Data-driven calibration of atomic volumes for efficient prediction of material properties from chemical formula

Atomic volume, a cornerstone of modern chemistry and materials science, governs a vast array of material properties. However, a comprehensive and accurate calibration of this critical parameter has remained elusive. Herein, we establish a set of atomic volumes for most elements within the periodic table by developing a data-driven framework for compact crystalline solids. Our approach calibrates these atomic volumes via self-consistent optimization on a large dataset of compact crystalline solid structures. We demonstrate that these calibrated atomic volumes enable the efficient and accurate prediction of material properties, such as densities and bulk moduli of compact crystalline solids solely from their chemical formula. Finally, we provide heat maps of densities and bulk moduli, facilitating the inverse design of materials with targeted properties. This work delivers a foundational dataset critical to chemistry and materials science.