High-quality quasi-monochromatic near-field radiative heat transfer designed by adaptive hybrid Bayesian optimization

The increasing demand for versatile and high-quality near-field radiative heat transfer (NFRHT) has created a critical need for a design approach that can handle numerous candidate structures. In this work, we employ and develop an adaptive hybrid Bayesian optimization (AHBO) algorithm to design the high-quality quasi-monochromatic NFRHT. The candidate materials include hexagonal boron nitride, silicon carbide, and doped silicon. The high-quality quasi-monochromatic NFRHT is optimized over 1.0 × 108 candidate structures to maximize the evaluation factor. It is worth noting that only 2.6% of the candidate structures needed to be calculated to identify the optimal structure. The optimal structure of quasi-monochromatic NFRHT is an aperiodic multilayer metamaterial that differs from conventional periodic multilayer structures. Moreover, we investigate the robustness and mechanisms of the optimal quasi-monochromatic NFRHT with respect to the vacuum gap distance and the temperature difference between the emitter and receiver. In addition, the high-quality multi-peak NFRHT is designed using the AHBO algorithm by improving the definition of the evaluation factor. The results demonstrate that the AHBO algorithm is efficient in designing high-quality quasi-monochromatic and multi-peak NFRHT, and it can be further expanded to other structural designs in the field of energy conversion.