Bridging the gap between atomically thin semiconductors and metal leads

The study addresses a central challenge in two-dimensional transition metal dichalcogenide semiconductor (TMDSC) devices: forming efficient electrical contacts to metal leads without detrimental interface barriers (tunnel or Schottky). Conventional contacts suffer from Fermi-level pinning and defect-induced gap states, resulting in large contact resistance and limited low-temperature performance, which hinders exploration of quantum transport phenomena. The authors propose a local bonding distortion (LBD) strategy to engineer the metal–TMDSC interface at the atomic level to create nearly barrier-free, robust ohmic contacts from room to cryogenic temperatures. The goal is to enhance carrier injection, reduce contact resistance towards the quantum limit, and enable high-mobility transport necessary for probing unconventional physics in TMDSCs.

Prior efforts to minimize contact barriers fall into: (I) direct metallization via chemical doping or phase transformation (e.g., inducing metallic 1T phases) and (II) van der Waals (vdW) contacts using tunnel barriers (e.g., h-BN) or graphene/soft-landed metal electrodes. While graphene electrodes with additional gating can mitigate Schottky barriers in few-layer MoS2, the inherent vdW gap (3–4 Å) still imposes a tunneling barrier limiting carrier injection. Edge contacts can in principle improve orbital overlap with transition-metal atoms but typically suffer from Fermi-level pinning and Schottky barriers due to dangling bonds. Plasma-induced phase transitions (e.g., argon-plasma to 1T) often create vacancies to stabilize phases, introducing defects. There remains a need for a controllable, defect-minimizing approach that enhances orbital hybridization and reduces injection barriers in both edge and top-contact geometries, especially effective at cryogenic temperatures.

Concept and mechanism: The LBD strategy locally distorts the TM–chalcogen coordination in TMDSCs from trigonal prismatic (2H, semiconducting) to octahedral derivatives (T/T′-like, semi-metallic) within a nanoscale region at the metal contact. This rearranges the d-orbital energy splitting of TM4+ ions, partially filling degenerate orbitals to yield itinerant electrons and a dispersed density of states (DOS) within the pristine band gap, enabling nearly barrier-free injection by enhancing orbital overlap/hybridization with the metal. Unlike bulk 1T transitions or vacancy-inducing treatments, the LBD is nanoscale, defect-free (no vacancies/dangling bonds), and exhibits tunable work function.

Process implementation: A referenced reactive ion etching (RIE) workflow is used to define contact windows with precise depth control by simultaneously etching a reference region whose optical contrast is monitored. After patterning a thin PMMA mask by e-beam lithography, CHF3/O2 RIE shapes the device and opens windows. A soft O2 plasma treatment is then applied in a soft-landing configuration (parallel electric field along sample surface, reduced plate bias, low power) to induce LBD on exposed TMDSC edges (edge contacts) or top surface (top contacts). Immediately afterwards, metal electrodes are deposited under high vacuum (∼3 × 10^-7 Torr) at slow rate (0.2 Å/s) to seal the windows and minimize deposition-induced effects.

Geometries: Both edge-contact (BN/TMDSC/BN stack cut by referenced RIE to expose edges) and top-contact (top BN locally removed to expose top TMDSC surface) configurations are demonstrated. Edge contacts maximize orbital overlap to metal; top contacts offer larger injection area though retain a vdW gap.

Atomic-scale characterization: Cross-sectional STEM-ADF imaging at 60 kV with aberration correction resolves the interface. In MoS2 edge contacts, a ∼1 nm-wide distorted region shows Mo and S atoms displaced from 2H positions into octahedral coordination; terminal Mo atoms bond to metal atoms at ∼1.8 Å (shorter than vdW gap), evidencing strong hybridization. In WSe2 top contacts, the topmost layer within the contact window undergoes LBD forming mixed 1T/1T′-like structures without defects; underlying layers remain 2H. 2D Gaussian fitting of atomic columns quantifies bond-length/angle variations. EELS detects oxygen in the distorted region, consistent with oxygen-substitution-induced bond rearrangement stabilizing the distortion; Raman spectroscopy corroborates local structural changes.

Device fabrication and materials: BN-encapsulated few-layer MoS2 and WSe2 heterostructures are assembled by dry transfer in a glovebox (top BN 5–8 nm; bottom BN 12–20 nm). Contacts use Ti/Au or other metals (e.g., Pd). RIE system (STS Pro) parameters: Etching—200 W at 13.56 MHz, CHF3:O2 = 40:4 sccm, base pressure 6 mTorr, substrate bias 20 W, 5–15 s; Plasma treatment—100 W at 13.56 MHz, O2 100% at 40 sccm, base pressure 6 mTorr, 0 W substrate bias, 10 s.

Electrical measurements: Two-probe I–V (Keithley 6430); four-probe conductance and magnetotransport (lock-in, SR830 and DS360). Cryogenics: Oxford Instruments down to 1.4 K, 15 T; He3 holder for 0.3 K. Mobilities extracted: field-effect (μFE from conductance vs gate and Hall-calibrated capacitance), Hall (μHall from Hall coefficient), quantum (μQ from SdH onset B as μQ = 7/B). Contact resistance Rc extracted from R2p–R4p differences, scaled by geometry. Transmission line measurements (TLM) performed for cross-validation.

Computations and simulations: DFT (Quantum ESPRESSO v6.1) with SG15 ONCV pseudopotentials, SOC included; structures relaxed with optB88-vdW; electronic structures with PBE. Multislice STEM ADF simulations (Kirkland codes) include frozen phonons and source-size convolution. Cross-section sample prep by FIB with protective Pt straps and low-kV thinning; final Ar ion cleaning.

