An elementary proof of the oscillation lemma for weak Markov systems (Q919558)

Let \(f_ 0,f_ 1,...,f_ n\) be linearly independent set of real functions on a subset of the line. \(F_ k=span\{f_ 0,f_ 1,...,f_ k\}\) is said to be an alternation space if no member has an alternation of length \(k+2:\) i.e. there do not exist points \(t_ 0<t_ 1<...<t_{k+1}\) such that \(f(t_ i)f(t_{i+1})<0\). If for each k, \(F_ k\) is an alternation space then \(F_ n\) is called an alternation Markov space. Even if \(F_ n\) is an alternation space it may contain a function f that changes direction \(k+2\) times; that is, that \([f(t_ i)- f(t_{i-1})][f(t_{i+1})-f(t_ i)]<0.\) If no such f exists, \(F_ n\) is called an oscillation space. There is an intriguing chain of ideas characterizing oscillation spaces and their relation to alternation spaces. One of the links is the theorem that if \(F_ n\) is an alternation Markov space with \(f_ 0\) equal 1 then \(F_ n\) is an oscillation space. Zielke's original proof used a Gauss kernel approximation of alternation spaces. This paper presents a more basic proof.