Sets with at most two-valued metric projection on a normed plane (Q2396428): Difference between revisions

For a nonempty subset \(M\) of a real normed space \(X\), denote by \(P_M(x):=\{y\in M :\|x-y\|=d(x,M)\}\) the metric projection on \(M\), where \(d(x,M):=\inf\{\|x-z\| : z\in M\}\) is the distance from \(x\) to \(M\). The set \(M\) is called proximinal (Chebyshev) if \(P_M(x)\neq\emptyset \) (resp., \(\#(P_M(x))=1\)) for all \(x\in X\). It is obvious that a proximinal set must be closed and every closed nonempty subset of a finite-dimensional normed space is proximinal. If \(X\) is finite dimensional, smooth and strictly convex, then a subset of \(X\) is Chebyshev iff it is closed and convex. If \(\dim X=2\), then \(X\) is smooth iff every Chebyshev subset of \(X\) is closed and convex. The author considers subsets \(M\) of \(X\) satisfying the condition \[ 1\leq\#(P_M(x))\leq 2\leqno{(\ast)} \] for all \(x\in X\). A subset \(M\) of a vector space \(X\) is called \(k\)-convex if every point in the convex hull of \(M\) is a convex combination of \(k\) points in \(M\). By Carathéodory's theorem, every subset of an \(n\)-dimensional vector space is \((n+1)\)-convex. In a previous paper [Mosc. Univ. Math. Bull. 68, No. 6, 275--280 (2013; Zbl 1312.52004); translation from Vestn. Mosk. Univ., Ser I 68, No. 6, 14--19 (2013)], the author proved that any subset of the Euclidean plane \(\mathbb{R}^2\) that satisfies the condition \((\ast)\) is 2-convex, and that there exists a 2-convex subset of \( \mathbb{R}^2\) not satisfying \((\ast)\). The main result of this paper asserts that a 2-dimensional normed space \(X\) is smooth iff every closed subset \(M\) of \(X\) satisfying \((\ast)\) is 2-convex.