简体中文

Resolving Chemical-motif Similarity with Enhanced Atomic Structure Representations for Accurately Predicting Descriptors at Metallic Interfaces

Prof. Tao Wangs group at the Center of Artificial Photosynthesis for Solar Fuels at Westlake University (CAP for Solar Fuels @Westlake) recently published a research paper in Nature Communications, which is titled “Resolving chemical-motif similarity with enhanced atomic structure representations for accurately predicting descriptors at metallic interfaces.


This work has successfully tackled the challenge of resolving chemical-motif similarity in highly complex catalytic systems by integrating equivariant message-passing-enhanced atomic structure representation. The team developed an equivariant graph neural network (equivGNN) model, achieving mean absolute errors of <0.09 eV for different descriptors at metallic interfaces. The models high prediction accuracy and ease of implementation across various systems demonstrate its robustness and potentially broad applicability, laying a reasonable basis for accelerating catalyst design.


Assistant Researcher Dr. Cheng Cai from the CAP for Solar Fuels @Westlake is the first author, and Professor Tao Wang is the corresponding author.