Vanguard Design of Smart Windows: Multi-Objective Optimisation of Dynamic Glazings for Radiative Cooling.

Abstract

Windows represent the primary thermal vulnerability in sustainable building envelopes, standing at the intersection of conflicting architectural and thermodynamic demands : providing high visual transparency while simultaneously managing massive solar and thermal radiative fluxes. While contemporary research frequently focuses on optimizing single functionalities in isolation, designing a truly multi-functional smart window requires satisfying these competing constraints simultaneously across distinct spectral regimes. This thesis addresses this bottleneck through a twofold contribution : the development of a fully computational, multi-objective optimization framework and the design of a performant physical blueprint for a dynamic radiative cooling glazing system. Methodologically, the traditional trial-and-error approach to multilayer optical design is su- perseded by an automated workflow coupling the Transfer Matrix Method (TMM) with a Non-dominated Sorting Genetic Algorithm (NSGA-II). This framework successfully navigates the vast, multidimensional geometric space of layer thicknesses to map entire Pareto-optimal fronts. Physically, the algorithmic mapping revealed that the direct integration of a thermo- chromic vanadium dioxide (VO2) layer severely degrades visible transparency (Tlum < 70%). To overcome this limitation, the structural paradigm was shifted toward a resonant design, culminating in the Fabry–P´erot Angular-Selective Thermochromic (FAST) window architecture. Optimization of a ZnSe-based FAST stack unlocked a large Long-Wave Infrared (LWIR) swit- ching capacity of up to ∆εLWIR = 0.50 while maintaining positive solar modulation (∆Tsol > 0). Crucially, the optimization unveiled a fundamental physical trade-off : the resonant, isotropic enhancement of emissivity within the Fabry–P´erot cavity inherently dilutes the directional emission engineered by the top Epsilon-Near-Zero (ENZ) layers, forcing the Angular Enfor- cement Factor below unity (AEF < 1) at maximum switching. Nevertheless, the workflow successfully identified viable intermediate Pareto states that, for the first time, simultaneously reconcile all competing criteria : satisfactory luminous transparency (Tlum > 70%), high color fidelity ( CRI > 90), positive solar modulation, highly positive LWIR switching (∆εLWIR ≫ 0), and finite angular selectivity (AEF > 1). Ultimately, this work demonstrates that advanced genetic algorithms can successfully architect existing materials into high-performance resonant structures, laying a comprehensive foundation for next-generation energy-efficient smart glass.