Skip to content

Multiscale computational understanding and growth of 2D materials: a review

Review of atomistic (DFT, empirical and reactive MD including ReaxFF lines), mesoscale (e.g., phase-field), and continuum transport models for 2D material growth, plus machine-learning angles for structure–property mapping and discovery.

Summary

Since graphene’s isolation, 2D semiconductors and heterostructures have motivated a broad computational ecosystem. The abstract frames the review around three methodological tiers: (i) nanoscale atomistic simulations—DFT and MD with empirical and reactive potentials (explicitly naming reactive interatomic potentials); (ii) mesoscale methods such as phase-field models; and (iii) macroscale continuum approaches coupling thermal and chemical transport. It further discusses machine learning to connect structures and properties and to guide discovery of new 2D materials, closing with an outlook on computational support for 2D synthesis and growth. The introduction stresses that growth perfection and properties are exquisitely sensitive to processing, enumerating desired traits (mobility, bandgap tunability, flexibility, optical coupling) and noting top-down vs bottom-up synthesis routes (exfoliation vs CVD/ALD) with contrasting quality vs scalability trade-offs.

Methods

4 — Review / perspective (primary article type). The article is a narrative review in npj Computational Materials (DOI 10.1038/s41524-020-0280-2) that surveys 2D material growth and synthesis across length and time scales. It organizes the field into: (i) atomistic tools—DFT and molecular dynamics with empirical and reactive interatomic potentials (including ReaxFF-line models in cited work); (ii) mesoscale phase-field and related morphology models; and (iii) continuum thermal and chemical transport coupled to reactor-relevant heat/mass fluxes. It also treats machine learning combined with computation and experiment for structure–property links and new-material discovery, and includes an outlook on computation-guided 2D synthesis and growth after the graphene-era expansion of layered systems (per abstract and Introduction in the VOR at pdf_path). Reproducibility for any numerical value means opening the cited primary work—this review’s scope is curatorial and comparative, not a single laboratory protocol.

1 — MD application (as one protocol in this document). N/A — the article is bibliography-driven; it does not report one NVE/NVT/NPT molecular dynamics run with a single code, time step (fs), production (ns), and thermostat/barostat table for a single 2D growth benchmark.

2 — Force-field training. N/A — no de novo ReaxFF (or other) force-field fit is a result of this manuscript; such lines are cited from the literature.

3 — Static QM / DFT (a unified DFT study in this document). N/A — the DFT content is synthetic (e.g. GGA–PBE-class examples and limitations in the ATOMISTIC section); cutoffs, k-meshes, and structures belong to cited primaries and Table 1-style summaries in the review itself, not a one-off DFT project performed for this manuscript alone.

Findings

Outcomes, comparisons, and sensitivity. The review frames wafer-relevant outcomes—uniformity, defect burden, grain texture—as emerging from coupled heat and mass transport, nucleation, and surface reaction kinetics, so that useful multiscale models must link DFT/atomistic barriers to FEM- or continuum-class transport in realistic geometries. Machine learning is presented as a practical layer for exploring high-dimensional synthesis parameter spaces when paired with expensive DFT or reactive MD. Challenges called out in the text include substrate defects and wrinkling, van der Waals interactions across scales, kinetics specific to monolayer growth, and reproducing flexural (quadratic) phonon behavior with classical or reactive potentials. Comparisons to experiment and agreement between methods are citation-specific; no single number in this wiki should replace a cited primary study.

Limitations and outlook (as authored). A review selects exemplars; ranks of codes or potentials are not a universal table here. Corpus honesty: use the VOR file at pdf_path for figure and table numbering; the sibling proof PDF is listed under 2020momeni-npj-computat-multiscale-computational-2 if a local duplicate bytes need governance only.

Limitations

As a review, it selects exemplars rather than ranking all codes/potentials; readers must consult primary studies for quantitative performance. Adri van Duin co-authorship signals ReaxFF ecosystem ties but does not make every section ReaxFF-specific.

Relevance to group

Adri van Duin co-authorship ties the ReaxFF / reactive MD ecosystem to 2D growth modeling communities spanning Penn State and partner institutions—useful for cross-linking method pages with 2D materials theme hubs.

Reader notes (navigation)