Machine-learning-assisted environment-adaptive thermal metamaterials

Adaptive metamaterials have prevailed recently owing to their extraordinary features like dynamic response to external interference. However, highly complicated parameters, narrow working ranges, and supervised manual intervention are still long-term and tricky obstacles to the most advanced self-adaptive metamaterials. To surmount these barriers, we present environment-adaptive thermal metamaterials driven by machine learning, which can automatically sense ambient temperatures and regulate thermal functions promptly and continuously. Thermal functions are robust when external thermal fields change their directions, and simulations and experiments exhibit excellent performance. Based on this, we further design two metadevices with on-demand adaptability, performing distinctive features with isotropic materials, wide working temperatures, and spontaneous response. This work provides a paradigm for intelligent diffusion metamaterial design and can be extended to other diffusion fields, responding to more complex and variable environments.