Low-defect-density WS2 by hydroxide vapor phase deposition

As transistor gate lengths scale, ultra-thin channels are needed for electrostatic control, motivating interest in ~1 nm-thick TMD monolayers as next-generation channel materials. However, scalability via CVD commonly yields high defect densities (grain boundaries, point defects, strain) that degrade device performance. STM studies identify oxygen substitution at sulfur sites (O_S) as a dominant point defect in CVD WS2. Post-growth treatments (thiol chemistry and chalcogen annealing) can repair S vacancies but are ineffective for substitutional defects. WS2 is theoretically attractive due to high mobilities and saturation velocities, yet conventional growth via sulfidation of WO3 or oxygenated precursors still yields defect-rich films. Transport agents (H2O, O2) have been used to enhance metal-source volatilization, but their impact on defect formation has been underexplored. This study investigates whether using hydroxide vapor phase deposition (OHVPD) with volatile tungsten hydroxide intermediates can fundamentally alter sulfurization kinetics to minimize defects and improve electronic quality.

Prior work has established that mechanically exfoliated TMD monolayers outperform synthetic CVD films in electrical quality, but lack scalability. CVD WS2 often contains O-substitutional defects and other imperfections; STM has identified O_S as prolific in WS2 and oxygen-doped MoS2. Post-growth defect passivation (e.g., thiols) can mitigate charged S vacancies and improve mobility, while heavy oxygen doping or high S-vacancy densities can change band structure but are unsuitable for scalable electronics due to charged-defect scattering. Growth facilitation with H2O or O2 can assist precursor volatilization and domain growth but their effect on intrinsic defect populations has been unclear. Theoretical and experimental studies suggest neutral substitutional O_S and Mo_W have limited impact on band structure compared to charged defects; thus, minimizing charged defect density is critical for high mobility.

Growth: OHVPD employed a homemade 3-inch CVD system with high-purity W foil (99.95%) and sulfur. Moisture was introduced via Ar flow (180 sccm) at atmospheric pressure. S was heated to 180 °C upstream; W foil was held at 1050 °C (center), and sapphire substrates downstream at 950–800 °C. H2 (20 sccm) was supplied during growth for 15 min, followed by natural cooling in Ar/H2 without water vapor. In situ, W reacts with H2O to form volatile tungsten hydroxide WO2(OH)2, which is transported and reduced/sulfurized to WS2. For comparison, conventional CVD used WO3 powder (900 °C, 15 min, Ar/H2 at 10 Torr) and sulfur. Simulation of sulfurization kinetics: NEB calculations compared sulfurization at edges containing W–O versus W–OH. W–O bond length (2.061 Å) is shorter than W–OH (2.152 Å). Kinetic barriers to remove O and form H2O are higher for W–O (1.44 eV and 1.39 eV for key steps) than for W–OH (0.94 eV and 0.71 eV), and W–O requires two H atoms versus one for W–OH, indicating energetically favorable sulfurization via hydroxide intermediates. Materials characterization: Optical microscopy and AFM imaged domains (micron-scale) and continuity (inch-scale films). Raman (532 nm) and PL mapping assessed optical quality; low-temperature (4 K) PL deconvolution quantified exciton/trion/defect-bound exciton features. STM/STS at 77 K (imaging) and 5.3 K (dI/dV) on samples grown directly on HOPG quantified defect types/densities following established classifications (O_S top/bottom, Mo_W, positively/negatively charged defects, S vacancies). For transfer impact, monolayer WS2 grown on sapphire was transferred by PDMS-assisted method onto HOPG; STM quantified post-transfer defects (t-CVD, t-OHVPD). Device fabrication and measurement: Four-terminal back-gated FETs on 300 nm SiO2/Si used PMMA and e-beam lithography; contacts were Al/Au (5/65 nm). Measurements were in vacuum (≤1e-4 Torr) from 15–300 K; field-effect mobility μ_FE extracted from μ_FE = (dσ/dVg)/Cox at n ≈ 4.7×10^12 cm^-2. Short-channel back-gated FETs (L_CH = 100–400 nm) on SiNx (100 nm)/p++-Si used He-ion beam lithography; contacts were Bi/Au (20/15 nm). Transfer curves and output characteristics were recorded in high vacuum (10^-5–10^-6 Torr). Trap and charged impurity densities were estimated using a CI-limited mobility model, with Cox = 1.15×10^-8 F cm^-2 for 300 nm SiO2. Oxidation modeling: Formation energies and NEB barriers for O_S formation via O2 dissociation on pristine and O_S-containing WS2 were computed, showing thermodynamic favorability under O-rich conditions and reduced barriers with pre-existing O_S.

