AI-driven Simulation-Scale Bridging and Experimental Data Interpretation
The AI capabilities at the Center for Nanoscale Materials (CNM) tackle the challenges of “Predicting Transformations Across Scales” and “Extracting Rules from Multimodal Data.” To address the first challenge, Argonne researchers are developing a software called Bridging Length-scales via Atomistic Simulation Toolkit (BLAST). This high-performance, user-friendly tool allows users to create their own interatomic models by providing a simplified machine learning framework to generate training data sets, optimize potential functions and cross-validate model predictions. To address the second challenge, Argonne researchers are developing Fully-Automated Nanoscale to Atomistic Structure from Theory and eXperiment (FANTASTX), which combines atomistic-level modeling with machine learning, especially pattern recognition and multi-objective supervised learning. The result will be a powerful machinery that allows real-time interpretation of materials properties and processes from experiments performed at the CNM and elsewhere.