistics.

Figure 3c shows the normalized current distribution statistics for a series of MoS2/Gr heterojunctions with different twist angles θ. The intensity comes from the scale of heterojunction and graphene area we chose to make statistics. It can be clearly seen that the two peaks are moving apart to each other when the MoS2/Gr heterojunction is twisted from 0° to 30°. As illustrated in Fig. 3d, we also calculated the resistances of the MoS2/Gr heterojunctions with different twist angles, which shows an obvious 60° period. (Details of resistance calculation, please see Supplementary Note 5) The resistance at θ = 30° is about 5 times larger than that at θ = 60° (0°). Error bar in Fig. 3d is derived from the full width at half maximum of the MoS2 peaks, which can characterize the dispersion of the heterojunction resistance. From the results, we can see that the dispersion of the heterojunction resistance has no clear correlation with twist angles. In addition to the measurements of these MoS2/Gr heterojunctions fabricated by AFM-tip-facilitated rotation, we also measured a polycrystalline MoS2/Gr heterojunction (Supplementary Figure 5), in which different twist angles naturally exist in an individual sample. Current mapping also shows bright and dark areas clearly separated by the 30° grain boundaries, consistent with the above observations. Note that we reproduced the above experiments many times on different samples. In each time we at least captured five current maps for an individual twist angle, and the overall results show good consistency.

Theorical simulations

As observed above, vertical conductivities across the MoS2/Gr heterojunctions are highly twist-angle dependent. Since the tip is in direct contact with the sample surface during C-AFM scanning and tip-to-MoS2 interface is not likely to have any lattice orientation dependent effects, this interface-modulated conductivities could be related to the stacking configurations between graphene and MoS2. The tunneling current can be described as25,26,27:

$$I\left( V \right) \propto {\int} {{d}E \cdot DoS_{\mathrm{g}}} \left( E \right) \cdot DoS_{\mathrm{t}}\left( {E - eV} \right) \\ \cdot \left[ \, {f\left( {E - eV} \right) - f\left( E \right)} \right] \cdot T\left( E \right)$$
(1)

where \(DoS_{\mathrm{g}}\left( E \right)\) and \(DoS_{\mathrm{t}}\left( {E - eV} \right)\) are the available tunneling density of states in graphene and the metal tip. f(E) is the Fermi–Dirac distribution. T(E) is the transmission coefficient. It can be seen that the tunneling current is dominated by the transmission coefficient under a certain tip-to-sample bias.

To unravel the variation of transmission coefficient in the MoS2/Gr heterojunctions, we thus performed density functional theory (DFT) simulations (please see Methods and Supplementary Note 7). For simplicity, we just calculated the transmission coefficients for θ = 0° and θ = 30° configurations and got T0 = 0.0015 and T30 = 0.000705, respectively. Two numbers are in the ratio 2.21:1, suggesting the tunneling current of the 0°-twisted heterojunction will be 2.21 times larger than that of the 30°-twisted heterojunction under a same bias. This result is in fair agreement with our observations, although our experimental results indicate that the ratio should ∼5. This experiment theory inconsistency might come from choosing structural parameters used for simulations, considering that little variation of structural parameters could lead to significant change on the simulated results. In our model, the periodic symmetry has been broken along the transport direction, but preserved on the flat straighten to the transport direction. As a result, a 2D Brillouin zone exists. Figure 4a and b show the simulated hot mappings of transmission coefficients in K-space for the 0°- and 30°-twisted heterojunctions, respectively, providing a hint on which momentum area dominates the transport mostly. It can be clearly seen that K points dominate transport mostly for the 0°-twisted heterojunction; in sharp contrast, Γ points contributes to transport mostly for the 30°-twisted heterojunction. In order to give a clearer relationship between the twist angle and the transmission coefficient, we also applied Wentzel–Kramers–Brillouin (WKB) method to generate transmission data under different twist angles as shown in Fig. 4c. We can clearly see that the transmission coefficient decreases monotonically as twist angles varying from 0° toward 30°, consistent with our experiment results. (Please see Methods for more details of WKB calculations.)

