“…An important choice to be made is whether to use continuous or discontinuous finite elements. Although the choice of continuous finite element approximations for the Lagrange multipliers μ k was advocated, for instance, in [1], [2], we believe that allowing discontinuities is a healthy choice, at least when the interface Σ is not smooth (which means, in this context, that its normal unit vector n has jumps that do not tend to zero with the mesh size), as it will almost always be the case when dealing with more than two subdomains. Indeed, in view of (2.35), the solution λ k has to jump whenever n jumps (as we expect each σ σ σ(u k ) to be smooth).…”