Walks on ordinals and their characteristics (Q2372415)

The first uncountable ordinal, \(\omega_1\), behaves in a markedly different way than the first infinite ordinal, \(\omega_0\). Every continuous real-valued function on the space~\(\omega_1\) is constant on a tail; the function \(n\mapsto 2^{-n}\) is continuous on~\(\omega_0\) and injective. Kőnig's Lemma on branches in infinite trees fails for trees of height~\(\omega_1\), as witnessed by Aronszajn trees. The function \(n\mapsto n-1\) is injective and maps every finite nonzero ordinal to a smaller one; the Pressing-Down lemma states that a `pressing-down' function on~\(\omega_1\) must be constant on a big set. The author of this book has spent much time investigating the reason behind this behaviour of~\(\omega_1\) and makes an excellent case for having discovered from which all things flow. The idea is to assign to every countable ordinal~\(\alpha\) a cofinal subset~\(C_\alpha\); if \(\alpha=\beta+1\) put \(C_\alpha=\{\beta\}\) and if \(\alpha\)~is a limit ordinal let \(C_\alpha\) be some simple increasing sequence of successor ordinals, cofinal in~\(\alpha\). Given ordinals \(\alpha\) and \(\beta\) with \(\alpha<\beta\) one can `walk' from~\(\beta\) down to~\(\alpha\), guided by the sets \(C_\eta\): start at \(\beta_0=\beta\), then let \(\beta_1=\min\{\xi\in C_{\beta_0}:\xi\geq\alpha\}\), \(\beta_2=\min\{\xi\in C_{\beta_1}:\xi\geq\alpha\}\), \dots, until \(\beta_n=\alpha\) for some~\(n=n(\alpha,\beta)\), which will happen of course. Thus one obtains 1)~the set \(\operatorname{Tr}(\alpha,\beta)=\{\beta_i:i\leq n\}\), the trace of the walk; 2)~the number \(n(\alpha,\beta)\) of steps from~\(\beta\) to~\(\alpha\); 3)~the sequence \(\rho_0(\alpha,\beta)=\langle k_i:i<n\rangle\) of finite ordinals from which~\(\operatorname{Tr}(\alpha,\beta)\) may be reconstructed: \(k_i=|C_{\beta_i}\cap\alpha|\), the position of~\(\beta_{i+1}\) in~\(C_{\beta_i}\). This list is not exhaustive but it will do for this review. Rather than giving an exhausting enumeration of all results I will try to convey its spirit by describing some of its highlights (in this reviewer's opinion). To begin, the function~\(\rho_0\) yields, without any real effort, an Aronszajn tree. First order \(\omega^{<\omega}\) lexicographically; this gives a dense countable order and for each~\(\beta\) the sequence \(s_\beta=\langle \rho_0(\alpha,\beta):\alpha<\beta\rangle\) is strictly increasing. The initial segments of these sequences form an Aronszajn tree; the level sets are automatically countable as are the branches. The lexicographic order of \(\{s_\beta:\beta\in\omega_1\}\) yields a linearly ordered set~\(C(\rho_0)\) whose square is the union of countably many chains. Under \(\roman{MA}_{\aleph_1}\) this order or its reverse embed into each uncountable linear order whose square is the union of countably many chains. This is one step on the way to \textit{J. T. Moore}'s result [Ann. Math. (2) 163, No. 2, 669--688 (2006; Zbl 1143.03026)] that, under the Proper Forcing Axiom, every uncountable linear order contains a copy of \(\omega_1\) or its reverse, of \(C(\rho_0)\) or its reverse, or of some \(\aleph_1\)-dense subset of the real line. A combination of \(\operatorname{Tr}\) and \(\rho_0\) yields an older result by the author [Acta Math. 159, 261--294 (1987; Zbl 0658.03028)]: a colouring \(c:[\omega_1]^2\to\omega_1\) such that \(c``[X]^2=\omega_1\) for \textit{each} uncountable subset~\(X\) of~\(\omega_1\) (an `anti-Ramsey theorem' for \(\omega_1\), so to speak). First put \(\Delta_0(\alpha,\beta)=\min\{\xi\leq\alpha: \rho_0(\xi,\alpha)\neq\rho_0(\xi,\beta)\}\) and then let \([\alpha\beta]=\min\bigl(\text{Tr} (\Delta_0(\alpha,\beta),\beta)\setminus\alpha\bigr)\). For each uncountable~\(X\) the set \(\{[\alpha\beta]:\alpha,\beta\in X\}\) contains a closed and unbounded subset, so that a partition of~\(\omega_1\) into \(\aleph_1\) many stationary sets yields the colouring~\(c\). The family of sequences \(\langle n(\xi,\alpha):\xi<\alpha\rangle\) derived from the number-of-steps function \(n(\alpha,\beta)\) is an example of a coherent-but-not-trivial object: it is coherent in the sense that \(|n(\xi,\alpha)-n(\xi,\beta)|\) is uniformly bounded whenever~\(\alpha<\beta\); it is non-trivial in the sense that each total function \(g:\omega_1\to\omega_1\) has infinite distance to at least one sequence. This idea is lifted to arbitrary regular cardinals and yields a way of recognizing non-trivial sequences \(\langle C_\alpha:\alpha<\theta\rangle\). The book is worth reading for these and related questions alone but there is more. Each of the functions \(\text{Tr}\), \(n\), \(\rho_0\) and others may be used to construct Aronszajn trees, ordered sets like~\(C(\rho_0)\) and so on. Under \(\roman{MA}_{\aleph_1}\) and related axioms these objects are, if not isomorphic, all equally canonical. For instance, the ordered sets all share the decomposability and minimality properties of~\(C(\rho_0)\). Likewise, in the case of Aronszajn trees, those constructed in the book are archetypal in the sense that one cannot expect to construct very dissimilar objects in~\(\roman{ZFC}\) alone. There is more, much more, that can be done using walks as described above and it goes without saying that the structure of the sequence \(\langle C_\alpha:\alpha<\theta\rangle\) determines the outcome of the constructions. Thus there is a whole chapter on square-sequences with refinements of older results and new applications. This book is simply recommended reading for anyone with an interest in uncountable constructions and Chapters~2 and~3 may even be considered essential reading. The latter chapter develops the function~\(\rho\), which is as close to a (natural) metric on~\(\omega_1\) as one can get. The basic properties of~\(\rho\) are that \(\rho(\alpha,\beta)\geq\max\rho_0(\alpha,\beta)\) and that it obeys the `triangle inequalities' \(\rho(\alpha,\gamma)\leq\max\{\rho(\alpha,\beta),\rho(\beta,\gamma)\}\) and \(\rho(\alpha,\beta)\leq\max\{\rho(\alpha,\gamma),\rho(\beta,\gamma)\}\). From~\(\rho\) one defines (again canonically) a Hausdorff gap, a Cohen-name for a Souslin tree and some rather rigid Banach spaces. I hope to have enticed the reader into at least picking up this book.