Seminars

Mechanical metamaterials derive their behavior from architecture (geometry) rather than chemical composition, opening a route to designing mechanical properties directly. With additive manufacturing (3D printing) now reaching a critical point where almost any intricate 3D geometry can be fabricated with ease, the central challenge is inverse design: finding geometies that realize targeted responses. I will present our recent progress on the inverse design of mechanical properties across increasing levels of complexity—from linear elasticity and full stress–strain curves to nonlinear dynamics and fracture. We are developing universal tools to facilitate the inverse design process for general metamaterials, including general graph representations and machine-learning–based, physics-guided inverse design methods.
Beyond passive properties (stiffness, strength, toughness), I will show how increasing architectural complexity eventually enables intelligence. This capability—often termed physical intelligence in robotic materials—is a property of the “body” that can share the computational burden with the “brain.” We propose a systematic way to train physical intelligence by coupling differentiable simulation with neural controllers: gradients flow from the “brain” to sculpt the “body,” yielding metamaterials that sense, actuate, compute, remember, and adapt. This body–brain co-design improves robustness and efficiency, pointing toward deployable meta-machines that perform reliably in unpredictable environments.