Symmetry and Prediction: Bridging DFT and AI to Forge New Frontiers in Low-Dimensional Structures
Prof. Aldo Humberto Romero
Director Research Computing
Physics and Astronomy Department
West Virginia University
In this presentation, I will unveil an innovative strategy that combines a thorough, symmetry-based search with the cutting-edge capabilities of graph neural networks to pioneer materials' structural search and design. My discourse will methodically detail our approach, emphasizing the distinctive geometries we have uncovered for one-dimensional and two-dimensional materials. I will also introduce newly identified crystal families with binary and ternary compositions, marking a significant advancement in the field. Beyond the introduction, I will examine graph neural networks in depth, elucidating their critical role and burgeoning significance within crystallography. As I illustrate the broad applicability of this method in materials science, I also consider its implications for other domains, such as developing universal potentials applicable to low-dimensional systems. To support the robustness of this method, I highlight two distinct families that our team has discovered, each showing fascinating excitonic properties. Moreover, my talk will venture into an uncharted scientific territory, showcasing previously undisclosed results. I aim to demonstrate how leveraging existing databases can forge powerful synergies between theorists and experimentalists, guiding them toward discovering compelling and novel materials. Through this exploration, we hope to inspire a collaborative frontier for innovation in the material sciences.
Hosted by Prof. Ghimire