Bayesian Optimization of Carbon Nanolattices Machine Learning designs new nanolattice geometries with the strength of carbon steel, but the density of Styrofoam, offering record strength-to-weight of lightweight materials. By implementing multi-objective Bayesian optimization in combination with two-photon polymerization and pyrolysis, these ultrahigh specific strength carbon nanolattices more than double the performance of benchmark materials. More details can be found in article number 2410651 by Peter Serles, Tobin Filleter, Seunghwa Ryu, and co-workers.

Nanoarchitected materials are at the frontier of metamaterial design and have set the benchmark for mechanical performance in several contemporary applications. However, traditional nanoarchitected designs with conventional topologies exhibit poor stress distributions and induce premature nodal failure. Here, using multi-objective Bayesian optimization and two-photon polymerization, optimized carbon nanolattices with an exceptional specific strength of 2.03 MPa m³ kg^-1 at low densities <215 kg m^-3 are created. Generative design optimization provides experimental improvements in strength and Young's modulus by as much as 118% and 68%, respectively, at equivalent densities with entirely different lattice failure responses. Additionally, the reduction of nanolattice strut diameters to 300 nm produces a unique high-strength carbon with a pyrolysis-induced atomic gradient of 94% sp² aromatic carbon and low oxygen impurities. Using multi-focus multi-photon polymerization, a millimeter-scalable metamaterial consisting of 18.75 million lattice cells with nanometer dimensions is demonstrated. Combining Bayesian optimized designs and nanoarchitected pyrolyzed carbon, the optimal nanostructures exhibit the strength of carbon steel at the density of Styrofoam offering unparalleled capabilities in light-weighting, fuel reduction, and contemporary design applications.