Tuning transport across MoS2/graphene interfaces via as-grown lateral heterostructures

An unexploited property of graphene-based heterojunctions is the tunable doping of the junction via electrostatic gating. This unique property may be key in advancing electronic transport across interfaces with semiconductors. Here, we engineer transport in semiconducting TMDs by constructing a lateral heterostructure with epitaxial graphene and tuning its intrinsic doping to form a p–n junction between the graphene and the semiconducting TMDs. Graphene grown on SiC (epitaxial graphene) is intrinsically doped via substrate polarization without the introduction of an external dopant, thus enabling a platform for pristine heterostructures with a target band alignment. We demonstrate an electrostatically tunable graphene/MoS2p–n junction with >20× reduction and >10× increased tunability in contact resistance (Rc) compared with metal/TMD junctions, attributed to band alignment engineering and the tunable density of states in graphene. This unique concept provides improved control over transport across 2D p–n junctions.


INTRODUCTION
Engineering p-n junctions in semiconductors requires choosing the right material combination for a target band alignment. This challenge is only amplified for two-dimensional (2D) materials, specifically semiconducting transition metal dichalcogenides (TMDs) owing to their ultra-thin nature, their layered structure, and the difficulties involved with achieving sufficiently high doping densities 1 . While there are previous studies showing the direct synthesis and properties of lateral p-n junctions in TMD-TMD heterostructures (e.g., MoS 2 /WSe 2 , MoS 2 /WS 2 ), they lack wide tunability in relative band alignments that enable minimization of interlayer resistance. Graphene grown on silicon carbide (SiC), dubbed epitaxial graphene (EG) 2 , provides an additional chemical degree of freedom to control the potential for only the graphene component in the heterostructure. A unique property of EG, not exploited previously in 2D heterostructures, is the ability to electrostatically modulate the intrinsic doping from nto p-type due to the spontaneous polarization of SiC and tunability of the EG/SiC interface via intercalation 3 . Building from our successful synthesis of lateral heterostructures between EG and MoS 2 4 , we demonstrate the formation of a low resistance lateral p-n junction between EG and molybdenum disulfide (MoS 2 ), which allows for configurable electronic properties. The EG is grown via silicon sublimation on SiC substrate to contact MoS 2 synthesized via powder vaporization. We demonstrate that the EG/MoS 2 heterostructure has a pristine interface and allows for the EG to reduce the contact resistance by >20×. This is an improvement over utilizing chemical vapor deposited (CVD) graphene, where gaseous carbon sources (e.g., methane, ethane, etc), and metal substrates (e.g., copper or nickel) are used and the graphene product is subsequently transferred onto the device substrate; the transfer process typically involves a polymer-based wet chemical etch 5 , which often introduces contamination into the system and deteriorates the quality of the contacts [6][7][8][9] . With EG contacts, the challenges involved with controllable large-area transfer is eliminated making this a scalable, as-grown process.

