We study variational problems for second order supremal functionals $\mathrm F_\infty(u)= \|F(\cdot,u,\mathrm D u,\mathrm{ A}\!:\!\mathrm D^2u)\|_{ \mathrm L^{ \infty}(\Omega)}$, where $F$ satisfies certain natural assumptions, $\mathrm A$ is a positive matrix, and $\Omega \Subset \mathbb R^n$. Higher order problems are very novel in the Calculus of Variations in $\mathrm L^{ \infty}$, and exhibit a strikingly different behaviour compared to first order problems, for which there exists an established theory, pioneered by Aronsson in 1960s. The aim of this paper is to develop a complete theory for $\mathrm F_\infty$. We prove that, under appropriate conditions, ``localised" minimisers can be characterised as solutions to a nonlinear system of PDEs, which is different from the corresponding Aronsson equation for $\mathrm F_\infty$; the latter is only a necessary, but not a sufficient condition for minimality. We also establish the existence and uniqueness of localised minimisers subject to Dirichlet conditions on $\partial \Omega$, and also their partial regularity outside a singular set of codimension one, which may be non-empty even if $n=1$.
Categories: math.AP