On dimensional properties of the generalized Smirnov's spaces (Q2268628): Difference between revisions

M.G. Charalambous (1997) introduced the transfinite dimensions modulo \(\mathcal P ,\mathcal P\)-trind and \(\mathcal P\)-trInd, by starting the induction with \(\mathcal P\)-trind \(X\) \((\mathcal P\)-trInd \(X\) ) = -1 when \(X \in \mathcal P\) instead of \(X = \varnothing\). The main objective of the paper under review is to show a series of counterexamples such that \(\mathcal P\)-trind \(X\) and \(\mathcal P\)-trInd \(X\) are different. First some useful sum theorems are proven. We call a decomposition \(X = F \cup (\bigcup_{i=1,2,\dots} E_i)\) A-special when for each \(i\) the set \(E_i\) is clopen in \(X\); B-special when also \(\lim i \to \infty\) diam\((E_i) =0\) and C-special when each closed set \(G\) disjoint from \(F\) has \(G \subset \bigcup_{ i=1,2,\dots,n} E_i\) for some \(n\). When a metric space admits a B-special decomposition then \[ {\mathcal P}-\text{trind}\, X \leq \max\{{\mathcal P}-\text{trind}\, F, \text{sup}_{ i=1,2,\dots}\{{\mathcal P}-\text{trind} E_i \}\}. \] If the composition is also C-special and \(F\) compact then also \[ {\mathcal P}-\text{trInd}\, X \leq \max \{{\mathcal P}-\text{trInd}\, F, \text{sup}_{ i=1,2,\dots} \{{\mathcal P}-\text{trInd}\, E_i \}\}. \] When a space \(X\) is A- and C-special with \(\dim F =n\), then we can find B-special spaces \(Z_k = F \cup (\bigcup_{ i=1,2,\dots} E_{i,k})\), for \(k=1,\dots,n+1\), such that \(X = \bigcup_{ k=1,\dots,n+1} Z_k\). To construct the generalized Smirnov Space \(S_Y^\alpha\) we have for \(\alpha = 0\) \(S_Y^\alpha = Y\), for a non-limit ordinal \(\alpha\) we have \(S_Y^\alpha = S_Y^{\alpha-1} \times I\) and for limit ordinals \(\alpha\) the space \(S_Y^\alpha\) is the one-point compactification of the topological sum \(\bigoplus_{\beta < \alpha} S_Y^\beta\). We consider for the class \(\mathcal P\) the additive Borel class \(A(\beta)\) or the multiplicative Borel class \(M(\beta)\). For each countable ordinal \(\beta\) spaces \(P_\beta \in A(\beta)\) and \(Q_\beta \in M(\beta)\) are defined. Let us denote for every ordinal \(\alpha = \lambda(\alpha) + n(\alpha)\), where \(\lambda(\alpha)\) is a limit ordinal and \(n(\alpha)\) an integer. For the space \(S_Y^\alpha\) such that \(Y = P_\beta (Y = Q_\beta)\) we have \(\mathcal P\)-trInd \(S_Y^\alpha \leq \alpha\) and we can decompose it in closed subspaces \(Z_i , i=1,\dots,n(\alpha) +1\) such that \(\mathcal P\)-trind \(Z_i \leq\) \(\mathcal P\)-trInd \(Z_i \leq \lambda(\alpha)\) for \(\mathcal P = A(\beta) (\mathcal P = M(\beta))\). We use the following quite strong sum theorem for \(\mathcal P\)-trind earlier proven by the first author which states that when \(X\) is the union of two closed subspaces \(X_1\) and \(X_2\) such that \(\mathcal P\)-trind \(X_1 \leq \alpha_1\) and \(\mathcal P\)-trind \(X_2 \leq \alpha_2\) and \(\alpha_1 \leq \alpha_2\) then \(\mathcal P\)-trind \(X \leq \alpha_2\) when \(\lambda(\alpha_1) < \lambda(\alpha_2)\) and \(\mathcal P\)-trind \(X \leq \alpha_2+1\) when \(\lambda(\alpha_1) = \lambda(\alpha_2)\). From this and the mentioned decomposition of \(S_Y^\alpha\) the main result of this paper follows. Let \(m\) be an integer such that \(0 \leq n(\alpha) \leq 2^m-1\). Then for \(Y = P_\beta (Y = Q_\beta)\) for \(\mathcal P = A(\beta) (\mathcal P = M(\beta))\) we have \(\mathcal P\)-trind \(S_Y^\alpha \leq \lambda(\alpha)+ m\), and moreover when \(n(\alpha) \geq 3\) we have \(\mathcal P\)-trind \(S_Y^\alpha < \mathcal P\)-trInd \(S_Y^\alpha\).