RESULTS AND DISCUSSION
Electrostatic doping modulation of graphene on SiC The existence of a layer of carbon atoms that are partially covalently bonded to the SiC substrate (buffer layer) between the graphene and SiC, in combination with the polarization of SiC, leads to a moderate n-type doping of as-grown EG 3 . The doping can be altered by intercalating hydrogen at the interface between the buffer and SiC, often accomplished by annealing the EG at elevated temperatures in a hydrogen-rich environment 3 . Hydrogen diffuses through defects and wrinkles in the EG, passivates dangling bonds at the buffer/SiC interface, and converts the buffer layer to an additional layer of graphene 3,5 . This hydrogen intercalated EG is termed quasi-freestanding EG (QFEG) 5 , and becomes p-type doped to compensate for the negative surface charge of the (H-passivated) Si-terminated SiC surface, so that a diverging electrostatic energy from charge-uncompensated surfaces is avoided 3 . Passivation of the buffer layer, and simultaneous formation of QFEG via hydrogen intercalation leads to characteristic shift of the C1s spectra in x-ray photoelectron spectroscopy (XPS), as shown in Fig. 1a 5 . The buffer layer initially present in the EG comprises of the S1 and S2 components, representing the sp 2 and sp 3 hybridized carbon in the partially covalently bonded layer. This buffer layer is removed upon hydrogenation, thus converting a monolayer of EG (+buffer) to a bi-layer QFEG system. Furthermore, in the transfer measurements of electrostatic double layer (EDL)-gated field-effect transistors (FETs), the location of the Dirac point shifts from negative for EG to positive for QFEG, further confirming the change in doping from n-type to p-type upon hydrogenation (Fig. 1b). Here, the term Dirac point is used to locate the voltage at which the current reaches its minimum in the ambipolar graphene, thus applicable for both monolayer EG and bi-layer QFEG. Electronic band alignments of n-type EG and p-type QFEG with conventional metals (e.g., titanium, gold), and commonly studied TMDs 10 are shown schematically in Fig. 1c 2 .
Device characteristics of EG/MoS 2 and QFEG/MoS 2 lateral heterostructures Direct synthesis of MoS 2 /EG lateral heterostructures enables configurable transport properties of MoS 2 because the carrier type of EG is predetermined by hydrogen intercalation, and the heterostructure system is reconfigurable via EDL gating. We employ p-type QFEG to synthesize p-n junctions with n-type MoS 2 . The advantage of using this system is twofold: (1) this system exhibits a significant reduction in contact resistance, when compared with conventional metals, and (2) unlike the EG/MoS 2 system, the QFEG/MoS 2 is a p-n junction, which leads to greater tunability via EDL gating. In fact, the contact resistances of QFEG/ MoS 2 can change by >10× upon application of gate voltage V g (Fig. 2a). Thus, the controlled doping modulation of EG allows for the formation of these unique graphene/MoS 2 heterostructure p-n junctions, which have lower contact resistances than conventional metals.
To evaluate the tunability, three types of structures are constructed and characterized: QFEG/MoS 2 ( Fig. 2 and Supplementary Fig. 2), EG/MoS 2 ( Fig. 2 and Supplementary Fig. 2), and metal/MoS 2 ( Supplementary Fig. 2). EG (and QFEG) is grown (and hydrogenated) on SiC substrates and patterned into rectangular contact pads using optical lithography. Subsequently, MoS 2 is grown using powder vaporization of molybdenum tri-oxide (MoO 3 ) and sulfur (S), as described in ref. 4 . Such lateral heterostructures exhibit a pristine interface with an overlap of MoS 2 onto the edge of the graphene pattern (see Supplementary Note 1) 4 . Achieving a clean interface enables efficient electronic coupling between EG (or QFEG) and MoS 2 , leading to an effective graphene contact to MoS 2 . We note that the MoS 2 grown via powder vaporization (PV) is n-type doped due to the combined effects of a low vacuum-level-referenced band edge positions 11 (relative to other TMDs) and the presence of defects and typical non-uniformity based on the PV method 12 .
The p-n junction barrier, and consequently the resistance at the junction (contact resistance (R c )), is evaluated via conventional transfer length measurements, also known as transmission line measurements (TLM). EDL-gated transistors are fabricated post growth using a side gate geometry (inset, Fig. 2a). QFEG and EG exhibit comparable contact resistance values at a gate voltage (V g ) of +3 V, and >20× reduction in R c compared with conventional titanium/gold (Ti/Au) metal stacks (see Supplementary Note 2). This tunability can be attributed to the low density of states in graphene 13,14 compared with metal. The constructed lateral heterostructure that allows the EDL gating of the EG as well as MoS 2 , providing a means of electrostatic modulation of the metal/ EG/MoS 2 band alignment and a gate-dependent contact resistance. Thus, EG and QFEG are considered an electrostatically tunable contacts. Such tunable contacts are key in various chemical and optical sensors [15][16][17][18] . The ability to tune the Fermi levels using V g and locking the system at a predetermined electronic configuration is of great interest in the sensor community. Graphene offers a unique contact to other 2D materials such as MoS 2 because the Fermi level of graphene can be readily modulated, along with the semiconducting MoS 2 , providing an additional mode of tunability absent in metal contacts. Electrostatically tunable EG provides an additional "knob" with its unique property of enabling electrostatic modulation of the intrinsic doping from nto p-type 3 .
The QFEG/MoS 2 interface resistance can be modulated by >10× as a function of V g (−3 to +3 V), providing evidence of a highly tunable p-n junction. The large modulation of QFEG/MoS 2 is correlated with a larger electronic barrier as compared with the case of EG, thus making it sensitive to an external electric field and enabling efficient charge separation. The sheet resistance (R s ) (Fig.  2b) exhibits a similar trend as R c , also representative of the barrier at the QFEG/MoS 2 p-n junction. Although, the R c and R s values are lower for the EG/MoS 2 case, the absence of a p-n junction does not allow this lateral EG/MoS 2 heterostructure to be employed in many cases aside from unipolar FETs (see Supplementary Note 3). For functionality that requires a p-n junction and charge separation, the QFEG/MoS 2 system is a unique solution that takes advantage of the high conductivity of graphene that allows for improved electronic and thermal transport in the system, which would not be possible in a TMD/TMD p-n junction.
Visualization and calculations of band alignments Electronic band structure alignments are key to understanding the heterojunctions formed at interfaces of 2D materials. To further our understanding of the tunability of the p-n QFEG/MoS 2 lateral heterostructure system, we investigate the relative electronic band alignments experimentally and verify it using density functional theory (DFT). Angle-resolved photo-emission spectroscopy (ARPES) is used to directly visualize the band alignments, and contrast QFEG/MoS 2 and EG/MoS 2 junctions. As shown in Fig.  3a, b, the graphene 19 as well as the averaged MoS 2 valence bands are compared, including the MoS 2 at the interface and away from it 20 . ARPES across the QFEG/MoS 2 interface in Fig. 3a reveals two valence bands filled below the charge neutrality point typical of ptype bi-layer graphene, whereas the MoS 2 valence band now extends up to approximately −1 eV binding energy at Γ. In contrast, for ARPES across the EG/MoS 2 interface shown in Fig. 3b, we observe a complete Dirac cone filled above the Dirac point at the K-point, indicative of n-type monolayer graphene, while the intense valence band signature of MoS 2 appears at approximately −1.3 eV at Γ. The band bending diagrams in Fig. 3c, d quantify the band alignment for the QFEG and EG cases, respectively, where the MoS 2 levels can be deduced from the respective work functions of QFEG and EG, reported to be ϕ QFEG ¼ À4:7 eV and ϕ EG ¼ À4:2 eV (see Supplementary Note 4 for comparisons with theoretical values from literature). We focus on quantifying the valence band maximum (VBM) level relative to vacuum (i.e., before contact) rather than the Fermi level in MoS 2 since the former is intrinsic to the material. For EG, µ-ARPES reveals a spread in the VBM from −1.5 to −2.2 eV, as indicated by the two green arrows and consistent with direct measurements of band bending from nano-ARPES data (published elsewhere) where an upwards bending of MoS 2 bands toward the MoS 2 /graphene interface was found. Thus, the MoS 2 VBM relative to the vacuum level can be estimated to be À4:2 À 1:5 ¼ À5:7 eV. Similarly, for QFEG, we estimate the band bending from the VBM spread in µ-ARPES from −1.0 to −1.8 eV, yielding an estimated MoS 2 VBM of −5.7 eV. Although, estimating VBM positions from the magnitude of band spreading in µ-ARPES is less direct than nano-ARPES (since the band bending may not reach asymptote values limited by the laser spot size), the fact that the EG and QFEG cases gave identical values suggest reasonable accuracy. The measured VBM is compared with the vacuum-level-referenced VBM calculated for monolayer, bi-layer, and bulk MoS 2 as shown in Fig. 3e from manybody perturbation theory within the GW approximation to the electronic self energy [21][22][23][24] . Considering that further accounting for the dielectric screening from the substrate 25 and carriers 26 would yield a VBM higher by~0.2 eV, the quasiparticle ionization potential of the bilayer case at −5.9 eV appears to be the best match. This is also corroborated with a cross-sectional transmission electron micrograph in Supplementary Note 5 showing a bilayer MoS 2 in the channel.