Controls and reproducibility: Imaging conditions validated to avoid e-beam artifacts; pristine monolayer MoS2 imaged as control. Devices show reproducible contact quality across multiple samples and after storage; minimal hysteresis indicates absence of trap states from LBD.

• LBD creates a nanoscale (∼1 nm) semi-metallic region at the metal–TMDSC interface without vacancies or dangling bonds, enabling strong orbital hybridization and nearly barrier-free injection.
• Atomic-scale STEM shows direct Mo–metal bonding at ∼1.8 Å in edge-contacted MoS2; WSe2 top-contact LBD forms mixed 1T/1T′-like distorted layer confined to the topmost layer in the contact area.
• 3L-MoS2 edge-contacted FETs: robust linear (ohmic) I–V from 300 K to 2 K; contact resistance Rc ≈ 90 Ω·μm (∼270 Ω absolute at 2 K), approaching quantum limit (estimated ∼30 Ω·μm at n = 1×10^13 cm^-2). Field-effect mobility μFE = 556 cm²V^-1s^-1 at 300 K and 23,700 cm²V^-1s^-1 at 2 K; temperature exponent γFE ≈ −1.9, indicating strong phonon scattering. Channel current at 2 K is ∼200× that at 300 K.
• 1L-MoS2 top-contacted devices (with LBD): high-quality quantum Hall detection, excellent subthreshold swing; μFE ≈ 9,900 and μHall ≈ 9,200 cm²V^-1s^-1 at cryogenic temperatures.
• 5L-WSe2 top-contacted FETs: linear I–V down to 0.3 K; Rc ≈ 700 Ω·μm (limited by vdW gap in top contact). Achieved μFE ≈ 358,000 cm²V^-1s^-1, μHall ≈ 151,000 cm²V^-1s^-1, and quantum mobility μQ ≈ 25,000 cm²V^-1s^-1 at 0.3 K; clear SdH oscillations with onset near |B| ≈ 0.4 T.
• TLM confirms Rc: 85 ± 15 Ω·μm in edge-contacted 3L-MoS2; 100–200 Ω·μm in n-type top-contacted WSe2 across carrier densities and temperatures; two- vs four-terminal extraction consistent.
• Device polarity with LBD is contact-metal dependent (e.g., Ti/Au yields n-type access; Pd yields p-type in WSe2), consistent with dispersed DOS in LBD region coupling to different metal work functions.
• Soft O2 plasma both supplies kinetic energy for bond distortion and enables limited O substitution for local bond-length adjustment, stabilizing distortion; Raman and EELS support oxygen incorporation and structural changes.
• No measurable gate-sweep hysteresis; contact quality persists after long-time storage; multiple devices show high performance with reasonable variation.

By locally distorting the coordination of TM–chalcogen polyhedra to octahedral derivatives, the LBD region becomes semi-metallic with a dispersed DOS across the band gap. This DOS bridges the metal Fermi level and TMDSC band edges, enhancing orbital hybridization with the metal and enabling efficient carrier injection to conduction or valence bands with minimal Schottky/tunnel barriers. The observed robust ohmic behavior from room temperature to cryogenic temperatures, low Rc approaching quantum limits (in MoS2 edge contacts), and extremely high mobilities (in WSe2 top contacts) directly address the long-standing contact bottleneck in TMDSC devices. Unlike bulk 1T/H interfaces with fixed work function (leading to metal-independent polarity), the LBD’s distributed DOS allows coupling to different metal work functions, enabling metal-dependent polarity (n- or p-type access). The method’s defect-free nature (no vacancies/dangling bonds), atomic-scale confinement, and compatibility with standard cleanroom processes make it broadly applicable. These advances unlock low-temperature quantum transport studies in TMDSCs (e.g., SdH oscillations at very low B, potential for fractional quantum Hall in monolayer MoS2) and improve practicality for electronic applications.

The work introduces a controllable, soft O2 plasma–induced local bonding distortion (LBD) method to form high-quality electrical contacts to atomically thin TMDSCs in both edge- and top-contact geometries. Atomic-resolution cross-sectional microscopy confirms a ∼1 nm, defect-free, octahedral-derivative, semi-metallic interfacial region that bridges metal leads with pristine TMDSCs. Devices exhibit robust ohmic characteristics, ultralow contact resistance (down to 90 Ω·μm in 3L-MoS2), and record-high low-temperature mobilities (μFE up to 358,000 cm²V^-1s^-1 in 5L-WSe2), with clear quantum oscillations. The approach is reproducible, cleanroom-compatible, and enables metal-dependent polarity, positioning it as a versatile solution for contact engineering and as a platform for quantum transport studies in TMDSCs. Future work could optimize LBD size and composition to further reduce Rc, tailor p/n contacts via metal choice and LBD tuning, extend to other 2D semiconductors, and integrate at scale for van der Waals device architectures with locally engineered electronic properties.

• In top-contact geometry, a residual vdW gap leads to higher Rc (∼700 Ω·μm) than edge contacts, limiting ultimate injection efficiency.
• Device-to-device variation is larger in WSe2, likely due to a higher degree of mixed octahedral derivatives in LBD-treated WSe2 compared to MoS2.
• Monolayer MoS2 mobility remains relatively lower than few-layer due to weaker screening and stronger impurity scattering, despite improved contacts.
• The quantum mobility estimate from SdH onset is a lower bound due to measurement noise and environment; precise determination requires lower-noise setups.
• Process performance depends on precise RIE/plasma conditions; although reported reproducible, scalability may require tight process control and monitoring of oxygen incorporation.
• Contact resistance depends on carrier density and temperature; achieving quantum-limit Rc requires high carrier densities and low temperatures.