Discovery of Novel Materials with Giant Dielectric Constants via First‐Principles Phonon Calculations and Machine Learning

We report a materials informatics framework that combines first‐principles phonon calculations and machine learning to discover perovskite oxides with giant dielectric constants. Initially, phonon dispersions were calculated for 61 experimentally reported perovskite compounds from the ICSD database to validate the effectiveness of hierarchical clustering in classifying phonon dispersion behaviors. The analysis was then extended to a broader set of 525 experimentally reported and hypothetical perovskite structures. Six promising candidates were identified, among which RbNbO 3 was synthesized under high‐pressure conditions. Its dielectric properties were validated: a single‐crystal RbNbO 3 exhibited a room‐temperature dielectric constant of 800–1000, consistent with theoretical predictions. These results demonstrate the power of soft‐mode‐based informatics, combined with comprehensive phonon calculations, in discovering high‐performance dielectric materials.