\(\Gamma\)-convergence and integral representation for a class of free discontinuity functionals (Q6489760)

The authors analyze the \(\Gamma \)-limits of sequences of integral functionals of the form: \(E^{f,g}(u,A)=\int_{A}f(x,\nabla u)dx+\int_{A}f^{\infty }(x,\frac{dD^{c}u}{d\left\vert D^{c}u\right\vert } d\left\vert D^{c}u\right\vert +\int_{A\cap J_{u}}g(x,[u],\nu _{u})d\mathcal{H }^{d-1}\), defined for all bounded open subsets \(A\) of \(\mathbb{R}^{d}\), \( d\geq 1\), and for all functions \(u\) in a suitable function space contained in the space \(GBV_{\ast }(A)\) of generalized functions with bounded variation on \(A\). Here, \(\nabla u\) is the approximate gradient of \(u\), \( D^{c}u\) is the Cantor part of the distributional gradient \(Du\) of \(u\) when \(u \) belongs to \(BV(A)\), and a suitable measure that extends this notion when \(u \) belongs to \(GBV_{\star }(A)\), \(\left\vert D^{c}u\right\vert \) is the variation of the measure \(D^{c}u\), \(\frac{dD^{c}u}{d\left\vert D^{c}u\right\vert }\) is the Radon-Nykodim derivative of \(D^{c}u\) with respect to \(\left\vert D^{c}u\right\vert \), \(J_{u}\) is the jump set of \(u\), \( [u]\) denotes the amplitude of the jump, \(\nu _{u}\) is the approximate unit normal to \(J_{u}\), and \(\mathcal{H}^{d-1}\) is the Hausdorff measure of dimension \(d-1\). Given an open set \(A\subset \mathbb{R}^{d}\), \(\mathcal{A}(A) \) is the collection of all open subsets of \(A\) and \(\mathcal{A} _{c}(A)=\{A^{\prime }\in \mathcal{A}(A):A^{\prime }\subset \subset A\}\). Given \(A\in \mathcal{A}_{c}(\mathbb{R}^{d})\), the space \(GBV_{\ast }(A)\) is defined as the space of functions \(u:A\rightarrow \mathbb{R}\) such that \( u^{(m)}\in BV(A)\) for every \(m>0\) and \(\sup_{m>0}(\int_{A}\left\vert \nabla u^{(m)}\right\vert dx+\left\vert D^{c}u^{(m)}\right\vert (A)+\int J_{_{u}(m)}\left\vert [u^{(m)}]\right\vert \wedge 1d\mathcal{H} ^{d-1})<+\infty \), \(u^{(m)}\) being the truncation of \(u\) defined as \( u^{(m)}(x)=(u(x)\wedge m)\vee (-m)\). \(f(x,\xi )\) is supposed to have a linear growth with respect to \(\xi \) and \(f^{\infty }\) is its recession function with respect to \(\xi \). \(g\)\ is supposed to be bounded and to satisfy for small \(\left\vert \zeta \right\vert \) the inequalities \( c\left\vert \zeta \right\vert <g(x,\zeta ,\nu )<C\left\vert \zeta \right\vert \), for suitable constants \(0<c<C\). The authors first recall the main properties of \(GBV_{\star }\) functions and they propose a \(\Gamma \)-compact class \(\mathfrak{E}\) of local functionals related to \(GBV_{\star }\) functions, the set \(\mathbb{F}\) be the of functions \(f:\mathbb{R}^{d}\times \mathbb{R}^{d}\rightarrow \lbrack 0,+\infty )\) which are Borel measurable and Lipschitz continuous and satisfy: \(c_{1}\left\vert \xi \right\vert -c_{2}\leq f(x,\xi )\leq c_{3}\left\vert \xi \right\vert +c_{4}\) for every \( x,\xi \in \mathbb{R}^{d}\), and the set \(\mathbb{G}\) of functions \(g:\mathbb{R} ^{d}\times \mathbb{R}\times \mathbb{S}^{d-1}\rightarrow \lbrack 0,+\infty )\) which are Borel measurable, Lipschitz continuous, non-decreasing on \( [0,+\infty )\) and non-increasing on \((-\infty ,0]\) with respect to the second variable, and satisfying \(c_{1}(\left\vert \zeta \right\vert \wedge 1)\leq g(x,\zeta ,\nu )\leq c_{3}(\left\vert \zeta \right\vert \wedge 1)\), \( g(x,-\zeta ,-\nu )=g(x,\zeta ,\nu )\) for every \(x\in \mathbb{R}^{d}\), \(\zeta \in \mathbb{R}\), \(\nu \in \mathbb{S}^{d-1}\). Given \(f\in \mathbb{F}\) and \( g\in \mathbb{G}\), the above-indicated functional \(E^{f,g}\) belongs to the class \(\mathfrak{E}\). The authors prove that if \(E_{k}\) is a sequence in \( \mathfrak{E}\), there exist a subsequence and a functional \(E\in \mathfrak{E} _{sc}\), the class of functionals in \(\mathfrak{E}\) such that for every \(A\in \mathcal{A}(\mathbb{R}^{d})\) the functional \(E(\cdot ,A)\) is lower semicontinuous in \(L^{0}(\mathbb{R}^{d})\), such that for every \(A\in \mathcal{A}_{c}(\mathbb{R}^{d})\) the sequence \(E_{k}(\cdot ,A)\) \(\Gamma \) -converges to \(E(\cdot ,A)\) with respect to the topology of \(L^{0}(\mathbb{R} ^{d})\). They finally prove the following integral representation: given \( E\in \mathfrak{E}_{sc}\) which satisfies a continuity estimate with respect to translations, then \(E=E^{f,g}\), where \(f:\mathbb{R}^{d}\times \mathbb{R} ^{d}\rightarrow \lbrack 0,+\infty )\) is defined through \(f(x,\xi )=\limsup_{\rho \rightarrow 0^{+}}\frac{m^{E}(\ell _{\xi },Q(x,\rho ))}{\rho ^{d}}\), with \(m^{E}(w,A)=inf\{E(u,A):u\in BV(A)\), \(tr_{A}u=tr_{A}w\) \( \mathcal{H}^{d-1}\)-a.e. on \(\partial A\}\), \(g:\mathbb{R}^{d}\times \mathbb{R} \times \mathbb{S}^{d-1}\rightarrow \lbrack 0,+\infty )\) is defined through \( g(x,\zeta ,\nu )=\limsup_{\rho \rightarrow 0^{+}}\frac{m^{E}(u_{x,\zeta ,\nu },Q_{\nu }(x,\rho ))}{\rho ^{d-1}}\). The paper ends with the presentation of two results concerning the convergence of minimum values of some minimum problems related to the functional \(E^{f,g}\), considering the cases of absolute minimizers and a Dirichlet boundary condition.