A quadratically constrained minimization problem arising from PDE of Monge-Ampère type (Q849146)

A quadratically constrained eigenvalue minimization problem is considered. These problems arise in the numerical solution of fully nonlinear three dimensional Monge-Ampère type equations, known as the Dirichlet problem for the \(\sigma_2\)-operator. The nonlinear Dirichlet problem with the elliptic \(\sigma_2\)-operator is linearized in a neighborhood of the solution. The theory and a solution technique is developed for the least-squares minimization of the linearized \(\sigma_2\) problem. The minimization subproblem has to be solved many times during the numerical solution of the Monge-Ampère equation. Therefore an efficient numerical technique is required for the solution of minimization problem. It turns out that the proposed algorithm is finite, of complexity \(\mathcal{O}(n^3)\) and requires additionally solving a simple scalar secular equation. As a numerical example, two dimensional minimization to the solution of the Dirichlet problem for the two-dimensional Monge-Ampère equation, is considered. The numerical results indicate the excellent convergence behavior of the algorithm.