Fig. 5. Inverse-designed truss metamaterials based on gradient optimization.

a Maximizing Young’s modulus E₂₂, b minimizing Poisson’s ratio ν₂₁, and c maximizing the bulk-to-shear modulus ratio K_V/G_V of truss lattices. Each example shows the property evolution vs. the number of optimization iteration steps, including a few selected structures at the indicated points, as well as the property path compared to the training dataset in the relevant property spaces (each dot represents a truss in the training data, which is color-coded by the (dimensionless) radius of the beams with circular cross-section). Source data are provided as a Source Data file.