PARCE: Accurate Inference of Crystal Structure Information From Phonon Vibrational Modes

ABSTRACT An accurate description of crystal structures is a prerequisite for experimental synthesis and understanding of the physicochemical properties of materials. However, conventional X‐ray diffraction (XRD) often requires stringent experimental conditions and offers limited access to certain types of data, highlighting the need for complementary optical measurement techniques. In this study, we developed a multidescriptor framework by integrating key crystallographic descriptors, including space groups, Pearson symbols, Wyckoff sequences, and lattice‐geometry descriptors, to classify over 30,000 crystals into physically interpretable structure clusters. Subsequently, an accuracy‐adaptive ensemble network based on residual architectures was implemented to capture structural “fingerprints” within phonon vibration modes and Raman vibration modes, which achieves a classification accuracy exceeding 60% for cluster‐resolved structural information inferred from the computed phonon data.