Outlook
In summary, we have established a facile route to tune the transport properties of MoS 2 -based devices by introducing tunable graphene in the system leading to the synthesis of tunable p-n junctions. This allows for improved electronic conduction compared with conventional p-n junctions with two semiconducting components. This study exploits the property of tuning the doping of EG via hydrogen intercalation to form p-type doped QFEG and using it as one of the components of the p-n junction. We have also compared the QFEG/MoS 2 p-n junction with the previously synthesized EG/MoS 2 n + -n junction and studied the electronic band alignments via ARPES and corroborated it using theoretical calculations. Along with this in-depth knowledge of band alignments in as-grown heterostructures and the band alignment diagram shown in Fig. 1c, this study enables application-specific growth of heterostructures with target band alignments and study of interfaces.

Synthesis of EG and QFEG
EG is grown in a three-phase, hot-zone, graphite furnace (Thermal Technology LLC) via silicon sublimation from the 6H SiC(0001) face. The SiC is first cleaned using organic solvents and then annealed in 10% hydrogen (balance argon) at 1500°C for 30 min to eliminate subsurface damage due to chemical and mechanical polishing. The H 2 is then removed from the system, and the temperature is increased to 1800°C for 10 min at 500 Torr to form the graphene layers. This process yields low defect density multi-layer EG. Upon annealing the EG in 100% H 2 at 900°C for 30 min at a pressure of 600 Torr, the hydrogen intercalates and converts the n-type EG to a p-type QFEG 4 .

