A new approach to dimension (Q752534)

The author proposes a new topological dimension theory based on the philosophical constraint that a continuous injective mapping from one space to another ought not decrease dimension. In particular, dimension dm is defined so that any separable metric space X which admits a continuous injective mapping into the Cantor set must have \(dm(X)=0\). Therefore, any totally disconnected space S is required to have \(\dim (S)=0\). Although agreeing on compact metric spaces, this distinguishes the author's theory from both small and large inductive dimensions and covering dimension as there exist totally disconnected separable metric spaces of all dimensions [\textit{L. Rubin}, \textit{R. M. Schori} and \textit{J. J. Walsh} (preprint 1979), see also Schori and Walsh in Topology Proc. 3, No.2, 495-506 (1979; Zbl 0442.54024)]. The author uses earlier examples due to Kuratowski and Erdős to establish this distinction. The author's theory is shown to possess many of the usual properties which are commonly held to define a topological dimension theory; topological invariance, monotonicity of dimensions to subspaces, sum theorems and the logarithmic inequality for topological products. However, the author has failed to prove or disprove the analogous inequality to \(ind(X\cup Y)\leq ind(X)+ind(Y)+1\) for pairs of separable metric spaces. Therefore, while it is shown that any separable metric space Y with \(dm(X)=n\) can be written as the disjoint union \(\cup \{Z_ i:\;i=0,...,n\}\) where \(\dim (Z_ i)\leq 0\) for each i, the converse has not been established.