Machine-Assisted Design of Patchy Polygons for Mesoscale Assembly of Superlattices

Creating self-assembled materials with programmable structure and function demands understanding and control over how constituent colloidal building blocks assemble. Despite experimental advances, predictive design of such assemblies remains limited due to the complexity of interactions and the computational cost of accurate modeling. Moreover, the mapping between particle design and assembly outcome is highly complex, making inverse design of targeted structures a significant challenge. Here, we develop a coarse-grained patchy polygonal particle model inspired by DNA origami tiles to investigate their surface-confined assembly. By systematically varying particle shape and patch functionality, we reveal how local design rules govern emergent two-dimensional order. Using a machine-learning–based inverse design framework, we identify particle geometries and patch arrangements that produce desired lattice symmetries, such as Archimedean and Pythagorean tilings, and uncover design principles for improving assembly yield and enabling lattice reconfigurability. Our approach provides a data-driven pathway toward programmable and adaptive two-dimensional materials.