Locally Lipschitz stability of a parametric semilinear elliptic optimal control problem with mixed constraints (Q6167078): Difference between revisions

The subject of the paper is minimization of the cost functional \[ J(y, u, \lambda) = \int_\Omega \big(\eta(x, y(x), \lambda(x)) + \zeta(\lambda(x))u(x) + \mu(\lambda(x))u^2(x)\big)dx \tag{1} \] where \(\Omega\) is an open and bounded domain in \(\mathbb{R}^n\) with boundary \(\partial \Omega\) \((n = 2, 3)\). The problem is: given \(\lambda(\cdot) \in L^\infty(\Omega)\) find a control function \(u(\cdot) \in L^2(\Omega)\) that minimizes (1) where \(y(x) \in H^2(\Omega) \cap H_0^1(\Omega)\) satisfies \begin{align*} Ay(x) + h(x, y(x), \lambda(x)) &= u(x) + \lambda(x) \quad (x \in \Omega) \quad y(x) = 0 \ \ (x \in \partial \Omega) \\ f(x, y(x), \lambda(x)) + \delta(x)u(x) &\ge 0 \quad (x \in \Omega) \\ u(x) &\ge 0 \quad (x \in \Omega). \end{align*} The operator \(A\) is defined by \[ Ay(x) = - \sum_{i, j = 1}^n D_j \big(a_{ij}(x) D_iy(x)\big) + a_0(x)y(x) \quad \big( a_{ij}(x) = a_{ji}(x) \big) \] with smooth \(a_{ij}(\cdot)\) satisfying \(m|\xi|^2 \le \sum_{i, j}a_{ij}(x) \xi_i \xi_j,\) \(\xi = (\xi_1, \dots \xi_n)\) \((m > 0)\) and \(a_0(\cdot) \in L^\infty(\Omega),\) \(a_0(x) \ge 0\) on \(\Omega,\) \(a_0(x) > 0\) in a set of positive measure. The subject of this paper is continuous dependence of the solution of the minimization problem with respect to the parameter \(\lambda(\cdot)\) in a neighborhood of the solution \((y_0(\cdot), u_0(\cdot))\) corresponding to \(\lambda(x) = \lambda_0(x).\) Under suitable conditions on all coefficients and a second order optimality condition for \((y_0(\cdot), u_0(\cdot))\), involving the Lagrangian in a critical cone, the author shows that solutions \((y(\cdot), u(\cdot))\) to the optimal problem exist for \(\|\lambda - \lambda_0\|_\infty \le \epsilon\) \((\epsilon > 0\) and satisfy an inequality of the form \[ \|(y, u) - (y_0, u_0)\|_{\infty} \le K \|\lambda - \lambda_0\|_\infty. \] There are also estimates in the \(L^2\) norm and in other norms.