Machine learning-guided hierarchically structured lotus-leaf-inspired polysiloxane composite films for amphiphobic self-cleaning radiative cooling.

Passive radiative cooling holds profound promise as a zero-energy and carbon-neutral thermal management technology. However, its real-world outdoor deployment is severely hampered by surface soiling, which progressively degrades spectral selectivity and sub-ambient cooling performance. In response, we architect a hierarchically structured self-cleaning radiative cooling composite (SRCC) film, directly mimicking the topographic hierarchy of lotus leaves. To enhance the efficiency of microstructure design, we developed a deep-learning surrogate model based on finite-difference time-domain (FDTD) simulations and integrated it with Bayesian optimization to perform a machine-learning-driven inverse design of the top-layer periodic cylindrical microstructure array. This optimization yielded optimal geometric parameters: a cylinder height of 8 μm, a bottom radius of 2.5 μm, and a period of 4 μm. The resulting top-layer microstructures, coupled with Mie scattering from the porous polydimethylsiloxane (PDMS) matrix, enable precise dual-band spectral regulation, achieving a solar reflectance of 94.45% and an atmospheric-window emissivity of 96.5%. Field validation demonstrates that the SRCC film achieves an average sub-ambient temperature difference of 4.8 °C throughout the day, with a peak difference of 11.5 °C. Crucially, the engineered silica-modified interface endows the film with robust superhydrophobicity (water contact angle: 160.75°) coupled with notable oleophobicity (oil contact angle: 109.2°), effectively repelling both droplet-borne and particulate contaminants while preserving optical clarity. Comprehensive accelerated aging, tape-peeling, and water-jet impact assays confirm the exceptional durability of both wettability and optical properties. Furthermore, annual building energy simulations affirm the film's universal energy-saving viability across all climate zones of China and representative climate zones worldwide. Taken together, the multi-level synergistic design strategy provides a generalizable theoretical framework and engineering pathway for the development of radiative cooling materials that reconcile high cooling efficiency with enduring self-cleaning functionality.