Note on \((m,q)\)-isometries on an hyperspace of a normed space (Q2346166)

In this paper the author considers the hyperspace \(k(X)\) consisting of all non-empty compact convex subsets of a normed space \(X\), endowed with the Hausdorff distance given by \(h(C,D):=\inf\{\epsilon>0:C\subset D+\epsilon B_X\text{ and }D\subset C+\epsilon B_X\}\), where \(B_X\) is the unit closed ball of \(X\). After discussing several properties of this distance \(h\), the author introduces linear maps \(\mathcal{T}:k(X)\to k(X)\) and a norm of \(\mathcal{T}\) by \[ \|\mathcal{T}\|:=\sup_{\{0\}\neq C\in k(X)}\frac{\|\mathcal{T}C\|}{\| C\|}=\sup_{C\in k(X),\| C\|=1}\|\mathcal{T}C\|. \] Here \(\| C\|:=\sup_{x\in C}\| x\|=h(C,\{0\})\), for any \(C\in k(X)\). \(\mathcal{T}\) is called bounded if \(\|\mathcal{T}\|<\infty\). Observing some results regarding equivalence of boundedness and continuity of \(\mathcal{T}\) which are similar to analogous facts about linear operators between normed spaces the author finally proves that \(T\in L(X)\) is an isometry if and only if the map \(k(T):k(X)\to k(X)\) defined by \(k(T)C:=TC\), \(\forall\,C\in k(X)\), is an isometry. Here \(L(X)\) denotes the class of all bounded linear operators \(T:X\to X\). Concerning the similar question about \((m,q)\)-isometries (\(m\geq1\) integer, \(q>0\) real) on the hyperspace \(k(X)\) the author shows that for \(T\in L(X)\) if \(k(T)\) is an \((m,q)\)-isometry then \(T\) is also an \((m,q)\)-isometry although the converse is not true as is shown by considering an arbitrary weighted shift operator \(S_w\in L(\ell_2)\) which is a \((2,2)\)-isometry whereas \(k(S_w)\) is not a \((2,2)\)-isometry. A linear map \(\mathcal{T}:k(X)\to k(X)\) is called an \((m,q)\)-isometry if \[ \sum_{i=0}^m(-1)^{m-i}\binom{m}{i}h(\mathcal{T}^{i}C,\mathcal{T}^{i}D)^{q}=0\quad(C,D\in k(X)). \] A similar definition follows for \(T\in L(X)\) with only replacing \(h\) by the distance induced by the norm of \(X\). Finally the author considers the Radström space \(\widehat{k(X)}\) associated to the hyperspace \(k(X)\) which is the quotient space \(\frac{k(X)\times k(X)}{\sim}\) induced by the equivalence relation \(\sim\) defined by : \((C,D)\sim(E,F)\Longleftrightarrow C+F=D+E\), where \(C,D,E,F\in k(X)\) and which becomes a normed space with some suitably defined operators and norm. Moreover \(k(X)\) can be embedded isometrically into \(\widehat{k(X)}\). For a \(\mathcal{T}\in L(k(X))\) defining \(\widehat{\mathcal{T}}:\widehat{k(X)}\to\widehat{k(X)}\) by \(\widehat{\mathcal{T}}[C,D]:=[\mathcal{T}C,\mathcal{T}D]\), where \([C,D]\) is the equivalence class containing \((C,D)\) with respect to \(\sim\), the author proves that \(\mathcal{T}\) is a strict \((m,q)\)-isometry if and only if \(\widehat{\mathcal{T}}\) is a strict \((m,q)\)-isometry, where \(\mathcal{T}\in L(k(X))\) is called a strict \((m,q)\)-isometry if it is not an \((m-1,q)\)-isometry.