Skip to content

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

Abstract

NEMD with ReaxFF grain-boundary thermal conductances feeds finite-element models of polycrystalline MoS₂ (and validation on graphene/h-BN) to report effective in-plane thermal conductivity vs grain statistics.

Summary

This uncorrected proof matches the Applied Materials Today article on multiscale thermal transport in polycrystalline MoS₂. Twenty MoS₂ grain-boundary models (5–7, 4–4, 4–6, 4–8, 6–8 ring motifs from literature TEM/DFT references) span a range of defect concentrations. Reactive NEMD supplies thermal conductance across boundaries and pristine reference values; results upscaled through finite-element continuum meshes of polycrystalline films. The workflow is cross-checked against graphene and h-BN polycrystals (comparison to equilibrium MD noted in the discussion). Adri C. T. van Duin is a co-author on the ReaxFF side. The study is motivated by thermal management in 2D electronics: in-plane κ of TMD films can fall sharply when grain sizes approach device dimensions, so multiscale coupling is presented as a way to connect DFT/ReaxFF interface conductances to device-scale temperature maps.

Methods

Force-field training / fitting. ReaxFF for MoS₂ is used as published (prior mechanical validation cited in the article); this contribution does not introduce a new QM refit.

MD application (atomistic dynamics). Engine / code: LAMMPS with ReaxFF drives nonequilibrium molecular dynamics (NEMD) on ~20 MoS₂ grain-boundary supercells (see Fig. 1 inventory: 4–4, 4–8, 5–7, 4–6, 6–8 motifs, symmetric/asymmetric variants) plus pristine single-layer reference cells. Boundaries / periodicity: In-plane periodic models along the grain-boundary direction. Timestep: Δt = 0.25 fs. Protocol / ensembles: initial room-temperature (~300 K) equilibration; end atoms fixed; further NVT equilibration with a Nose–Hoover thermostat; production NEMD partitions the cell into 22 slabs with hot (310 K) and cold (290 K) NVT-controlled end slabs (ΔT = 20 K) and NVE integration on the interior 20 slabs to reach steady-state heat flux \(J_x\) and extract κ via Fourier’s law (Eq. (2)) for pristine films, and grain-boundary conductance from the temperature drop at the interface (Eq. (3)). Barostat / hydrostatic pressure: N/A — no global pressure servo; the NEMD geometry instead establishes a stress/flux response consistent with fixed-end thermal biasing. Thermostat summary: Nose–Hoover during NVT equilibration; NVT hot/cold slabs during production NEMD; NVE in the interior slabs. Electric field: N/A — not applied. Enhanced sampling: N/A — direct NEMD. Duration: N/A — explicit ns totals are tied to prior protocol [66]/SI—confirm numerics on the VOR PDF.

Static QM / DFT. Literature TEM/DFT informs GB construction; DFT is not the thermal-transport calculator here.

Review / non-simulation framing. FE continuum models (Neper tessellations, ~10⁴ grains) upscale atomistic conductances to device-scale effective κ maps; graphene and h-BN benchmarks compare NEMD+FE vs EMD. Corpus note: this slug tracks an uncorrected proof PDF; use [[2017mortazavi-applied-mate-strong-thermal]] for the version-of-record bytes and pagination.

Findings

Outcomes & mechanisms. Reactive NEMD yields thermal conductances for diverse MoS₂ grain boundaries and pristine references; feeding those into FE meshes reproduces how polycrystalline microstructure suppresses effective in-plane κ, with especially strong reductions when grain sizes fall below ~100 nm in their models.

Comparisons. Graphene and h-BN polycrystal calculations contextualize the multiscale approach against fully atomistic EMD baselines as reported.

Sensitivity & design levers. Non-uniform grain-size distributions reroute heat flux compared with uniform approximations at similar averages—texture matters.

Limitations & outlook (as authored). Phonon-limited classical picture; electronic thermal transport omitted; NEMD vs EMD can shift absolute κ—see article discussion.

Corpus / KB honesty. Proof PDF; cite [[2017mortazavi-applied-mate-strong-thermal]] for authoritative tables after corpus alignment.

Limitations

Proof PDF; absolute κ can differ between EMD and NEMD implementations as discussed; primarily relative trends for polycrystallinity. Phonon isotope effects, substrate coupling, and electron–phonon transport in MoS₂ are outside the classical ReaxFF NEMD scope used here. Interlayer thermal resistance in few-layer films is another degree of freedom not exhaustively explored in the proof extract.

Relevance to group

Group co-authorship; couples ReaxFF transport data to continuum device-scale predictions.

Citations and evidence anchors