Supremally generating cones of the space of continuous functions (Q756656): Difference between revisions

C(T) denotes the (real) Banach space of continuous real-valued functions on the compact metric space T, and H denotes a convex cone in C(T). H is coinitial if for each \(f\in C(T)\), H contains an element \(h\leq f\). Connections with the Korovkin theorems on convergence of sequences of operators and functionals lead to the study of the following notion: a coinitial cone H in C(T) generates C(T) with order p if, whenever \(f\in C(T)\), \(\epsilon >0\), and points \(z_ 1,...,z_ p\) of T are given, then H contains an element \(h\leq f\) such that \(h(z_ k)>f(z_ k)-\epsilon\) \((k=1,...,p)\). The bulk of this paper consists of characterizing this generating property. The main characterization, from which the others are derived, is Theorem 1: The coinitial cone H generates C(T) with order p iff given arbitrary positive numbers \(\epsilon\), \(a_ 1,...,a_ p\) and points \(z_ 1,...,z_ p\) of T with pairwise disjoint neighborhoods \(V_ 1,...,V_ p\), there is \(h\in H\) such that \(h(z_ k)>a_ k- \epsilon\) and \(h\leq a_ k\) on \(V_ k\) \((k=1,...,p)\), and \(h\leq 0\) off the union of the \(V_ k\). Examples to which some of the theorems apply are given. A few observations on some details are in order. In the proof of sufficiency in Theorem 1, one needs \(\epsilon_ 1<\min \{\epsilon,a_ 1,...,a_ p\}\), not \(\epsilon_ 1<\min \{\epsilon,\max \{a_ 1,...,ap\}\}\), to ensure that \(a_ k-(\epsilon_ 1/2)\) be positive for every k. The system \(\Phi_{n,p}\) introduced on page 923 is linearly independent only if all the \(f_ k\) are nonconstant; in any case, there is duplication in the list presenting \(\Phi_{n,p}\). Finally, in connection with Theorem 4, some readers may be surprised to learn that if T is infinite then there are in C(T) infinite-dimensional locally compact closed convex cones containing -1; for instance, if \(T=[a,b]\) let H consist of those h such that \(| h(x)-h(y)| +h(a)d(x,y)\leq 0\) for all x, y in T.