“…In the numerical computations, we apply finite-element approximation spaces based on rectangular elements, and we opt to approximate the error in (40) by u h/2 − u h , where u h/2 is an approximation obtained from a Galerkin approximation in a finite-element space V h/2 0 in which each element is uniformly divided into 4 elements; see Figure 2. The refined finite-element space V h/2 0 moreover serves to construct an approximation to the dual solution in (40). In summary, considering an approximation u h ∈ V h 0 and a refined approximation u h/2 ∈ V h/2 0 , the worst-case multi-objective error estimate pertaining to u h is r(u h ), z h/2 with z h/2 according to Figure 1 plots the worst-case multi-objective error estimate, r(u h ), z h/2 , versus the dimension of the finite-element approximation space, dim V h 0 , for finite-element approximations with polynomial degrees p ∈ {1, 2, 3, 4} on uniform meshes with mesh width h = 2 −2 , 2 −3 , .…”