Skip to content

Strong thermal transport along polycrystalline transition metal dichalcogenides revealed by multiscale modeling for MoS₂

Abstract

Same study as the proof ingest: ReaxFF NEMD thermal conductances for MoS₂ grain boundaries feed FE models of polycrystalline films; validates multiscale route against graphene/h-BN.

Summary

Thermal management in two-dimensional transition metal dichalcogenides depends strongly on microstructure: grain boundaries and defects scatter phonons and reduce in-plane thermal conductivity relative to pristine single crystals. This Applied Materials Today article develops a multiscale workflow in which atomically resolved thermal conductances from nonequilibrium molecular dynamics (NEMD) with a ReaxFF parametrization for MoS₂ feed into finite-element continuum models of polycrystalline films reminiscent of chemical-vapor-deposited microstructures. The authors construct on the order of twenty representative grain-boundary configurations informed by microscopy and density-functional-theory literature, extract thermal boundary resistances or conductances as needed for continuum meshes, and then solve heat transport on statistically structured polycrystals. Cross-material benchmarking against graphene and hexagonal boron nitride polycrystals, including discussion of equilibrium versus nonequilibrium estimators for thermal conductivity, positions the MoS₂ results within a broader two-dimensional thermal-transport context. This wiki slug attaches to the version-of-record PDF filename Mortavazi_AMT_MoS2_thermal_2017.pdf; a parallel ingest slug [[2017mortavazi-venue-paper]] may exist for alternate corpus bytes of the same study.

Methods

Force-field training / fitting. ReaxFF for MoS₂ is used as published (prior mechanical validation cited in the article); this Appl. Mater. Today paper does not report a new QM refit of the reactive potential.

MD application (atomistic dynamics). Engine / code: LAMMPS carries nonequilibrium molecular dynamics (NEMD) with the ReaxFF bond-order energy expression (bond/valence/torsion/over- and under-coordination/lone-pair terms plus vdW and Coulomb contributions, Eq. (1) in the paper). System size & composition: Twenty distinct MoS₂ grain-boundary supercells (armchair/zigzag families with 4–4, 4–8, 5–7, 4–6, 6–8 cores, including symmetric and asymmetric variants where applicable) plus pristine single-layer MoS₂ reference cells for bulk κ extraction—see Fig. 1 and accompanying text for the structural inventory. Boundaries / periodicity: In-plane periodic supercells along the grain-boundary line (as stated for the constructed GB models). Timestep: Δt = 0.25 fs. Protocol / ensembles: structures are first equilibrated at room temperature (~300 K); end atoms are fixed, followed by additional NVT equilibration with a Nose–Hoover thermostat. For the NEMD production stage, the cell is partitioned into 22 slabs along the transport direction: the first and last slabs are held at 310 K and 290 K, respectively (ΔT = 20 K between hot/cold slabs), using NVT control on those slabs, while the interior 20 slabs evolve in the microcanonical (NVE) ensemble to establish a steady-state heat flux \(J_x\) and a linear temperature profile in the pristine case (Fourier-law extraction, Eq. (2)). Barostat / global hydrostatic pressure: N/A — no Parrinello–Rahman pressure servo; NVT hot/cold reservoirs impose the thermal bias while interior slabs remain NVE. Thermostat summary: Nose–Hoover in NVT equilibration; NVT thermostats on hot/cold slabs during NEMD production. Electric field: N/A — not applied. Replica / enhanced sampling: N/A — direct NEMD rather than umbrella or replica-exchange methods. Duration: N/A — total trajectory lengths beyond the steady-state criterion are tied to prior protocols [66] and are not re-tabulated numerically on this wiki page—use the article/SI for exact run lengths.

Static QM / DFT. Literature DFT/TEM motivates GB geometries and defect topologies; DFT is not rerun here as the κ engine.

Review / non-simulation framing. Finite-element (FE) continuum models (2D tessellated polycrystals with Neper, ~10⁴ grains, ~60 triangular elements per grain on average) consume the NEMD grain-boundary conductances to predict effective in-plane κ vs grain statistics; additional graphene and h-BN multiscale benchmarks compare NEMD+FE against fully atomistic EMD references as described in the article. Corpus note: this slug uses the version-of-record PDF Mortavazi_AMT_MoS2_thermal_2017.pdf; [[2017mortavazi-venue-paper]] tracks an uncorrected proof duplicate of the same study.

Findings

Outcomes & mechanisms. Reactive NEMD supplies thermal conductance values for ~20 representative MoS₂ grain boundaries plus pristine SL MoS₂; inserting those conductances into FE meshes yields effective in-plane thermal conductivity maps for polycrystalline films. Effective κ drops sharply when grain sizes approach ~100 nm and smaller because boundary scattering density increases.

Comparisons. The workflow is cross-checked on polycrystalline graphene and h-BN, comparing the multiscale (NEMD+FE) route to fully atomistic EMD effective conductivities as reported in the manuscript.

Sensitivity & design levers. Broadened grain-size distributions alter intrafilm heat-flux patterns relative to uniform microstructure approximations at comparable average grain sizes; non-uniform texture therefore matters for thermal routing.

Limitations & outlook (as authored). The model emphasizes phonon-dominated transport in the insulating regime and does not treat electronic thermal conduction or hot carriers; absolute κ values can depend on NEMD vs EMD estimators—follow the article’s method discussion.

Corpus / KB honesty. Numeric κ and conductance tables should be read from the DOI-linked article/PDF; this page summarizes the protocol narrative for navigation.

Limitations

The model focuses on phonon-dominated thermal transport in the insulating regime and does not couple to electronic thermal conduction or hot-carrier effects. Absolute thermal conductivity numbers can differ between EMD and NEMD protocols; readers should follow the article’s discussion of method dependence rather than treating any single κ value as universal.

Relevance to group

Adri C. T. van Duin appears among the coauthors; the work is a concrete example of ReaxFF-derived thermal properties feeding engineering-scale continuum predictions for two-dimensional materials.

Citations and evidence anchors

  • DOI: 10.1016/j.apmt.2017.02.005

Reader notes (navigation)