Massive Discovery of Low-Dimensional Materials from Universal Computational Strategy

TL;DR

This work combined universal machine-learning interatomic potentials (UMLIPs) and an advanced, interatomic force constant (FC) -based dimensionality classification method to make a massive discovery of novel low-dimensional materials.

Abstract

Low-dimensional materials have attractive properties that drive intense efforts for novel materials discovery. However, experiments are tedious for systematic discovery, and present computational methods are often tuned to two-dimensional (2D) materials, overlooking other low-dimensional materials. Here, we combined universal machine-learning interatomic potentials (UMLIPs) and an advanced, interatomic force constant (FC) -based dimensionality classification method to make a massive discovery of novel low-dimensional materials. We first benchmarked UMLIPs' first-principles-level accuracy in quantifying FCs and calculated phonons for 35,689 materials from the Materials Project database. We then used the FC-based method for dimensionality classification to discover 9139 low-dimensional materials, including 1838 0D clusters, 1760 1D chains, 3057 2D sheets/layers, and 2484 mixed-dimensionality materials, all of which conventional geometric descriptors have not recognized. By calculating the binding energies for the discovered 2D materials, we also identified 887 sheets that could be easily or potentially exfoliated from their parent bulk structures.

Figures (4)

• Figure 1: Validation of methodology. Comparing DFT and ML (MatterSim) for: a) MaxFC and MinFC [Equation (\ref{['eq:minmax']})], b) distribution of MaxFC and MinFC, c) maximum phonon frequencies at $\Gamma$-point, and d) number of Phonon database structures in each dimensionality group. Panels (a) and (c) also show the related RMSE.
• Figure 2: High-throughput calculation workflow and resulting dataset statistics. (a) Workflow for screening and high-throughput calculations. Each step indicates the number of structures involved. (b) Number of low-dimensional materials discovered in the screened dataset by FCDimen. "Other" corresponds to mixed dimensionalities with more than two types of dimensionalities (012D, 013D, and 023D). (c) Number of low-dimensional materials classified using the $16$ most common space groups. (d) Heat map with occurrences of each element in low-dimensional materials in the dataset.
• Figure 3: Chemical structures of representative low-dimensional materials in the dataset (top and side views). Legends show dimensionalities, chemical formulas, and MP IDs.
• Figure 4: Discovered 2D materials in detail. (a) Comparison of the binding energies of some known 2D materials, calculated using ML (with D3–BJ damping) and DFT (DFT-D3, DF2‑C09, and rVV10). DFT-calculated binding energies using DF2‑C09 and rVV10 are taken from Ref. Mounet2018. (b) Number of 2D structures as a function of exfoliation energies using ML D3-BJ. Structures are grouped into easily exfoliable ($E_b \leq35$ meV/Å$^2$), potentially exfoliable ($35<E_b<125$ meV/Å$^2$), and strongly bound ($E_b>125$ meV/Å$^2$) 2D materials. (c) Correlation between binding energies and MinFC for easily and potentially exfoliable 2D materials. (d) Easily and potentially exfoliable 2D materials categorized into $15$ most common space groups. (e) Chemical structures (top and side views) of selected easily exfoliable 2D materials from the most common species (binary, ternary, and quaternary). Legends indicate space groups, chemical formulas, MP IDs, and binding energies ($E_b$ in meV/Å$^2$).