| 초록 |
Recently, a combination of the density-functional theory calculation and the automation technique provides a high-throughput screening for a catalysis design. With this method, we have recently developed novel metallic catalysts, which has been collaborated with experiments. On the other hand, a machine-learning has been rapidly penetrating the catalyst design. We have recently developed a machine-learning model (slab graph convolutional neural network, SGCNN) that can rapidly predict adsorption energies at accuracy level of DFT from the catalyst surface-adsorbate structures. Using the SGCNN, we can not only design various catalysts for nitrogen reduction reaction (NRR) and oxygen reduction reaction (ORR), but also predict Pourbaix diagrams of nanoparticle catalysts to evaluate their thermodynamic stability under electrochemical environments. And, in the last part of this work, I will discuss our machine-learning model for inverse design of materials. |
