Block-Krylov discrete-dipole and volume-integral-equation solvers for multi-orientation light scattering across the dielectric–metal contrast range

Simulation-based inference of the physical properties (size, refractive index, shape) of individual environmental particles from the observation signals of recent optical particle analyzers, such as Complex Amplitude Sensing version 2 (CAS-v2), requires a versatile light-scattering solver that (i) handles arbitrary particle shape and internal composition and (ii) efficiently executes the simulation for many particle orientations. Existing open-source volume-integral solvers based on the discrete dipole approximation (DDA) or the volume integral equation method (VIEM) satisfy (i) but solve one orientation at a time, making (ii) difficult in practice. We develop two volume-integral scattering solvers — block-DDA_Py on a cubic lattice and block-VIEM.jl on a tetrahedral mesh — that handle many particle orientations efficiently using block-Krylov subspace methods. The two solvers are suited, in accuracy and cost, to complementary refractive-index regimes: DDA to low-contrast targets ( | m p / m m | ≲ 2 ) and VIEM to high-contrast targets ( | m p / m m | ≳ 2 ). We benchmark both solvers on four particle shapes (sphere, oblate spheroid, sphere doublet, Gaussian random sphere) over the wavelength-scale size range targeted by the current CAS-v2 instruments. We demonstrate that the block-Krylov approach improves convergence (fewer iterations to reach a given residual) and yields a monotonic per-orientation speedup at least up to right-hand-side block sizes of order 1 0 2 , making the orientation sweeps required by simulation-based inference tractable on a single workstation.