Designing ultra rare, multimodal mechanical metamaterials

Controlling both the structure and enumeration of zero modes in metamaterial design is difficult; designs that feature both the desired number and spatial structure of zero modes are typically very rare and thus hard to find without a design method. Here, we introduce a general design method that combines neural networks, a genetic algorithm, and cluster analysis to exploit previously unreachable parts of the design space and find ultra rare multimodal designs. In particular, we design metamaterials with a few spatially complex zero modes corresponding to a smiley face and a frowny face. Our method provides a crucial step towards systematic design of complex multifunctional materials.