Stochastic homogenization of degenerate integral functionals with linear growth (Q6039200)

The authors study a problem of stochastic homogenisation. They consider the family (in \(\varepsilon > 0\)) of non-convex, vectorial, random integral functionals defined in \(L^1_{\mathrm{loc}} (\mathbb{R}^d ;\mathbb{R}^m) \times {\mathcal A} \to [0, +\infty]\) (\({\mathcal A}\) the set of open and bounded subsets of \(\mathbb{R}^d\) with Lipschitz boundary) \[ F_{\varepsilon} (\omega) (u, A) := \left\{ \begin{array}{ll} {\displaystyle \int_A f \left( \omega, \frac{x}{\varepsilon}, \nabla u \right) dx } & \text{ if } u \in W^{1,1}(A;\mathbb{R}^m) , \\ + \infty & \text{ otherwise } \end{array} \right. \] with \(\omega \in \Omega\) (\((\Omega, {\mathcal F},\mathbf{P})\) a probabilty space) satisfying an ergodicity condition and \(f\) satisfying the degenerate condition \[ \alpha \big| \xi \Lambda (\omega, x) \big| \leqslant f \left( \omega, x, \xi \right) \leqslant \big| \xi \Lambda (\omega, x) \big| + \lambda (\omega, x) \] for suitable functions \(\lambda\) and \(\Lambda\) and constant \(\alpha > 0\). The homogenised functional is deterministic and depends on a function \(f_{\mathrm{hom}}\) which is not satisfying a degenerate condition as \(f_{\varepsilon}\) do, but \[ \alpha c_0 | \xi | \leqslant f_{\mathrm{hom}} \left( \xi \right) \leqslant C_0 | \xi | + C_1 \] with suitable constants \(c_0, C_0, C_1\) depending on \(\lambda\) and \(\Lambda\). They also explicit the \(\Gamma\)-limit \(F_{\mathrm{hom}}\) showing that this is finite on \(BV (A;\mathbb{R}^m)\) and not only in \(W^{1,1}(A; \mathbb{R}^m)\).