Fig. 4
figure 4

Theorical simulations. a, b Simulated K-space transmission hot mappings for the 0°- and 30°-twisted MoS2/Gr heterojunctions. c Transmission coefficient as a function of twist angles. Red points are calculated by WKB method and blue points are calculated by DFT

Full size image

Considering the evolution of MoS2/Gr heterojunctions with twist angles from 0° to 30°, we can simply treat them by rotating the Brillouin zone of graphene. At θ = 0°, K points for graphene and MoS2 are aligned. During rotating, graphene’s K points will move away from MoS2’s K points; at θ = 30°, graphene’s K points are at MoS2’s Γ points due to the band folding. From K point to Γ point, there is a k-path along which MoS2’s bandgap changes from minimum to maximum. As a result, the transmission coefficient of this system changes from the largest to the smallest, which is consistent with the above experimental observations shown in Fig. 3d.

Discussion

We investigated MoS2/Gr heterojunctions with different twist angles in terms of their vertical conductivities. We found that the vertical conductivities of these heterojunctions can be modulated up to ∼5 times by twist angles with the highest/lowest conductivity occurs at a twist angle of 0°/30°. DFT simulations suggest that this twist-angle-dependent conductivities originates from the transmission coefficient difference in the MoS2/Gr heterojunctions. Our work provides a guidance in using the MoS2/Gr heterojunction for electronics, especially concern on reducing the contact resistance in MoS2 devices as well as other TMDCs devices contacted by graphene.

Methods

Sample preparations

The MoS2 growth was performed in a three-temperature-zone chemical vapor deposition (CVD) chamber. S (Alfa Aesar, 99.9%, 4 g) and MoO3 (Alfa Aesar, 99.999%, 50 mg) sources were loaded in two separate inner tubes and placed at zone-I and zone-II, respectively. Substrates were load in zone-III. During the growth, Ar/O2 (gas flow rate: 75/3 sccm) are used as carrying gases and temperatures for the S-source, MoO3-source, and substrates are 115 °C, 530 °C, and 930 °C, respectively. Thin graphene flakes were mechanically exfoliated from both HOPG and Graphenium graphite (Manchester Nanomaterials) and placed on SiO2 substrates.

Sample characterizations

PL and Raman characterizations were performed in a Horiba Jobin Yvon LabRAM HR-Evolution Raman system using a 532-nm laser excitation wavelength. SAED was performed in a TEM (Philips CM200) operating at 200 kV.

AFM and C-AFM scanning

AFM imaging was performed by Asylum Research Cypher S with AC160TS tip. In C-AFM mode, we used ASYELEC-01 tip with Ti/Ir coating and the holder 901.730. The bias was applied from sample to tip with a current limit of ±20 nA and noise of 1.5 pA.

DFT calculations

We used DFT within the Keldysh non-equilibrium Green’s function (NEGF) formalism28. Succinctly, NEGF-DFT calculates density matrix by NEGF as \({\rho \sim {\int } {\mathrm d}EG^{ < }}\), and transmission coefficient as \(T = T_{r}\left[ {G^{r}\Gamma _LG^{a}\Gamma _R} \right]\). While, \(G^{r,a, < }\) are the retarded, advanced, and lesser Green’s functions, respectively, and ΓL,R are the self-energy of the left (L) and right (R) leads. Furthermore, the conductance \(G = T \cdot e^2{\mathrm{/}}h\), where e is the electron charge and h is the Planck constant. In our model. We choose multilayer graphene for both left lead and right lead, considering the lattice match problem. The structure has mirror symmetry centered on the plane of molybdenum atoms. The structure parameters have been chosen as: superlattice constant: 1.2816 nm; thickness of MoS2: 0.3227 nm (0°) and 0.3174 nm (30°); interlayer spacing of MoS2/Gr heterojunction: 0.3110 nm (0°) and 0.3131 nm (30°); interlayer spacing of graphene: 0.34 nm, all same with ref. 29.

WKB calculations

In WKB calculations, MoS2 was treated as a limited barrier; thus, the transmission coefficient has a simple relation with the twist angle, described as \(T = ae^{ - b\sqrt {E_{\mathrm{g}}} }\). Where, a, b are pending constants, T is transmission, and Eg is the bandgap. Since we already have two transmissions hot maps for two twist angles, i.e., 0° and 30°, which have been used to determine a and b in the equation. In order to obtain the relation between the tunneling gap and the twist angle, band-folding technique has been used.