SENSITIVITY CALCULATION IN SPECTRO-RADIATIVE COUPLING: A COMPARATIVE STUDY FOR THREE MONTE CARLO PATH INTEGRALS

Radiative transfer (RT) modeling is crucial in many fields such as combustion chambers, fire safety, climate studies, Earth remote sensing, planetary atmosphere modeling or heat transfer in porous media. These diverse applications require efficient and accurate computation not only of radiative quantities but also of their sensitivities to various model parameters. In complex settings (3D media, multiple-scattering environments,..) Monte Carlo (MC) ray tracing is particularly well-suited. Highly heterogeneous fields can be handled using spatial fictitious events—null collisions—introduced to ensure unbiased calculations. Acceleration techniques can be used to minimize the amount of null collisions and reduce the CPU cost. However, prior research has shown that minimizing these null collisions hinders accurate sensitivity calculations when using the standard method of path-integral differentiation. In this work, we encounter similar sensitivity calculation challenges, where highly varying spectra necessitate the introduction of spectral fictitious events—null transitions—combined with acceleration techniques that minimize them. We explore two alternative path-integral reformu- lations to address this issue: one inspired by solutions developed for the spatial domain and another one derived from transport modeling of sensitivity. A comparative study on an academic test case evaluates their respective strengths and limitations.