Jonggeol Na

CatDiT can generate diverse catalyst candidates that significantly outperform traditional methods, achieving a 1.5-fold enrichment in desired properties for nitrogen reduction.

CatRetriever connects slab-level generative models with bulk structures, achieving over 91% accuracy in retrieving parent bulk candidates.

Achieving 98% structural validity and a four-fold improvement in binding energy conditioning, this model revolutionizes how we design catalysts by directly generating structures tailored to desired properties.