Bridging Crystal Structure and Material Properties via Bond-Centric Descriptors

Abstract

Although chemical bonding is the fundamental mechanistic bridge connecting atomic structure to macroscopic material properties, current data-driven materials science largely treats it as an implicit "black box". Existing machine learning (ML) models rely predominantly on geometric coordinates, forcing them to implicitly relearn complex quantum mechanics from scratch. This lack of intermediate physical features limits model interpretability and generalizability, particularly when training data is scarce. To solve this problem, we introduce MattKeyBond, a bond-centric materials database that explicitly maps the local electronic landscape and bonding interactions of materials. Building on this, we propose Bonding Attractivity (BA), a novel element-specific descriptor that quantifies the intrinsic capability of atoms to form covalent networks. By providing pre-calculated, energy-dimensional bonding descriptors, MattKeyBond transforms the implicit "black box" into physically interpretable features. This strategy relieves ML models from the burden of deducing physical laws from pure geometry, enabling accurate predictions even with limited data and seamlessly integrating electronic structure theory into modern AI workflows.

Figures (10)

• Figure 1: MattKeyBond: Enhancing implicit learning with interpretable electronic and bonding descriptors.
• Figure 2: The evolution of electronic structure from isolated atomic orbitals to hybridized molecular orbitals and finally to Bloch states in a periodic crystal. The reconstruction of the Reduced Density Matrix (RDM) highlights two distinct energetic pathways driving bond formation: charge transfer (represented by blue arrows), and orbital hybridization (represented by orange arrows). The last column lists a conceptual classification of chemical bonds based on these two primary dimensions.
• Figure 3: The materials screening criterions and high throughput calculation/analysis workflow.
• Figure 4: Detailed bonding analysis of the nearest C-C bond in graphene as a representative example. (a) Crystal structure of graphene with the target C$_1$-C$_2$ atomic pair highlighted. (b) The real-space inter-atomic Hamiltonian matrix and (c) RDM for the C$_1$-C$_2$ pair in the atomic orbital basis ($s, p_z, p_y, p_x$). (d) Comparison of the electronic band structures calculated by DFT (gray lines) and our constructed CWF (purple circles). (e) Bond-resolved COHP curves in energy space, clearly distinguishing the contributions and strengths from $\sigma$ (black), $\pi$ (red), $\pi'$ (blue), and $\sigma'$ (green). (f) Quantitative decomposition of the bond strength via the SVD of RDM.
• Figure 5: Periodic table of Bonding Attractivity. The three numbers in each element block are, respectively, the primitive bonding attractivity $\eta_A^{0}$, the characteristic decay length $L_A$, and the valence-state modulation factor $M_A$. The background color encodes $\eta_A^{0}$.