Synthesis of MoS 2
Powder vaporization in a horizontal quartz tube furnace is used to synthesize MoS 2 , using 2-3 mg of MoO 3 powder and 200 mg S as the precursors. The system is purged with argon to atmospheric pressure (~710 Torr), and S is independently heated to 130°C to provide a continuous S flow during the MoS 2 growth, which occurs at 800°C for 15 min 4 .

Device fabrication
Prior to MoS 2 growth, both EG and QFEG are patterned using standard ultraviolet photolithography. A TLM pattern of graphene contacts with a variable channel spacing (ranging from 0.75 to 10 µm) is produced using a photoresist double stack (PMGI SF2 + SPR 3012) exposed using a Stepper 8500. A mixture of oxygen and argon (O 2 /Ar) is used for a reactive ion etch to remove the EG outside of the patterns, leaving behind a series of periodically spaced graphene rectangles of fixed length (5 µm), which ultimately constitute the contacts to the MoS 2 channel for transistor fabrication. After the synthesis of the MoS 2 /EG or MoS 2 /QFEG heterostructure, an (SF 6 )/O 2 reactive ion etch is used to remove the excess MoS 2 , thereby isolating any MoS 2 growth strictly between the graphene electrodes. Subsequently, contact regions are lithographically patterned, briefly exposed to an O 2 plasma, and Ti/Au (5/15 nm) metal is deposited via electron-beam evaporation, followed by lift off in PRS 3000 photoresist remover 4 .

Electrolyte deposition and I-V measurements
The polymer electrolyte preparation details are the same as published by Xu et al., where the electrolyte is prepared and deposited in an argon-filled glove box with concentrations of H 2 O and O 2 < 0.1 parts-per-million (ppm) 27 . The solid polymer electrolyte (PEO) 76 :CsClO 4 is deposited on the device by drop-casting onto ∼1 × 1 cm 2 substrates in an argon ambient and annealed on a hotplate at 80°C. The sample is then transferred from the glove box to the probe station through an Ar-filled load lock. The entire process of electrolyte preparation, deposition, transfer to the probe station, and measurement is completed under an inert gas environment with no sample exposure to ambient. Electrical measurements are performed on a Lake Shore cryogenic vacuum probe station (CRX-VF) under ∼10 −6 Torr at 300 K using a Keysight B1500A semiconductor parameter analyzer. For transfer curves, the voltage sweep rate is optimized at ∼0.67 mV/s to minimize hysteresis. Four TLM sets (0.75-10 µm) were measured both for the heterostructures and the MoS 2 -only devices to extract contact resistance and sheet resistance. For transfer curves, five devices were measured for each channel length, and several such devices with different channel lengths were measured. Supplementary Note 2 shows the repeatability of the data and its variation with different voltage sweep rates and hysteresis.

ARPES measurements
ARPES experiments were performed at the μ-ARPES end-station of the Microscopic and Electronic Structure Observatory at the Advanced Light Source in Berkeley, USA. As-grown samples were introduced into ultra-high vacuum (<5 × 10 −11 mbar) and outgassed at 100°C. Data were collected at 80 K temperature with p-polarized light of hν = 145 eV photon energy, employing a hemispherical Scienta R4000 electron analyzer equipped with custom made deflectors 19,20 . These enable the collection of ARPES spectra over an entire Brillouin Zone without moving the sample and thus allow to retain the X-ray focus on the small sample structures without drift. The synchrotron beam-spot size was on the order of 20 μm, on the order of the feature size of the graphene/MoS 2 structure. ARPES thus integrates over band bending effects across the interface, which explains the valence band smearing in the data of Fig. 3d.

Theory calculation
For the GW calculations for monolayer (bilayer) MoS 2 , we employ a 18 × 18 × 1 k-point grid, a truncated Coulomb interaction 28 , a planewave energy cutoff of 15 Ha, and 250 (500) empty bands for the evaluation of the dielectric matrix within the random-phase approximation. The Godby-Needs-generalized plasmon-pole model 29 was used to extend the static dielectric matrix to finite frequencies. Correction to the dielectric matrix and the self energy were applied with and extrapolar energy of 2.7 Ha to accelerate the convergence of absolute band levels with respect to the empty bands included in the summation 30 . Quasiparticle shifts (relative to DFT levels) are then superimposed on DFT levels referenced from the vacuum level to obtain the final level alignments. All calculations are performed using the ABINIT code 31 .

DATA AVAILABILITY
The data that support the findings of this study are available from the authors on reasonable request, see author contributions for specific data sets.