“…A common feature of most of the recently developed methods is to try to circumvent the inceptive ill-posedness of shape optimization problems which manifests itself, in numerical practice, by the occurrence of many local minima, possibly far from being global. Probably the most successful approach is the homogenization method [1,6,7,19,23]: it allows to find a global minimizer in most instances, at the price of introducing composite materials in the optimal shape (a tricky penalization procedure is required for extracting a classical shape out of it). Unfortunately, the rigorous derivation of the homogenized or relaxed formulation of shape optimization is complete only for a few, albeit important, choices of the objective function (mostly self-adjoint problems like compliances or eigenvalues optimization).…”