• OHVPD growth: Volatile WO2(OH)2 intermediates from W + H2O enable more favorable sulfurization kinetics than WO3-based CVD; key kinetic barriers for W–OH conversion are lower (0.41, 0.94, 0.71 eV) than for W–O (0.17, 1.44, 1.39 eV), supporting reduced oxygen incorporation.
• Optical quality: Raman statistics (n=50 per method) show narrower A1g peak for OHVPD-WS2 (avg FWHM 4.2 cm^-1) vs CVD-WS2 (5.5 cm^-1), and lower normalized LA(M) intensity in OHVPD, indicating fewer defects. Room-temperature PL exhibits higher peak energy and narrower FWHM for OHVPD-WS2. At 4 K, OHVPD has markedly reduced trion (X_T) and defect-bound exciton (X_D) intensities versus CVD.
• Defect densities by STM (on HOPG, 40×40 nm^2 fields, >20 images each): • CVD-WS2: O_S(top) 3.52×10^12 cm^-2; O_S(bottom) 3.46×10^12 cm^-2; Mo_W 9.85×10^11 cm^-2; NCD 3.0×10^10 cm^-2; PCD ~0; S vacancies 8×10^9 cm^-2; total 8.009×10^12 cm^-2. • OHVPD-WS2: O_S(top) 1.199×10^12 cm^-2; O_S(bottom) 1.177×10^12 cm^-2; Mo_W 9×10^9 cm^-2; NCD 8×10^9 cm^-2; PCD ~0; S vacancies ~0; total 2.393×10^12 cm^-2. • After transfer (PDMS) to HOPG: t-CVD total 2.1063×10^13 cm^-2 with charge impurities ~2.5×10^11 cm^-2; t-OHVPD total 2.488×10^12 cm^-2 with charge impurities ~2.0×10^10 cm^-2. t-CVD shows O_S(top) ≫ O_S(bottom), indicating top-side oxidation during transfer; NCD increases.
• Transport: OHVPD-WS2 shows a metal–insulator transition at Vg ≈ 60 V (n ≈ 4.3×10^12 cm^-2) and ≥10× higher conductivity than typical CVD-WS2. Extracted densities (model-based): OHVPD trap density Nt ≈ 3.6×10^12 cm^-2; charged impurity density Ncr ≈ 8.7×10^10 cm^-2 (lowest among reported MoS2/WS2 monolayers in cited comparisons). For CVD-WS2: Nt ≈ 8.2×10^12 cm^-2; Ncr ≈ 2.2×10^12 cm^-2.
• Mobility: Four-probe μ_FE ≈ 198 cm^2 V^-1 s^-1 at 300 K and 789 cm^2 V^-1 s^-1 at 15 K for OHVPD-WS2 at n ≈ 4.7×10^12 cm^-2, among the highest for synthetic monolayer WS2 and comparable to exfoliated WS2. CVD-WS2 yields ~17 cm^2 V^-1 s^-1 (300 K) and 105 cm^2 V^-1 s^-1 (15 K).
• Short-channel FETs (L_CH = 100 nm): OHVPD-WS2 achieves I_on = 403 μA/μm at V_ds = 1 V with I_on/I_off ~10^8, surpassing CVD-WS2 processed identically. Device hysteresis is low (normalized width ~40 mV per MV cm^-1).

Using water vapor to generate volatile tungsten hydroxide intermediates fundamentally changes sulfurization chemistry, lowering kinetic barriers for O removal and suppressing oxygen incorporation and other impurities (e.g., Mo from WO3). STM confirms an approximately order-of-magnitude reduction in total defect density for OHVPD WS2 versus CVD, especially in charged defect densities that most strongly scatter carriers via Coulomb interaction. Neutral substitutional defects (O_S, Mo_W) do not introduce in-gap states and minimally affect band structure at observed densities, aligning with the high mobilities achieved. The reduced charged impurity density enables a clear metal–insulator transition and near-exfoliated mobility values. Transfer processes can induce additional O_S(top) and NCDs, particularly for CVD films with higher initial O_S densities; OHVPD films are more robust to transfer-induced degradation, though gentle transfer remains important. The electrical gains (high μ_FE, high I_on, and large I_on/I_off) demonstrate that growth-route defect minimization is a pivotal lever toward scalable, electronic-grade 2D TMDs.

The study introduces hydroxide vapor phase deposition (OHVPD) for monolayer WS2, leveraging volatile WO2(OH)2 intermediates to provide energetically favorable sulfurization and substantially reduce defect densities. OHVPD-WS2 exhibits superior optical quality, an order-of-magnitude lower total and charged defect densities than CVD, room-temperature four-probe electron mobility ~198 cm^2/Vs (789 cm^2/Vs at 15 K), and high short-channel current (≈403 μA/μm at 1 V) with ~10^8 I_on/I_off. These metrics approach those of exfoliated flakes while maintaining scalability, highlighting OHVPD as a route toward electronic-grade 2D semiconductors. Future work should focus on further minimizing charged impurities and interface traps (including from WS2–dielectric interfaces and lithography), optimizing gentle transfer/encapsulation, and extending hydroxide-mediated growth to other TMDs and wafer-scale integration.

• Quantitative STEM imaging was avoided due to potential electron-beam damage and difficulty distinguishing O_S from S vacancies; defect quantification relied on STM, which samples surfaces and may not capture all subsurface/interface defects.
• Post-transfer and device fabrication introduce additional charge impurities and O_S(top), particularly in higher-defect CVD films; some extracted charged impurity density likely arises from WS2–SiO2 interfaces and lithographic processes.
• Mobility modeling neglected phonon-limited scattering (assumed much higher than experimental values) and focused on charged impurity limitation, which may not capture all scattering mechanisms under different conditions.
• Generalization to other TMDs was only briefly demonstrated (MoS2 optical quality); broader validation and uniformity at wafer scale require further study.