Stochastic effects often occur in real-life applications of electromagnetic scattering. Examples are the propagation of radio waves in the atmosphere or in saline fluids like sea water, or penetration of electromagnetic energy in a human body for hyperthermia. In these examples, only knowledge about the average regarding the propagation medium properties and its spatial distribution is available. Therefore we adopt a stochastic modeling approach to characterize electromagnetic scattering for such problems.

We distinguish between three cases in which stochastic variation occurs, i.e. in the excitation field, in the target, and in the environment. The latter two cases are computationally in the most challenging and also the most interesting from an engineering point of view. A further distinction can be made between intentional stochastic effects, such as occur in various EMC/EMI testing procedures for example in a Vibrating Intrinsic Reverberation Chamber (VIRC) or a mode-stirred chamber, and unintentional stochastic effects, e.g. due to manufacturing tolerances, mechanical vibrations, and inhomogeneities in medium properties. Outside EMC/EMI, especially the manufacturing tolerances and medium inhomogeneities play an importentrole, e.g. in the production of solar cells, lasers, and integrated structures.

The aim of the project is to develop efficient and robust computational modeling-and-design methods that are capable of taking into account stochastic variations in both the electromagnetic field and the scattering setup, for application in an industrial or medical context e.g. for risk assessment or yield optimization.

Classical methods that are frequently applied are sampling methods, which include Monte-Carlo methods and techniques like sparse-grids and space-filling curves. Sampling methods have the distinct advantage that they can use existing electromagnetic solvers as a black box and all samples can be evaluated independent from each other. Nevertheless, these methods require a substantial amount of samples to reach a sufficent level of accuracy in e.g. the stochastic moments of the observable of interest and we are looking for alternative approaches.

This research is partially funded and performed as part of BESTCOM, which is part of the Interuniversity Attraction Pole Programme initiated by the Belgian Science Policy Office.