Atomic Structure-Free Representation of Active Motifs for Expedited Catalyst Discovery

• DOI: 10.1021/acs.jcim.1c00726
• Publication Date: 2021-08-23T11:51:33Z
• Authors: Dong Hyeon Mok; Seoin Back
• Keywords: Representation

To discover new catalysts using density functional theory (DFT) calculations, binding energies of reaction intermediates are considered as descriptors to predict catalytic activities. Recently, machine learning methods have been developed reduce the number computationally intensive DFT calculations for a high-throughput screening. These require several steps such bulk structure optimization, surface modeling, and active site identification, which could be time-consuming candidate materials increases. bypass these processes, in this work, we report an atomic structure-free representation motifs energies. We identify atoms their nearest neighboring positioned same layer sublayer, properties collected construct fingerprints. Our method enabled quicker training (200–400 s CPU) compared previous deep-learning models predicted CO H with mean absolute errors (MAEs) 0.120 0.105 eV, respectively. is also capable creating all possible without any predicting trained model. The energy distributions can suggest promising candidates accelerate catalyst discovery.