Direct visualization of out-of-equilibrium structural transformations in atomically thin chalcogenides

The study investigates how equilibrium versus non-equilibrium thermal diffusion conditions affect structural phase transformations in atomically thin MoS2. TMDCs such as MoS2 can exist in multiple phases (2H, 1T, 1T′, 2H′, 3R), with 2H being thermodynamically stable and 1T/3R metastable. Prior studies have largely examined amorphous-to-crystalline kinetics, point defect dynamics, and dislocation kinetics under equilibrium conditions, leaving open questions about nucleation and stabilization of metastable phases. Here, the authors hypothesize that changing the heating rate of bilayer/few-layer MoS2 can effectively tune sulfur concentration via evaporation/redeposition dynamics, thereby altering the phase and morphology post-synthesis. They relate observations to the Mo–S bulk phase diagram and known sulfur vacancy energetics, and leverage in situ, real-time aberration-corrected STEM to directly visualize atomic-scale transformations under fast (non-equilibrium) and slow (equilibrium) heating.

• TMDCs exhibit multiple structural phases; 2H is stable, 1T and 3R are metastable and can revert to 2H upon energy input (refs 6–10).
• In situ microscopy studies have focused on point-defect kinetics and structural transformations under equilibrium or beam-induced conditions in TMDCs (refs 4–5, 16–27), with limited insight into metastable phase nucleation mechanisms.
• Prior observations include heterojunction formation and defect-driven atomic migration/grain boundary evolution (refs 28–30).
• Sulfur vacancies dominate in few-layer MoS2 due to low formation energies; annealing at 500–1000 °C causes sulfur loss on ~30 min timescales (refs 34–35).
• Melting point depression in nanoscale materials and effects of stacking (2H′, 3R) are known (refs 36–37, 6–9). Remaining gaps include how processing (heating rate, environment) maps to phase evolution and metastable phase formation in confined dimensions.

• Materials: Mechanically exfoliated monolayer to few-layer MoS2 prepared via Scotch tape method. Transferred onto MEMS heater-embedded TEM chips (Hummingbird Scientific) using dry PDMS-assisted transfer. Post-transfer cleaning: 300 °C, 2 h in Ar+H2 in quartz tube furnace.
• In situ non-equilibrium heating: Microfabricated MEMS heater inside TEM, heating rate up to 200 °C/s; used 25 °C/s ramps (e.g., from 500 to 650–700 °C). Temperature calibration from manufacturer and verified by Raman thermometry (Si particles). High vacuum in TEM column (~10−10 mbar order noted; text indicates high vacuum ~10 mbar typo—relative higher vacuum than ex situ).
• Ex situ equilibrium heating: Quartz tube furnace (rough vacuum ~10 mbar) and ultra-high purity Ar (99.995%) environments. Heating at 25 °C/min to target temperatures (e.g., 525–550 °C) with dwell times (5–10 min) to study time dependence.
• Microscopy: JEOL F200 S/TEM and probe-corrected JEOL NEOARM STEM at 200 kV. JEOL NEOARM: condenser aperture 40 μm, camera length 4 cm, probe current 120 pA. Imaging on Gatan OneView IS and Ultrascan cameras; ADF/HAADF-STEM detectors used.
• In situ imaging: BF-TEM time series at 650 °C using Gatan OneView IS, 50 fps, 2k×2k; sequences captured coalescence dynamics.
• Compositional analysis: EDS (dual detector) on F200.
• Image processing and simulations: Image smoothing (adaptive Gaussian blur in ImageJ). Image simulations using QSTEM with experimental parameters; combined with ImageSim for focal series reconstruction and QMB model builder.
• Island statistics: ADF-STEM images segmented via morphological filters and connected components (scikit-image). Perimeter and area measured, normalized to average perimeter (260 px) to compute perimeter/area shape factor; histogram fitted with inverse Gaussian distribution.
• Control experiments: Slow heating in vacuum vs UHP Ar to examine vapor pressure effects; assessment of beam effects via comparing irradiated vs non-irradiated regions (no significant differences post-heating).

