Parabolic second-order tangent sets of semialgebraic sets and applications to polynomial optimization

Abstract

We study parabolic second-order tangent sets of semialgebraic sets and their use in local polynomial optimization. For a basic closed semialgebraic feasible set, we compare the true parabolic tangent set with the algebraic second-order linearized set determined by gradients and Hessians of the active constraints. Under directional rank stability and semialgebraic parabolic arc-realizability, the outer, inner, and arc-generated tangent sets coincide with this algebraic model. Exact formulas are obtained for smooth hypersurfaces, regular complete intersections, smooth inequality systems, and stratified semialgebraic sets. These formulas yield algebraically checkable second-order necessary conditions and sufficient conditions for quadratic growth in polynomial optimization. Examples show how the theory detects curvature, flatness, branch dependence, and the failure of ordinary quadratic scaling.