The dimension Dind of finite topological \(T_0\)-spaces (Q2144650)

A. V. Archangelskii introduced the dimension Dind and the properties of this function have been discussed for various classes of spaces. In this paper the authors proceed with the study of finite \(\mathrm{T}_0\)-spaces and the behaviour of Dind on this class. The results are surprising and are not in line with the results for the class of normal Hausdorff spaces, where we have \(\operatorname{Ind}\leq\operatorname{Dind}\) and for transfinite values we can even have \(\operatorname{Ind} < \operatorname{Dind}\) for a compactum. A space has \(\operatorname{Dind} X\leq k\) iff for every finite cover \(\mathcal{U}\) there exists a family \(\mathcal{V}\) of pairwise disjoint open sets such that \(\mathcal{V} < \mathcal{U}\) and \(\dim(X\backslash\bigcup\{V : V\in \mathcal{V}\})\leq k-1\) and \(\operatorname{Dind} X = -1\) when \(X=\emptyset\). The cover \(\mathcal{V}\) is called a partition of \(X\). We recall that a space \(X\) is \(\mathrm{T}_0\) when for any two distinct points \(x\) and \(y\) in \(X\) we have an open set \(U\) such that \(x\in U\) and \(y\in U\) or we have \(x\notin U\) and \(y\notin U\). In the paper the minimal open sets \(U_x\) which contain a point \(x\) in a finite space \(X\) play an important role and simplify many proofs. For each integer \(n\) a finite \(\mathrm{T}_0\) space \(X\) is shown to exist such that \(\operatorname{Dind} X = n\). The authors show many examples. In the relation with the large inductive dimension, \(\operatorname{Ind}\), the relation \(\operatorname{Dind} \leq \operatorname{Ind}\) holds for finite \(\mathrm{T}_0\)-spaces and such spaces are constructed for each integer \(n\) such that \(\operatorname{Dind} X = 1\) and \(\operatorname{Ind} X = n\). Moreover, for every \(m\) and \(n\) such that \(m \leq n\) a finite \(\mathrm{T}_0\)-space is constructed such that \(\operatorname{Dind} X = m\) and \(\operatorname{Ind} X = n\). We have \(\operatorname{Dind} X = 0\) if and only if \(\operatorname{Ind} X = 0\). For the small inductive dimension, \(\operatorname{ind}\), we have \(\operatorname{Ind} \leq \operatorname{ind}\) for finite \(\mathrm{T}_0\)-spaces, so we have \(\operatorname{Dind} \leq \operatorname{ind}\). Finite \(\mathrm{T}_0\)-spaces are constructed for each integer \(n\) such that \(\operatorname{Dind} X = 0\) and \(\operatorname{ind} X = n\). For the covering dimension, \(\dim\), we have \(\operatorname{Ind} \leq \dim\) for finite \(\mathrm{T}_0\)-spaces, so we have \(\operatorname{Dind} \leq \dim\). Finite \(\mathrm{T}_0\)-spaces are constructed for each integer \(n\) such that \(\operatorname{Dind} X = \operatorname{Ind} X = \operatorname{ind} X = \dim X = n\) and other examples of spaces with \(n\) elements such that \(\operatorname{Dind} X = \operatorname{Ind} X = \operatorname{ind} X = 1\) and \(\dim X = n-2\). We may construct a \(\mathrm{T}_0\)-space \(X\) by two disjoint copies of a base space \(Y\) and a third point \(\omega\) connecting the copies. Then \(d(X) = d(Y) +1\) for \(d=\operatorname{Dind}, \operatorname{Ind}, \operatorname{ind}\) and \(\dim\). The research concludes with a sum theorem which states that when \(X = X_1\cup X_2\), for two closed subspaces \(X_1\) and \(X_2\), then \(\operatorname{Dind} X \leq \operatorname{Dind} X_1 \cap X_2 + \max\{\operatorname{Dind} X_1 , \operatorname{Dind} X_2\}\). The authors conclude with the question if there exists an algorithm to compute the dimension Dind for any finite \(\mathrm{T}_0\)-space. One could suggest the algorithm of stating that Dind is the number of steps of eliminating the singleton sets \(\pm 1\). However this is not true as an example shows. This is one of the few questions the authors leave open after their complete study of Dind for finite \(\mathrm{T}_0\)-spaces.