30 September 2023 in npj Computational Materials
Structural search and feature extraction are a central subject in modern materials design, the efficiency of which is currently limited, but can be potentially boosted by machine learning (ML). Here, we develop an ML-based prediction-analysis framework, which includes a symmetry-based combinatorial crystal optimization program (SCCOP) and a feature additive attribution model, to significantly reduce computational costs and to extract property-related structural features. Our method is highly accurate and predictive, and extracts structural features from desired structures to guide materials design. We first test SCCOP on 35 typical compounds to demonstrate its generality. As a case study, we apply our approach to a two-dimensional B-C-N system, which identifies 28 previously undiscovered stable structures out of 82 compositions; our analysis further establishes the structural features that contribute most to energy and bandgap. Compared to conventional approaches, SCCOP is about 10 times faster while maintaining a comparable accuracy. Our framework is generally applicable to all types of systems for precise and efficient structural search, providing insights into the relationship between ML-extracted structural features and physical properties.
24 January 2023 in Physical Review B
Carbon biphenylene has stimulated substantial research because of extraordinary properties introduced by the metallic character, e.g., the ultrahigh electron thermal transport. Here, inspired by the synthesis of carbon biphenylene [Fan et al., Science 372, 852 (2021)], we identify the stability of biphenylene as and semiconductors with four-, six-, and eight-membered periodic rings of irregularly -hybridized atoms via structural searches. Unexpectedly, we confirm that biphenylene exhibits a peculiar funnel-shaped band structure, which is a direct consequence of the delocalization/localization of bonds formed by B-, N-, or C- electrons. The band structure greatly improves the thermoelectric performance by enhancing the power factor, although the lattice thermal conductivity is relatively large after including four-phonon scattering resistance because of low atomic masses. The similar behaviors are absent in biphenylene because of the localization of bonds formed by C- electrons, although with a stronger anharmonicity and thus a lower lattice thermal conductivity. The anomalous power factor can be explained by the constant approximation: the -type doping controls the carrier group velocities and thus realizes the tunability of tensor ratio . Our analysis suggests that the funnel-shaped electronic structure could be reproduced in two-dimensional semiconductor systems with the small electronegativity difference and the comparable stoichiometry. Our work realizes the thermoelectric improvement through controlling the shape of band structure, which provides insights for designing promising two-dimensional thermoelectric materials.
24 January 2024 in Physical Review B
Cu chalcopyrites exhibit excellent thermoelectric performance because of their high thermopower and low thermal conductivity. Conversely, despite that Ag chalcopyrites have even lower thermal conductivity than the Cu-based ones, they generally display lower thermoelectric performance. The underlying physics for the thermoelectric disparity between Cu- and Ag-based materials remains unclear. In this work we investigate thermal transport and thermopower of ternary chalcopyrites using first-principles methods. We reveal that strong anharmonicity from coupling leads to low thermal conductivity in Cu as well as Ag chalcopyrites, whereas weaker bond strength and heavier atomic mass contribute to even lower thermal conductivity in Ag chalcopyrites than the Cu ones. On the other hand, Cu chalcopyrites show superior thermopower owing to their higher band degeneracy and stronger coupling, which leads to enhanced mobility. We conclude that if Ag chalcopyrites could reach comparable hole concentrations as in Cu chalcopyrites, high thermoelectric performance in Ag chalcopyrites can also be achieved because of their low thermal conductivity. Understanding the disparity between Cu and Ag chalcogenides from a mechanistic perspective is important for future design of superior thermoelectric materials.
25 November 2024 in Physical Review Letters
Polymorphs commonly exist for various materials, enabling phase engineering for diverse material properties. While the crystal structures of different polymorphs can, in principle, be experimentally characterized, interpreting and understanding complex crystal structures can be very challenging. Using Ga_{2}O_{3} as a prototype, here we show that the crystal structure of γ-Ga_{2}O_{3} has long been misinterpreted from either theory or experiment. By lattice mapping based crystallographic analysis and combinatorial crystal structure search, we reveal the nature of γ-Ga_{2}O_{3} as a disordered form of β-Ga_{2}O_{3}, which is further confirmed by Monte Carlo simulations of the order-disorder transition. Our insights are critical for correct interpretation and characterization of γ-Ga_{2}O_{3}, and the established physics is expected to be widely applicable to a variety of polymorphic systems.
5 November 2021 in The Journal of Physical Chemistry Letters
22 May 2024 in Journal of the American Chemical Society