On threading continuous functions through compact sets (Q790264)

This paper deals with the following problem: Under what circumstances for a given compact set E of \(<0,1>\times<0,1>,\) for every point \((u,v)\in E,\) there exists a continuous function \(f:<0,1>\to<0,1>\) such that \(f(u)=v\) and the graph g(f) of f is a subset of E? A compact subset E of \(<0,1>\times<0,1>\) is called an admissible set iff the function \(E^*\) defined on \(<0,1>\) into the set of all non-empty compact subsets of \(<0,1>\), furnished with the Hausdorff metric, in the following way: \(E^*(x)=E_ x\) for each \(x\in<0,1>,\) where \(E_ x=\{y\in<0,1>:(x,y)\in E\},\) is continuous. The problem has an affirmative solution only if E is an admissible set. For the admissible set \[ E=(<0,1>\times \{0,1\})\cup(\{0\}\times<0,1>)\cup(\cup^{\infty}_{k=1}g(g_ k)), \] where \(g_ k\) is the function \(y=kx\) defined on \(<0,1/k>\) and for any \((0,v)\in \{0\}\times(0,1)\) does not exist a continuous function \(f:<0,1>\to<0,1>\) for which \(f(0)=v\) and \(g(f)\subset E\). First, the author proved that there is an affirmative answer for the problem if E is an admissible set for which \(E_ x\) is a nowhere dense subset of \(<0,1>\) for each \(x\in<0,1>\). The proof is constructive based on the limiting process of \(\epsilon\)-functions. Let \(\epsilon>0\). Then f is called an \(\epsilon\)-function iff: there exist a and b such that \(0\leq a<b\leq 1,\) f is defined and right continuous on \(<a,b)\), the graph g(f) of f is a subset of E, for each \(x\in<a,b)\) there exists \(\delta>0\) such that \(x+\delta \leq b\) and f is continuous on \((x,x+\delta)\) and \(0\leq f(x)- \overline{\lim}_{s\to x-}f(s)<\epsilon\) for every \(x\in(a,b)\). In the general case, for an admissible set E there is an affirmative answer iff E has the desired property locally; that is: for each \((u,v)\in E\) there exists an open interval I containing u and a continuous function g defined on \(I\cap<0,1>\) such that g(g)\(\subset E\) and \(g(u)=v\). The proof of the theorem in the general case is based on an inductive process on R and on the notion of a mostly continuous E-function from (u,v). A mostly continuous E-function from (u,v) is a pair \((f,\Gamma_ f)\), where f is a function defined on a non-degenerate interval \(<u,t_ f>\subset<u,1>, f(u)=v\), \(g(f)=E\), \(\Gamma_ f\) is a well-ordered, by the normal orderings on R, closed subset of \(<u,t_ f>\) not containing u, f is relatively continuous on \(<u,t_ f>-\Gamma_ f\) and for every \(t\in\Gamma_ f\) it holds: \[ \frac12 \left(\limsup_{s\to t-}f(s) + \liminf_{s\to t-} f(s)\right) = f(t) = \lim_{s\to t+} f(s) \] (where the last equality is ignored if \(t=t_ f)\). The theorem in the general case says: ''Let E be an admissible subset of \(<0,1>\times<0,1>\) having the local property and \((u,v)\in E\). Then there is a continuous function \(f:<0,1>\to<0,1>\) for which \(f(u)=v\) and g(f)\(\subset E.''\) In the general case, an admissible set E has local continuity property if each interior point of a vertical segment is a two-dimensional interior point of A and E has the last property if the function \(E^{c^*}\) defined on \(<0,1>\) by \(E^{c^*}(x)=\overline{E^ c_ x}\) is continuous, where \(\overline{E^ c_ x}\) is the closure of the set \(<0,1>-E_ x.\) An example shows that neither of these last conditions is necessary. At the end of the paper there are two notes - applications to path derivatives, and the following problem: Is it possible to phrase a sufficient condition in the general case in terms of local arc-wise connectedness of E? Does the local arc-wise connectedness at every point of E imply the local continuity property?