Prime ends for domains in metric spaces (Q387898)

A century ago, \textit{C. Carathéodory} defined his ends and prime ends [Math. Ann. 73, 323--370 (1913; JFM 44.0757.02)]. In a simply connected domain of the plane, Jordan arcs were used to define cross-cuts and their chains. Then, the concept of division of a chain by another chain created an equivalence relation that led to ends and, divisible only by themselves, prime ends. Under conformal mappings, prime ends of the image of the unit circle correspond to the points of the unit circle. With the Dirichlet boundary value problem in view, the authors of the paper under review define more general prime ends in metric spaces. The prime end boundary thus introduced is suitable for dealing with the Dirichlet problem in a less restrictive setting. While these new prime ends work well for finitely connected planar domains, as the Carathéodory prime ends did, it is worth noting that they are not the same. This lengthy paper provides an extensive comparison of the two concepts, as well as the study of modulus ends, singleton prime ends and the \textit{S. Mazurkiewicz} boundary [Fundam. Math. 33, 177--228 (1945; Zbl 0060.40009)] as it relates to the prime end boundary. In the same spirit, the concept of John domains [\textit{A. Ancona}, Publ. Mat., Barc. 51, No. 2, 345--396 (2007; Zbl 1134.31009)] is generalized to the concept of ``almost John domains'' in which prime ends are duly investigated. The gist of the authors definition of a prime end that makes it differ from Carathéodory's one is the abandonment of connectedness at the cross-cut stage of the definition. Where a cross-cut divides the domains into exactly two components, the ``acceptable set'' defined by the authors may create more than two components. While an acceptable set is connected, its relative boundary may not be and, consequently, not every Carathéodory prime end is a prime end in the authors' sense. On the other hand, this new theory does not recover Carathéodory prime ends in the simply connected planar case. A reader interested in the Dirichlet problem and the potential theory in general may find this and a few of the upcoming papers [\textit{A. Björn}, ``The Dirichlet problem for \(p\)-harmonic functions on the comb'' (in preparation); \textit{A. Björn} and \textit{J. Björn}, ``Obstacle and Dirichlet problems on arbitrary non open sets in metric spaces, and fine topology'', Preprint, \url{arXiv:1208.4913}; \textit{A. Björn}, \textit{J. Björn} and \textit{N. Shanmugalingam}, ``The Dirichlet problem for \(p\)-harmonic functions with respect to the Mazurkiewicz boundary, and new capacities'', Preprint, \url{arXiv: 1302.3887}; ``The Mazurkiewicz distance and sets which are finitely connected at the boundary'' (in preparation)] useful. The paper under review contains an interesting question: Is every end divisible by a prime end? The answer is yes for Carathéodory ends.