No label defined (Q4484764)

The problem is minimization of the functional \[ J(u) = \int_\Omega \sqrt{1 + |Du|^2} + \int_{\Omega_+} Q^2 d{\mathcal L}^n + \int_{\partial \Omega} |u - u_0|d {\mathcal H}^{n - 1} \] over \(u \in \text{BV}(\Omega),\) where \(\Omega\) is a \(n-\)dimensional domain with Lipschitz boundary \(\partial \Omega,\) \(u_0\in \text{BV}(\Omega)\) is nonnegative, \(Q \in L^2(\Omega)\) is positive and \(\Omega_+ = \{x \in \Omega; u(x) > 0\}.\) In the second integral \({\mathcal L}^n\) is the \(n\)-dimensional Lebesgue measure, and in the third \({\mathcal H}^{n - 1}\) is the \((n - 1)\)-dimensional Hausdorff measure; \(\text{BV}(\Omega)\) consists of all functions whose derivatives are signed measures of finite total variation in \(\Omega.\) The integrand \(\sqrt{1 + |Du|^2}\) is the total variation measure of \(({\mathcal L}^n, D_1 u, \dots, D_n u).\) A convexity property of the free boundary \(\Omega_0 = \{x \in \Omega; u(x) = 0\}\) was previously proved by Caffarelli and Friedman under the hypotheses \(n \leq 6\) and \(\sup_\Omega Q < 1,\) which produce Lipschitz continuity of the minimizer. In this paper the second condition is replaced by a bound from below, under which the minimizer may not even be continuous (as shown by the author in a previous paper).