Subdifferentials of a minimal time function in normed spaces (Q2389268)

Let \(X\) be a normed space with dual \(X^*\) and let \(U\subset X\) be a fixed nonempty bounded closed convex set. For a closed subset \(S\subset X\) the function \(T_S: X\to\mathbb{R}\) according to \[ T_S(x):= \text{inf}\{t\geq 0: S\cap(x+ tU)\neq\emptyset\} \] is called the minimal time function of \(S\) with respect to \(U\). Clearly, it is \(T_S(x)= 0\) if and only if \(x\in S\). If \(U\) is the unit ball of \(X\) then \(T_S\) is the usual distance function of \(S\), if \(U= \{0\}\), then \(T_S\) is the indicator function of \(S\). The authors provide representations of generalized subdifferentials (the proximal subdifferential and the Fréchet subdifferential) of \(T_S(x)\) in terms of the associated normal cones and the support function of \(U\). So for \(x\in S\) it is \[ \partial T_S(x)= N_S(x)\cap\Biggl\{x^*\in X^*: \sup_{u\in U}\,\langle -x^*, u\rangle\leq 1\Biggr\}. \] If \(x\not\in S\) then with \(r:= T_S(x)> 0\) and \(S(r):= \{x\in X: T_S(x)\leq r\}\supset S\) it is \[ \partial T_S(x)\subset N_{S(r)}(x)\cap\Biggl\{ x^*\in X^*: \sup_{u\in U}\,\langle -x^*, u\rangle= 1\Biggr\}, \] where equality holds if \(T_S\) is assumed to be calm. The results are extensions of well-known assertions for subdifferentials of distance functions.