“…There is a vast literature in this framework (see, for instance, [2,9,10,13,14,18,23,35,41]), motivated, for example, by the study of equilibria for liquid crystals and magnetostrictive materials, where the class of admissible fields is constrained to take values on a certain manifold M (commonly, M = S d−1 , the unit sphere in R d ). As in [2,9,35], the key ingredients in the proof of Theorem 1.2 are the density of smooth functions in W 1,1 (Ω; M) [11,12,29] and a projection technique introduced in [2,30,31]. However, new arguments are required as three main features of our problem prevent us from using immediately the relaxation results concerning the constraint case in the BV setting [2,9,35]: unlike [2,35], our starting point cannot be a tangential quasiconvex function as the energy density considered here (see (1.13)) fail always to satisfy such condition (see Remark 4.2); and unlike the general setting in the literature, (i) our manifold, M = [α, β] × S 2 , has boundary, (ii) the recession function f ∞ in our case (see (1.16)) does not satisfy a hypothesis of the type |f (r, s, ξ, η) − f ∞ (r, s, ξ, η)| C(1 + |(ξ, η)| 1−m ) for some C > 0 and m ∈ (0, 1) (for a.e.…”