A duality formula for multidimensional vector fields (Q1264947)

The main result of this paper is a theorem stating that the supremum of the appropriately normalized circulation of a vector field is equal to the \(L^{\infty}\)-distance from that field to the linear manifold of gradient vector fields. Let \(\Omega\) be a domain in \(R^{n}\), and \(R(y)\) be a continuous vector field on \(\Omega\). Then \[ \sup_{y(t)} { {\int_{t_0}^{t_1}R(y(t))u(t)dt} \over {\int_{t_0}^ {t_1}| u(t) | dt} }=\inf_{S(y)} \max_{\Omega} | R(y)-S(y) |, \] where the supremum on the left-hand side is taken over all closed contours \(y(t) \in \Omega\), \(y(t_0)=y(t_1)\), \(\dot {y}(t) = u(t)\), and the infimum on the right-hand side is taken over all gradient fields \(S(y)\) on \(\Omega\). The author also proves a generalization of this statement relating to the problem \(\sup_{y(t)} { {\int_{t_0}^{t_1}R(y(t))u(t)dt} \over {\int_{t_0}^{t_1}\varphi (u(t), y(t))dt}}\). The formulas obtained are applied to the study of optimality conditions for singular problems in optimal control theory.