Preprocessed data for Phonon Fine-tuning (PFT).
Abstract
Preprocessed MDR Phonon PBE dataset (https://alexandria.icams.rub.de/data/phonon_benchmark/) to be used for phonon fine-tuning, as introduced in "PFT: Phonon Fine-tuning for Machine Learning Interatomic Potentials" (https://arxiv.org/abs/2601.07742). The test set was constructed by combining the 91 materials with mp-ids that are not in MPtrj, and randomly selected materials for a total of 1000 calculations. The remaining calculations are split into train/validation with proportion 95/5. The training and validation files are preprocessed into .aselmbd files with the Hessian stored in atoms.info["hessian"].
Fields of Study & Subjects
Machine Learning in Materials Science
Advanced Chemical Physics Studies
Model Reduction and Neural Networks
Publication details
- Journal / source
- Figshare
- Publication date
- 1 January 2026
- Document type
- Dataset