The ADP dataset consists of over 200,000 experimental crystal structures curated from the Cambridge Structural Database (CSD). It focuses on Anisotropic Displacement Parameters (ADPs), which describe atomic thermal vibrations within crystal lattices. ADPs provide insights into material properties such as thermal motion, heat capacity, vibrational entropy, and thermal expansion.

The dataset was curated with strict filtering criteria: it includes only non-polymeric crystal structures with 3D atomic coordinates and anisotropic thermal displacements for all non-hydrogen atoms. Structures with high disorder, missing temperature data, or poor-quality diffraction data (e.g., R-factor > 5%) were excluded. It covers a wide temperature range (2K to 573K), but most data points fall within the 100K–300K range, which is common in crystallographic experiments.

It helps reduce computational costs compared to traditional Density Functional Theory (DFT) calculations, which are computationally expensive. The dataset’s diversity in elemental composition and thermal conditions makes it valuable for advancing machine learning models in crystallography and materials science.s