Formation of metal vacancy arrays in coalesced WS2 monolayer films
ADF-STEM shows linear arrays of W vacancies in coalesced CVD monolayer WS\(_2\); the authors connect them to gas-phase precursor access and substrate catalysis using ReaxFF MD of WS\(_2\) edges on (0001) α-alumina.
Summary¶
Monolayer MOCVD WS\(_2\) often grows as islands that later coalesce. The letter links ADF-STEM line defects to ReaxFF MD in which a gas-phase W–S–SH precursor (modeled with WS\(_2\)(SH)\(_2\)) approaches two parallel WS\(_2\) growth edges on (0001) α-Al\(_2\)O\(_3\). The key parameter is the lateral separation between those edges. For a 1.6 nm channel, the main-text reaction path (Fig. 2) lets the precursor access the sapphire surface, strip –SH groups, and exothermically add a WS\(_2\) unit; the illustrated trace lists barriers of about ~25 and ~75 kcal mol\(^{-1}\) for the two –SH removal steps, and a net release of about ~125 kcal mol\(^{-1}\) overall (Fig. 2(b), main text). For a 0.2 nm gap (Fig. 3), access to the catalytic substrate is blocked and a barrier > 100 kcal mol\(^{-1}\) in step (iii) disfavors the decomposition pathway, consistent with W-deficient lines in STEM along tight coalescence junctions. The W/S/H ReaxFF is merged with Hong-style Al/O (and related) parameters so the substrate participates in the chemistry, as written in the article.
Methods¶
1 — MD application (ReaxFF growth edges). Engine: ReaxFF molecular dynamics (main text; full thermostat/timestep/integrator details in Supplementary Information (not retyped here). System (atoms, supercell): WS\(_2\) armchair and zigzag edge pairs (both in the text/SI figures) on c-axis (0001) α-Al\(_2\)O\(_3\); W/S/H ReaxFF (trained in the W/S/H line cited (Ostadhossein, Yilmaz-line MoS\(_2\)-style, 2D Mater.) is merged with ReaxFF Al/O/C/H (Hong) so the sapphire + WS\(_2\) reactive interface is treated (main text paragraph). Bound / PBC: 3D PBC (standard slab-in-box for imaged supercells; lateral dimensions + gap geometry in Figs. 2–3). Reaction promotion: Targeted bond restraints (four in the 1.6 nm illustration: S-W (precursor)–W (edge) + S(–SH)-to-Al on (0001); rationale in the narrative) accelerate bond-formation; restraint energy is excluded from “barrier” analysis (as stated). Wide-channel run (≈1.6 nm): path (i)–(iv) (precursor to substrate (ii–iii), then row growth (iv), Figs. 2 a–b), with largest cited hurdle on second –SH removal (≈75 kcal mol\(^{-1}\), figure). Narrow-channel (0.2 nm): high (> 100 kcal mol\(^{-1}\)) barrier in (iii), incomplete WS\(_2\) row incorporation (Figs. 3a–b), paralleling W-deficiency arrays in STEM. Temperature (thermostat / target K in MD and CVD): N/A in this short note; see the 2D Mater. article and SI for CVD (K) and any NVT setpoints. Barostat: N/A in the restraint-driven kinetic sampling as excerpted; pressure: N/A; isotropic hydrostatic (NPT) is not a stated end-to-end protocol in the excerpted main text. Electric field: N/A. Replica/umbrella/metadynamics: N/A; restraints (not metadynamics) are used. Cumulative ps and 0.25 fs-type control (if any): N/A — see Supplementary Information; MOCVD (W(CO)\(_6\), H\(_2\)S, H\(_2\)) is laboratory context (Methods-level chemistry), not a full reactor ReaxFF** run.
2 — DFT/MD in this work. The kinetic arguments use ReaxFF reaction energies (main text), not VASP-style DFT PES in this letter; N/A — standalone ab initio (DFT) production in the 2D Mater. methods is not a separate “block” (check SI if they cite comparative **DFT)`.
3 — Force-field reparameterization: N/A (uses published W/S/ and merged oxide sets).
Microscopy (experiment). ADF-STEM (and TEM/STEM-class imaging) on monolayer CVD WS\(_2\); W-deficiency line and cluster morphologies (Figs. 1, 4), with CVD (MOCVD) growth on (0001) sapphire (article/SI for growth parameters).
Findings¶
- Imaging: long W-vacancy line defects and a minority of scattered vacancies; cluster-rich arrays also occur (Fig. 4), consistent with kinetic and morphological coalescence variability, not a single equilibrium point-defect picture.
- ReaxFF, wide (≈1.6 nm) channel: the (i)–(iv) path in Fig. 2 is exothermic by ~125 kcal mol\(^{-1}\) in the main-text trace, with (iii) ≈ 75 kcal mol\(^{-1}\) as the largest barrier (second –SH removal) and (i) ≈ 25 kcal mol\(^{-1}\) for the first –SH removal (Discussion).
- Narrow (0.2 nm) gap: a (> 100) kcal mol\(^{-1}\) barrier in the third step and incomplete precursor decomposition to a WS\(_2\) row-adding unit is argued to kinetically yield W-deficient line formation during late coalescence (Fig. 3, tied to the STEM “continuous array” case).
- Comparisons: the paper contrasts catalysis of –SH loss on sapphire (accessible in wide gaps) with H-abstraction off a facing edge in tight gaps, explaining the sensitivity of defect type to edge–edge separation and substrate access (armchair and zigzag edges treated, with zigzag in SI).
Corpus honesty: package-level ReaxFF control (timestep, thermostat) is not transcribed to this page; use the local PDF and SI for barrier curve digitization and figure numbering.
Limitations¶
ReaxFF accuracy for TMD–oxide interfaces is finite; gas-phase precursor chemistry and reactor flow are simplified in MD.
Relevance to group¶
van Duin-group collaboration on 2D TMD defects with ReaxFF growth modeling tied to electron microscopy.