A note on a sum theorem for dimension \(\mathcal K\)-Ind (Q409656)

In [Spaces which are weakly infinite-dimensional modulo simplicial complexes, Vestn. Mosk. Univ., Ser. I 2009, No. 3, 33--40 (2009)] the author developed a new inductive dimension function \({\mathcal K}\)-Ind, where \({\mathcal K}\) is a nonempty set of finite simplicial complexes.NEWLINENEWLINE Before its definition we need some auxiliary notions. For a space \(X\) the set of all closed subsets of \(X\) is called \(\text{Exp\,}X\) and we let \(\text{Fins}(\text{Exp\,}X)\) be the set of finite sequences of members of \(\text{Exp\,}X\). Complexes \(K\) are finite complete simplicial complexes. We put \(\text{Exp}_K(X)= \{\Phi\in\text{Fin}_s(\text{Exp\,}X)\): there is an embedding of the nerve \(N(\Phi)\subset K\}\).NEWLINENEWLINE For \(\Phi= (F_1,\dots, F_m)\in\text{Fin}_s(\text{Exp\,} X)\) a sequence \(u=(U_1,\dots, U_k)\), \(k\geq m\), is called a \(K\)-neighbourhood of \(\Phi\) if \(F_j\subset U_j\) and there is an embedding of the nerve \(N(u)\subset K\). A set \(P\subset X\) is said to be a \(K\)-partition of \(\Phi\) (notation: \(P\in\text{Part}(\Phi,K)\)) if \(P= X-\cup u\) for some \(K\)-neighbourhood \(u\) of \(\Phi\).NEWLINENEWLINE The dimension \({\mathcal K}\)-Ind is defined as follows. Let \({\mathcal K}\) be a non-empty set of complexes. To every space \(X\) one assigns the dimension \({\mathcal K}\)-\(\text{Ind\,}X\) which is an integer \(\geq-1\) or \(\infty\).NEWLINENEWLINE \({\mathcal K}\)-\(\text{Ind}= -1\) iff \(X= \varnothing\);NEWLINENEWLINE \({\mathcal K}\)-\(\text{Ind\,}X\leq n\), for some \(n= 0,1,\dots\), if for every \(K\in{\mathcal K}\) and \(\Phi\in\text{Exp}_K(X)\) there exists a partition \(P\in\text{Part}(\Phi,K)\) such that \(K\)-\(\text{Ind\,}P\leq n-1\);NEWLINENEWLINE \({\mathcal K}\)-\(\text{Ind\,}X=\infty\) iff \({\mathcal K}\)-\(\text{Ind\,}X> n\), for all \(n\geq -1\).NEWLINENEWLINE If the set \({\mathcal K}\) contains only one complex \(K\), we write \({\mathcal K}= K\) and \({\mathcal K}\)-\(\text{Ind}=K\)-Ind. We have that for a space \(X\), \([0,1]\)-\(\text{Ind\,}X=\text{Ind\,}X\). And in [Colloq. Math. 120, No. 2, 223--247 (2010; Zbl 1222.54035)] the author proved that \({\mathcal K}\)-\(\text{Ind\,}X= \text{Ind\,}X\) if and only if \({\mathcal K}\) contains a disconnected complex \(K\).NEWLINENEWLINE One of the main questions concerning the dimension \({\mathcal K}\)-Ind is that for perfectly normal spaces the finite closed sum theorem is still open. This is the main issue of the paper.NEWLINENEWLINE First (Theorem 3.1), the Finite Dowker Theorem is proven. Let \(Y\) be a closed subspace of a hereditarily normal space \(X\) such that \({\mathcal K}\)-\(\text{Ind\,}Y\leq n\) and \({\mathcal K}\)-\(\text{Ind\,}X-Y\leq n\). Then also \({\mathcal K}\)-\(\text{Ind\,}X\leq n\).NEWLINENEWLINE The following properties are used to obtain the main result:NEWLINENEWLINE \((\mu_0)\) For each subspace \(Y\) of \(X\) and every open subspace \(U\) of \(Y\), if \({\mathcal K}\)-\(\text{Ind\,}Y\leq n\), then \({\mathcal K}\)-\(\text{Ind\,}U\leq n\),NEWLINENEWLINE \((\mu^0_n)\) For each subspace \(Y\) of \(X\) and every open \(F_\sigma\)-subspace \(U\) of \(Y\), if \({\mathcal K}\)-\(\text{Ind\,}Y\leq n\), then \({\mathcal K}\)-\(\text{Ind\,}U\leq n\),NEWLINENEWLINE \((\sigma_n)\) For each subspace \(Y\) of \(X\) and every pair \(Y_1\), \(Y_2\) of closed subspaces of \(Y\) such that \(Y= Y_1\cup Y_2\), if \({\mathcal K}\)-\(\text{Ind\,}Y_i\leq n\) for \(i= 1,2\), then \({\mathcal K}\)-\(\text{Ind\,}Y\leq n\).NEWLINENEWLINE By Theorem 3.1 we have \((\mu_n)\to (\sigma_n)\) for hereditarily normal spaces. For hereditarily normal spaces the relation \((\sigma_n)\to (\mu^0_n)\) is proven in the paper. Since for perfectly normal spaces \((\mu_n)\) and \((\mu^0_n)\) are clearly equivalent, we obtain the main result:NEWLINENEWLINE Let \(X\) be a perfectly normal space. Then the finite closed sum theorem for \({\mathcal K}\)-Ind holds if and only if \({\mathcal K}\)-Ind is monotonic in \(X\).NEWLINENEWLINE The main question however remains open: Does the finite closed sum theorem hold in general for perfectly normal spaces and arbitrary \({\mathcal K}\)?