Smart window films with ultra-broadband radiative regulation for all-season building energy saving.

Windows are critical pathways for heat exchange and thermal radiation transfer across building envelopes, making temperature-adaptive spectral control essential for year-round building energy conservation. However, the strong coupling among visible, solar, and thermal-infrared responses, together with the high structural sensitivity and complex design space of multilayer films, makes coordinated multi-objective optimization challenging. To address this issue, this study proposes a Deep Q-network (DQN)-based reinforcement-learning inverse design framework coupled with the transfer matrix method (TMM). A two-stage optimization strategy is employed: dielectric materials and layer thicknesses are first optimized simultaneously using a symmetric dielectric/VO2/dielectric cavity, after which the selected material system is fixed to optimize multilayer structures with different layer numbers. The resulting CaF2/VO2 alternating symmetric system enables synergistic regulation across the solar and mid-to-far-infrared regions, with the five-layer CaF2/VO2/CaF2/VO2/CaF2 structure exhibiting the best overall performance. It achieves visible transmittances of 51.06% and 46.97% in the low- and high-temperature states, respectively, together with a solar modulation capability of 12.28% and an average emissivity modulation of 47.33% over 2.5-25 µm. Electric-field analysis attributes this multispectral response to the temperature-dependent reconstruction of optical interference and electromagnetic coupling induced by the VO2 phase transition, while angular analysis confirms stable temperature-selective behavior over a broad range of incident angles. EnergyPlus simulations further demonstrate energy-saving potential across different climate zones, with a maximum annual energy saving of 644.21 MJ per m2 per year in the hot semi-arid climate zone (BSh). These results demonstrate that the proposed DQN-assisted framework provides an effective strategy for designing temperature-adaptive smart windows with coordinated solar and ultra-broadband thermal-radiation regulation.