Around a Hurewicz's formula (Q408558)

The Hurewicz formula is the starting point of the present paper. W. Hurewicz stated the following theorem for dimension-lowering mappings: for a closed mapping \(f: X\to Y\) from a non-empty separable metric space \(X\) onto a separable metric space \(Y\) we have \(\dim X\leq\dim f +\dim Y\), where \(\dim f= \sup\{\dim(f^{-1}y): y\in Y\}\). In the paper under review the authors are interested in generalizing this theorem for topological variants \(d\) admitting integer resp. ordinal values or \(\infty\), and \(d(x)\neq\infty\) for a point \(x\). An additional generalization is applied by considering \(d({\mathcal A})f= \sup\{d(f^{-1} A): A\in{\mathcal A}\}\) for non-empty families \(\mathcal A\) of non-empty subsets of \(Y\). The following questions are asked: (A): What is the relationship between \(d({\mathcal A}_1)f\) and \(d({\mathcal A}_2)f\) for families \({\mathcal A}_1\) and \({\mathcal A}_2\)? (B): Let \(d\) and \({\mathcal A}\) be given. Is there an integer resp. ordinal valued function \(F\) such that \(d(X)\leq F(d({\mathcal A})f, d(Y))\)? For Question A transfinite topological invariants \(d\) are considered which can have the following properties: (P1) \(d(\bigoplus_{i=1,\dots, n}X_i)\leq\max\{d(X_i): i= 1,\dots, n\}\), (P2) \(d(\bigoplus_{i=1,2,3,\dots}X_i)= \infty\) whenever \(d(X_i)\geq i\) for \(i= 1,2,3,\dots\), (P3) For every integer \(n\geq 0\), \(d(\bigoplus\{X_\gamma: \gamma\in\Gamma\})\leq n\) whenever \(d(X_\gamma)\leq n\) for each \(\gamma\Gamma\). Mainly the collections \({\mathcal A}_0= \{\{y\}: y\in Y\}\}\), \({\mathcal A}_1= \{F: F\) is a closed discrete subset of \(Y\}\) and \({\mathcal A}_2= \{D: D\) is a discrete subset of \(Y\}\) are considered. For brevity \(d_if= d({\mathcal A}_i)f\) for \(i=1,2,3\). Clearly, \(d_0f\leq d_1f\leq d_2f\). For a compact space with (P1) we have \(d_0f= d_1f\). When \(d\) possesses (P1), (P2) and (P3) then i) if \(d_1f\neq\infty\) we have \(d_0f= d_1 f\). ii) if \(d_2 f\neq\infty\) then there exists a point \(y\in Y\) such that \(d(f^{-1}y)= d_0f= d_1 f=d_2f\). For the Smirnov compactum \(S^{\alpha+1}= S^\alpha\times I\) and the projection on the second factor \(\text{pr}_\alpha\) we see the following for a topological invariant \(d\) such that \(d(S^\alpha)= \alpha\), for \(\alpha<\omega_1\). i) if \(d\) possesses (P1) then \(d_1\text{pr}_\alpha=\alpha\), ii) if \(d\) possesses (P2) then \(d_2\text{pr}_\alpha=\infty\), for each \(\omega_0\leq\alpha<\omega_1\), iii) if \(d\) possesses (P3) then \(d_0\text{pr}_\alpha= d_1\text{pr}_\alpha= d_2\text{pr}_\alpha=\alpha\) for \(\alpha< \omega_0\). Another interesting example is when we have a one point extension \(X_*\) of the topological sum \(\bigoplus_{n=1,2,3,\dots}X_n\) of spaces such that \(d(X_n)=n\) with the extension point \(p\). In this case the closed map is the quotient mapping \(f_*\) which shrinks each subspace \(X_n\) to a point. The resulting space is \(Y_*\). For a topological invariant \(d\) we have i) if \(d\) possesses (P1) then \(d_1f_*= d_0 f_*= \max\{\omega_0, d(\{p\})\}\), ii) if \(d\) possesses (P2) then \(d_2 f_*=\infty\), iii) the restriction \(\psi\) of \(f_*\) on \(\bigoplus_{n=1,2,3,\dots} X_n\) has \(d_0\psi= \omega_0\) and if \(d\) possesses (P2), \(d_1\psi =\infty\). For Question B the main result is the generalization of the notion of fully closed mappings introduced by V. V. Fedorchuk. A closed mapping \(f: X\to Y\) is fully closed if for each pair of disjoint closed sets \(A\) and \(B\) of \(X\) the