• Heating-rate-dependent transformations:
  • Fast, non-equilibrium heating (25 °C/s to 650–700 °C): Continuous few-layer MoS2 disintegrates into highly crystalline, faceted nanoscopic islands (<20 nm; as small as ~5–10 nm) with hexagonal symmetry. Atomic-resolution HAADF-STEM and FFTs confirm crystallinity.
  • Slow, equilibrium heating (25 °C/min to 525–550 °C, 5–10 min): Formation of triangular etch pits and localized decomposition starting at edges/defects; progression to nanocrystalline regions (diffuse FFT spots) and fully amorphous MoS2; eventual Mo/MoSx nanocrystal formation (Mo-enriched EDX), indicating sulfur depletion and non-stoichiometric decomposition.
• Phase formation under fast heating:
  • Mixed 2H and 3R phases observed within nanocrystals; atomically sharp 2H/3R heterojunctions and transition regions identified via HAADF-STEM and QSTEM simulations (only 2H and 3R contrasts matched; no 1T/1T′ observed).
  • 3R phase occurs exclusively in fully/partially disintegrated regions; intact continuous regions remain 2H.
• Coalescence dynamics (650 °C hold): Rapid initial disintegration (<10 ms frame time), followed by coalescence over tens of seconds; example growth from multiple small crystallites to a ~10 nm hexagonal island over 55 s. Islands thicken over time (darker center contrast; reduced lattice fringe visibility) consistent with surface-diffusion-driven coarsening.
• Epitaxy/orientation:
  • Fully disintegrated regions on amorphous SiNx: islands randomly oriented ("quantum dot"-like).
  • Partially disintegrated regions: islands maintain perfect epitaxial alignment with underlying continuous MoS2.
• Shape statistics:
  • Automated analysis of 775 smallest nanostructures (90% of detected islands) shows shape factors predominantly between circle and regular hexagon, indicating rounded hexagonal morphology. Distribution fitted by inverse Gaussian with shape factor μ ≈ 2.1, consistent with near-stoichiometric conditions (mixed edge terminations).
• Mechanistic energetics and kinetics:
  • Under fast heating: S vacancies form quickly; S vacancy diffusion barrier ~0.8 eV vs S adatom diffusion barrier ~1.6 eV. Rapid temperature rise activates Mo adatoms (~700 °C) before substantial S desorption (S2, S8) occurs, enabling recombination and stoichiometric nucleation (CVD-like), producing vertically stacked layers with mixed 2H/3R stacking.
  • Under slow heating: Extended time at lower T allows volatile sulfur species to form and escape before Mo adatoms activate, driving the system into Mo-rich two-phase regimes (Mo + MoS2 or Mo + Mo2S3), producing amorphous/nanocrystalline MoSx and Mo nanocrystals.
• Environmental effects: Higher vapor pressure in Ar delays disintegration relative to vacuum, consistent with sulfur evaporation/re-deposition control.
• Consistency with bulk Mo–S phase diagram: Observed products align with predicted two-phase regions as sulfur content decreases at given temperatures (e.g., Mo2S3 and 3R at 500–1000 °C; Mo + Mo2S3 below ~60 at% S).

The results demonstrate that heating rate is a decisive control parameter for out-of-equilibrium structural evolution in atomically thin MoS2. Fast ramps conserve sulfur on the surface long enough for Mo adatoms to form and recombine, creating a local, stoichiometric, CVD-like environment that drives nucleation of highly crystalline, faceted nanostructures exhibiting mixed 2H/3R stacking and atomically sharp heterojunctions. The observation of 3R only in disintegrated regions suggests stabilization via rapid restacking and vertical mass transport during thermolysis. Epitaxial alignment in partially disintegrated regions indicates that residual continuous layers template island orientation; fully disintegrated regions lose this registry and exhibit random orientations. In contrast, slow heating allows sulfur to volatilize before Mo adatom formation, depleting S and pushing the composition into Mo-rich regimes consistent with the Mo–S phase diagram, leading to nanocrystalline/amorphous MoSx and Mo metal nanocrystals. Vapor pressure modulates these pathways, with inert Ar delaying sulfur loss versus vacuum. The mechanistic picture, supported by diffusion barriers (S vacancy ~0.8 eV; S adatom ~1.6 eV) and temperature-time profiles, explains the divergent morphologies and phases and addresses the central question of how non-equilibrium processing controls phase and nanostructure formation at the atomic scale.

High-rate in situ heating of atomically thin MoS2 induces spontaneous disintegration into highly crystalline, laterally confined nanostructures (<20 nm) with mixed 2H/3R phases and epitaxial alignment where underlying layers persist, while slow ex situ heating drives sulfur-depleted decomposition into nanocrystalline/amorphous MoSx and ultimately Mo nanocrystals. Direct, real-time STEM visualization links these outcomes to sulfur evaporation–redeposition kinetics and the Mo–S phase diagram. The work provides a top-down route to synthesize atomically thin, quantum-dot-like features, enabling studies and potential applications in optoelectronics, catalysis, lubrication, and phase-change memory. Future work could explore: tuning island size/phase via precise thermal profiles and environments; extending to other TMDCs; correlating structure with electronic/optical properties; and integrating these nanostructures into device architectures.

• Temporal resolution: The initial disintegration under fast heating occurs in less than a single 10 ms camera frame, limiting direct observation of the earliest transformation events at atomic resolution.
• Beam effects: Although control regions suggest minimal beam influence, high-voltage TEM/STEM could still contribute subtle effects not entirely excluded.
• Environmental/pressure differences: In situ high-vacuum conditions differ from ex situ furnace environments (vacuum vs UHP Ar), potentially affecting comparability; absolute temperatures for transitions vary with vapor pressure.
• Material scope: Experiments focus on exfoliated MoS2; generalization to other TMDCs and substrates requires further validation.
• Thickness range: Results emphasize bilayer/few-layer regimes; behavior may differ in monolayers or thicker films.