UniMatSim Dataset for High-Throughput Screening of Two-Dimensional Lieb-Lattice Materials

This dataset accompanies the study of UniMatSim, a unified framework for high-throughput materials simulations using universal machine learning interatomic potentials (UMLIPs). The dataset contains the initial crystal structures, high-throughput screening results, and subsequent density functional theory (DFT) calculation data for two-dimensional Lieb-lattice materials. The dataset is organized into the following main components: Initial candidate structures: 1,176 candidate compositions generated through elemental substitution of a two-dimensional Lieb-lattice template, provided in CIF and VASP formats. Uni2D screening results: Structural optimization, elastic stability, and phonon screening data from four independently trained Uni2D models. The strict intersection of the four screening sets contains 393 consensus candidates. DFT magnetic calculations: Magnetic energy comparisons and local magnetic moment data for 62 candidate materials. Electronic band structures: Spin-resolved band-structure data for 59 selected candidate materials. Phonon calculations: Phonon dispersion and density-of-states data for the representative material TaCo2Se4.