Relaxation for partially coercive integral functionals with linear growth

We prove an integral representation theorem for the $\mathrm{L}^1(\Omega;\mathbb{R}^m)$-relaxation of the functional $$\mathcal{F}\colon u\mapsto\int_\Omega f(x,u(x),\nabla u(x))\;\mathrm{dd } x,\quad u\in\mathrm{W}^{1,1}(\Omega;\mathbb{R}^m),\quad\Omega\subset\mathbb{R}^d\text{ open,}$$ to the space $\mathrm{BV}(\Omega;\mathbb{R}^m)$ under very general assumptions, requiring principally that $f$ be Carath\'eodory, partially coercive, and quasiconvex in the final variable. Our result is the first of its kind which applies to integrands which are unbounded in the $u$-variable and thus allows to treat many problems from applications. Such functionals are out of reach of the classical blow-up approach introduced by Fonseca & M\"uller [Arch. Ration. Mech. Anal. 123 (1993), 1--49]. Our proof relies on an intricate truncation construction (in the $x$ and $u$ arguments simultaneously) made possible by the theory of liftings as introduced in the companion paper arXiv:1708.04165, and features techniques which could be of use for other problems featuring $u$-dependent integrands.