WorksMathematics
Material optimization of optical properties of particulate products by transition matrix method and sequential global programming
Abstract
Material optimization of nanoparticle assemblies, such as structurally coloured photonic crystals, attracts increasing interest. We combine the transition matrix method (TMM), which avoids spatial discretization of particles and surrounding medium, with sequential global programming (SGP) to optimize optical properties of particulate products by iteratively assigning predefined catalogue materials to individual particles. We derive exact separable models, reproducing the fully coupled objective under single-particle material changes, and computationally accessible first-order correct approximations for general continuously differentiable objectives. These approximations aim to capture the nonlinear, potentially nonconvex dependence on individual material variables beyond their local first-order behaviour. Within SGP, globally minimizing the particle-wise subproblems can yield substantial improvements in few outer iterations, including steps away from unfavourable stationary points, as previously demonstrated for discrete dipole approximation-based SGP. The construction exploits the decomposition of the TMM system matrix into a material-dependent block-diagonal and a material-independent coupling matrix, while the state and adjoint problems remain fully coupled. Applications include extinction, scattering quantities, and colour objectives. In a 35-particle extinction minimization problem, the objective decreases by 31.0% in 13 iterations. A fine-grid evaluation of the fully coupled objective shows that no tested single-particle material change improves the final design. For a 627-particle photonic crystal, chroma increases by 58.4-122.1% for different particle diameters in 17-42 iterations. The optimized designs favour absorbing material in layers away from the incident light, consistent with experimental findings.