set \(f(A)\cap f(B)\) is discrete. We say that \(f\) has property \(Z\) iff \(f(A)\cap f(B)\) is a zero-set in \(Y\). M. G. Charalambous and J. Krzempek proved that if a mapping \(f: X\to Y\) is fully closed having property \(Z\) from a non-empty normal space \(X\) onto a space \(Y\) with \(\text{Ind}_0 Y<\infty\) then \[ \text{tr\,Ind}_0(X)\leq (\text{tr\,Ind}_0)_1 f+ \text{Ind}_0 Y. \] By our previous observations if also \((\text{tr\,Ind}_0)_1 f\neq\infty\) (resp. \(Y\) is compact) then \[ \text{tr\,Ind}_0 X\leq(\text{tr\,Ind}_0)_0 f+ \text{Ind}_0 Y. \] Generalization with respect to the dimension function is done by using the base dimension \(I(X,{\mathcal F})\) for a normal base \({\mathcal F}\) of \(X\) introduced by D. N. Georgiou, S. D. Illiadis and K. L. Kozlov for normal spaces. \(\text{Ind\,}X\), \(\text{Ind}_0X\) and the relative inductive dimension defined by Chigogidze can be seen as base dimensions resulting from a specific choise for a normal base \({\mathcal F}\). The main result of the paper is the following theorem: Let \(f: X\to Y\) be a closed mapping of a normal space onto a space \(Y\), and let \({\mathcal F}_X\) and \({\mathcal F}_Y\) be multiplicative perfectly normal bases on the spaces \(X\), \(Y\) resp. such that i) \(f^{-1}F\in{\mathcal F}_X\) for each \(F\in{\mathcal F}_Y\), ii) for any pair \(A\), \(B\) of disjoint elements of \({\mathcal F}_X\) we have \(C= f(A)\cap f(B)\in{\mathcal C}_Y\cap{\mathcal F}_Y\) and there exist elements \(F\), \(G\) of \({\mathcal F}_Y\) such that the sets \(F\cap(Y- C)\) and \(G\cap(Y- C)\) are disjoint and \(f(A)- C\subset F\cap(Y- C)\) and \(f(B)- C\subset G\cap(Y - C)\). Then \(I(X,{\mathcal F}_X)\leq I({\mathcal F}_X,{\mathcal C}_Y) f+ I(Y,{\mathcal F}_Y)+ 1\). Here \(I({\mathcal F}_X,{\mathcal C}_Y) f=\sup\{I(f^1C,{\mathcal F}_X\mid f^1(C): C\in{\mathcal C}_Y\cap{\mathcal F}_Y\}\). This general theorem has many applications in corollaries like: 3.1 \(\text{Ind}_0 X\leq \text{Ind}_0({\mathcal C}^0_Y)f+ \text{Ind}_Y+ 1\), where \({\mathcal C}^0_Y={\mathcal C}_Y\cap Z(Y)\). 3.2 \(\text{Ind}_0 X\leq fr+ \text{Ind}_0 Y+ 1\) when \(C= f(A)\cap f(B)\in Z(Y)\) and \(\text{Ind}_0(C)\leq r\), where \(fr= \sup\{\text{Ind}_0 f^{-1}D: D\in Z(Y)\) and \(\text{Ind}_0(D)\leq r\}\). 3.3 \(\text{Ind}_0 X\leq fp+ \text{Ind}_0 Y+ 1\), \(\text{Ind}_0 Y<\infty\) when \(C= f(A)\cap f(B)\in Z(Y)\) and \(C\) is countable (resp. compact, Lindelöf, paracompact, etc.), where \(fp= \sup\{\text{Ind}_0 f^{-1}D: D\in Z(Y)\) and \(D\) is countable (resp. compact, Lindelöf, paracompact, etc.)\}. 3.4 \(\text{Ind\,}X\leq \text{Ind}({\mathcal C}_Y)f+ \text{Ind\,}Y+ 1\), when \(\text{Ind\,}Y<\infty\), \(Y\) is perfectly normal and when \(A\) and \(B\) are disjoint elements of \(Z(X)\) we have \(C= f(A)\cap f(B)\in{\mathcal C}_Y\) and \(f^{-1}C\) is perfectly normal. 3.5 \(\text{Ind\,}X\leq(\text{Ind})_0 f+ \text{Ind\,}Y+ 1\), when \(\text{Ind\,}Y<\infty\), \(Y\) is perfectly normal and when \(A\) and \(B\) are disjoint elements of \(Z(X)\) we have \(C= f(A)\cap f(B)\) is countable and \(f^{-1}y\) is perfectly normal for each \(y\in Y\). The question remains to sharpen this result by removing the \(+1\). Observe that the above result forms a generalization of the notion of fully closed mappings. V. V. Chatyrko constructed a fully closed mapping \(f_{CH}\) from a compact space \(X\) with \(\text{Ind\,}X= 2\) to a compact space \(A\) with \(\text{Ind\,}A= 0\) such that \((\text{Ind})_0f_{CH}= 1\) showing that the Hurewicz formula does not hold in general.