Machine Learning-Assisted Ultraelastic and Vibration-Resolvable Microwebs

Bioinspired structural designs have introduced a new paradigm in material science and mechanical engineering. Among them, the emerging spiderweb-inspired structures have shown potential for creating artificial microstructures with enhanced tunability and functionality. However, the restricted structural elasticity of current spiderweb-like designs causes limited mechanical performances, especially at the micro/nanoscale. Here, we employ machine learning and kirigami micro/nanofabrication to develop an ultraelastic microweb. Data-driven optimizations enable efficient transformation of the natural configuration with limited elasticity into an artificial design with ultrahigh elasticity, achieving a remarkably low stiffness of ∼0.188 nN/nm. Both mechanical simulations and experimental characterizations confirm the superior mechanical properties of the optimized microweb, conclusively validating the optimization model with the combination of genetic algorithm and deep learning. Further dynamic vibration analyses reveal ultrasensitive low-frequency mechanical resonances of the microweb, benefited from the greatly enhanced structural elasticity. For proof-of-concept demonstrations, the mass sensing of micro-objects with a high sensitivity of −0.801 kHz/pg and diversified vibration-resolvable information encryption are realized, respectively. This work establishes a generalizable strategy for creating highly elastic microstructures, with broad implications in the areas of mechanical micro-/nano-sensors, micro-/nano-electro-mechanical systems (MEMS/NEMS), mechanical metamaterials, biological manipulations, information encryptions, and beyond.