Minimal noise subsystems

A system subjected to noise contains a decoherence-free subspace or subsystem (DFS) only if the noise possesses an exact symmetry. Here we consider noise models in which a perturbation breaks a symmetry of the noise, so that if S is a DFS under a given noise process it is no longer so under the new perturbed noise process. We ask whether there is a subspace or subsystem that is more robust to the perturbed noise than S. To answer this question we develop a numerical method that allows us to search for subspaces or subsystems that are maximally robust to arbitrary noise processes. We apply this method to a number of examples, and find that a subsystem that is a DFS is often not the subsystem that experiences minimal noise when the symmetry of the noise is broken by a perturbation. We discuss which classes of noise have this property.