(642b) AI-Enhanced Multiscale Design of Liquid Crystals

Liquid crystals (LCs) are an intermediate state of matter that combine properties of crystalline solids and isotropic liquids. Their macroscopic behavior emerges from molecular interactions, resulting in elastic and optical properties that amplify molecular events up to the visible range with speed and precision. While LCs are widely known for their role in display technologies, their relevance extends to biological systems, chemoresponsive sensors, and organic semiconductors. To design advanced LC-based systems, it is essential to model their macroscopic behavior while retaining critical molecular features. Traditional continuum approaches often fall short when applied to complex LC chemistries. This work focuses on integrating coarse-grained and continuum simulations with machine learning techniques. At smaller scales, coarse-grained models capture molecular-level features and behaviors, and are informed by molecular simulation data and ML-based parameterization strategies. In the continuum regime, we use a tensorial order parameter and Ginzburg-Landau relaxation to obtain equilibrium configurations at the micron scale. Polarized optical microscopy images of these configurations are simulated using the Ondris-Crawford method and validated against experiments. Active learning is used separately to automate feature detection in microscopy images and relate these features to relevant molecular parameters. This multiscale, data-informed framework enables inverse design and targeted characterization of complex LC systems.