A computational study proposes that pressure-induced metal–insulator transitions in ordered crystals fall into three broad classes: ionic, covalent and metavalent. The distinction is based not only on whether a band gap closes, but also on accompanying changes in atomic arrangement, lattice vibrations, electron–lattice coupling and electron sharing between neighboring atoms.
Three routes to becoming metallic
That produces three different routes to a metallic state. In ionic systems, the band gap closes continuously while the atomic arrangement and electron–lattice coupling change little. Covalent systems instead make a first-order transition to a higher-coordination metallic phase, meaning one with more neighboring atoms. Metavalent systems show continuous atomic rearrangement, complete lattice softening, a divergent mode-specific Grüneisen parameter and simultaneous changes in electronic structure, lattice dynamics and electron–lattice coupling.
NaCl, GeTe and GaAs serve as representative examples of the ionic, metavalent and covalent classes, respectively. The analysis also uses a one-dimensional hydrogen chain to reproduce the metavalent property pattern. The chain is treated as a toy model of the proposed bonding mechanism, offering a simplified picture rather than a material-by-material test.
The pressure paths are not alike
NaCl provides the clearest numerical illustration of the ionic route. In the simulation, its structural transition occurs at 27.9 gigapascals, while metallization occurs at 573 gigapascals. The separation is striking: in this model, the structural change comes long before the electronic transition.
GeTe supplies a contrasting example. Near its metal–insulator transition, the mode-specific Grüneisen parameter, γLO, rises above 50. This parameter tracks the response of a particular lattice vibration as the material is compressed; the reported surge is the numerical example behind the paper’s description of divergent anharmonicity—a strongly non-linear lattice response—near a metavalent transition.
The bonding map adds a second way to separate the routes. It uses ES, a shared-electron value for adjacent atoms: ionic solids become metallic at values much smaller than 1, whereas covalent and metavalent transitions proceed near the border between metavalent and metallic bonding. The authors present ES as a model-based descriptor of the pathway, not as a direct experimental measurement.
Following electrons and vibrations
To build the comparison, the study followed pressure or cell-volume series in calculations. It used density-functional theory, or DFT, with PAW pseudopotentials and the PBE-GGA approach to calculate energies and band gaps. Density-functional perturbation theory, or DFPT, supplied phonons, dielectric tensors and Born effective charges for the lattice-dynamics and dielectric-response analysis.
For the bonding analysis, ground-state wavefunctions were obtained with ABINIT and examined with DGrid using quantum theory of atoms in molecules, or QTAIM. The resulting quantities describe electron localization, delocalization and transfer between atoms. Additional calculations tested whether structure-specific relaxation procedures introduced systematic differences in structural evolution.
A framework still tied to the model
The abstract also associates metavalent behavior with Peierls distortions—lattice distortions on the insulating side—superconductivity on the metallic side and low lattice thermal conductivity near the transition. These are presented as associated features; the supplied analysis does not provide direct numerical analyses of superconductivity or lattice thermal conductivity.
Supplementary calculations report first-order transition pressures of 14.55 GPa for GaAs, 9.37 GPa for Si and 8.34 GPa for Ge. These values add concrete pressure markers to the covalent examples, alongside the broader three-way classification.
One important caveat is that the PBE-GGA approach is acknowledged to underestimate absolute band gaps, particularly for Ge. More broadly, the classification is based on the systems modeled and discussed, so it is best read as a framework for comparison rather than a universal rule for every metal–insulator transition.
Paper data and sources
Original title: Classification of Metal - Insulator Transitions: Relating characteristic Properties to Quantum Chemical Bonding Descriptors
Authors: Tim Bartsch, Carl-Friedrich Schön, Dasol Kim et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-25
DOI: Not available
Original paper · Full text