English

Method of nose stretching in Newton's problem of minimal resistance

Optimization and Control 2021-08-11 v2 Metric Geometry

Abstract

We consider the problem inf{ ⁣ ⁣Ω(1+u(x,y)2)1dxdy: the function u:ΩR is concave and 0u(x,y)M for all (x,y)Ω={(x,y):x2+y21}}\inf\big\{ \int\!\!\int_\Omega (1 + |\nabla u(x,y)|^2)^{-1} dx dy : \text{ the function } u : \Omega \to \mathbb{R} \text{ is concave and } 0 \le u(x,y) \le M \text{ for all } (x,y) \in \Omega =\{ (x,y): x^2 + y^2 \le 1 \} \, \big\} (Newton's problem) and its generalizations. In the paper \cite{BrFK} it is proved that if a solution uu is C2C^2 in an open set UΩ\mathcal{U} \subset \Omega then detD2u=0\det D^2u = 0 in U\mathcal{U}. It follows that graph(u)U(u)\rfloor_\mathcal{U} does not contain extreme points of the subgraph of uu. In this paper we prove a somewhat stronger result. Namely, there exists a solution uu possessing the following property. If uu is C1C^1 in an open set UΩ\mathcal{U} \subset \Omega then graph(uU)(u\rfloor_\mathcal{U}) does not contain extreme points of the convex body Cu={(x,y,z):(x,y)Ω, 0zu(x,y)}C_u = \{ (x,y,z) :\, (x,y) \in \Omega,\ 0 \le z \le u(x,y) \}. As a consequence, we have C_u = \text{\rm Conv} (\overline{\text{\rm SingC_u}}), where SingCuC_u denotes the set of singular points of Cu\partial C_u. We prove a similar result for a generalized Newton's problem.

Keywords

Cite

@article{arxiv.2003.06682,
  title  = {Method of nose stretching in Newton's problem of minimal resistance},
  author = {Alexander Plakhov},
  journal= {arXiv preprint arXiv:2003.06682},
  year   = {2021}
}

Comments

28 pages, 5 figures