Abstract Generated abstract
This paper studies preservation and characterization properties of Hewitt \(Q\)-spaces, also called functionally closed spaces, in general topology. It proves that a perfect image of a normal countably paracompact \(Q\)-space is again normal, countably paracompact, and a \(Q\)-space, giving a partial answer to a question on invariance under perfect mappings. The paper also characterizes \(Q\)-spaces among normal countably paracompact spaces by maximal countably centered systems of closed sets, and establishes several inverse image theorems for continuous or closed mappings whose fibers are \(Q\)-spaces. These results extend or refine earlier work of Ponomarev and Pasynkov on functionally closed spaces and mappings.
Full Text
UDC 519.50+519.54
MATHEMATICS
P. KENDEROV
ON \(Q\)-SPACES
(Presented by Academician P. S. Aleksandrov on 4 X 1966)
E. Hewitt, in the paper \((^{1})\), defined the class of \(Q\)-spaces. Most often these spaces are called functionally closed.
Definition. A completely regular space \(X\) is called a \(Q\)-space (functionally closed) if every maximal countably centered system of zero-sets has a nonempty intersection. Here a system \(\xi=\{Z_\alpha\}\) is called countably centered if every countable subsystem of this system of sets has a nonempty intersection. A set \(A\subseteq X\) is called a zero-set if on the space \(X\) there exists a continuous function such that the set \(A\) is the complete preimage of zero.
Theorem 1. A perfect image of a normal countably paracompact \(Q\)-space \(X\) is always normal, countably paracompact, and is a \(Q\)-space.
Proof. Let \(f:X\to Y\) be a perfect mapping, and let \(X\) be a normal countably paracompact \(Q\)-space. Then the space \(Y\) is normal and countably paracompact (for this it is sufficient only that the mapping \(f\) be closed). We shall prove that \(Y\) is a \(Q\)-space. Let \(\xi=\{Z_\alpha\}\) be an arbitrary maximal countably centered system of zero-sets of the space \(Y\). This means that in the space \(Y\) there exist continuous functions \(\varphi_\alpha(y)\) such that \(Z_\alpha=\varphi_\alpha^{-1}(0)\). Then \(\psi_\alpha(x)=\varphi_\alpha(fx)\) is a continuous function on the space \(X\), and
\[
\psi_\alpha^{-1}(0)=f^{-1}(\varphi_\alpha^{-1}(0))=f^{-1}Z_\alpha,
\]
therefore the system
\[
\eta=f^{-1}\xi=\{f^{-1}Z_\alpha\}
\]
is countably centered and consists of zero-sets of the space \(X\). The system \(\eta\), generally speaking, is not a maximal countably centered system of zero-sets. Complete it to a maximal countably centered system \(\hat{\eta}\) of zero-sets of the space \(X\). We shall prove that in our case the system obtained, \(\eta\), will be countably centered.
Consider an arbitrary countable subsystem \(\hat{\eta}_0=\{\Phi_{\lambda_i}\}\subseteq\hat{\eta}\) of the system \(\hat{\eta}\). Without loss of generality we may assume that
\[
\Phi_{\lambda_1}\supseteq \Phi_{\lambda_2}\supseteq \cdots \supseteq \Phi_{\lambda_i}\supseteq \cdots,
\]
since the system \(\hat{\eta}\), together with any two of its elements, also contains their intersection. Consider the system
\[
f\hat{\eta}_0=\{f\Phi_{\lambda_i}\}
\]
of subsets closed in \(Y\). Since for arbitrary \(i,\alpha\) one has
\[
\Phi_{\lambda_i}\cap f^{-1}Z_\alpha\ne\Lambda,
\]
it follows that
\[
f\Phi_{\lambda_i}\cap Z_\alpha\ne\Lambda
\]
for all \(i,\alpha\). We shall now prove that
\[
\bigcap_{i=1}^{\infty} f\Phi_{\lambda_i}\ne\Lambda.
\]
Suppose the contrary; let
\[
\bigcap_{i=1}^{\infty} f\Phi_{\lambda_i}=\Lambda.
\]
Then, by virtue of the normality and countable paracompactness of the space \(Y\) (see \((^{2})\)), there exist neighborhoods \(Of\Phi_{\lambda_i}\) of the sets \(f\Phi_{\lambda_i}\) for which
\[
\bigcap_{i=1}^{\infty} Of\Phi_{\lambda_i}=\Lambda.
\]
Further, by Urysohn’s lemma, there exist zero-sets \(Z_i\) of the space \(Y\) such that
\[
f\Phi_{\lambda_i}\subseteq Z_i\subseteq Of\Phi_{\lambda_i}.
\]
Since
\[
f\Phi_{\lambda_i}\cap Z_\alpha\ne\Lambda
\]
for all \(i,\alpha\), the zero-set \(Z_i\supseteq f\Phi_{\lambda_i}\),
system \(\xi=\{Z_\alpha\}\) is a maximal countably centered system of zero-sets, then \(Z_i\in\xi\), \(i=1,2,\ldots\). Since \(Z_i\subset Of\Phi_{\lambda_i}\), \(\bigcap_{i=1}^{\infty} Of\Phi_{\lambda_i}=\Lambda\), it follows that \(\bigcap_{i=1}^{\infty} Z_i=\Lambda\), and we have arrived at a contradiction with the countable centeredness of the system \(\xi\). Thus, necessarily \(\bigcap_{i=1}^{\infty} f\Phi_{\lambda_i}\ne\Lambda\). Let \(y_0\in \bigcap_{i=1}^{\infty} f\Phi_{\lambda_i}\) be an arbitrary point.
Consider the countable system \(\{f^{-1}y_0\cap \Phi_{\lambda_i}\}_{i=1}^{\infty}\) of nonempty closed subsets of the bicompactum \(f^{-1}y_0\). Since \(\Phi_{\lambda_1}\supseteq \Phi_{\lambda_2}\supseteq\cdots\supseteq\Phi_{\lambda_i}\supseteq\cdots\), we have
\[
\bigcap_{i=1}^{\infty}(f^{-1}y_0\cap \Phi_{\lambda_i})\ne \Lambda,
\]
and hence the system \(\hat\eta\) is countably centered. Thus, we have extended our countably centered system \(\eta=f^{-1}\xi=\{f^{-1}Z_\alpha\}\) of zero-sets of the space \(X\) to a maximal, also countably centered, system \(\hat\eta\) of zero-sets. In view of the fact that \(X\) is a \(Q\)-space, necessarily
\[
\bigcap_{\Phi_\lambda\in\hat\eta}\Phi_\lambda\ne\Lambda.
\]
Then \(\bigcap_\alpha f^{-1}Z_\alpha\ne\Lambda\), and consequently also
\[
\bigcap_\alpha Z_\alpha\ne\Lambda,
\]
i.e. \(Y\) is a \(Q\)-space. Theorem 1 is proved.
Remark 1. In the proof of Theorem 1 we used only the countable compactness of \(f^{-1}y_0\).
Remark 2. This theorem is a partial answer to a question of V. I. Ponomarev from (4): is functional closedness of a normal space preserved under a perfect mapping? In Frolik’s paper (3) a theorem is formulated which gives a complete answer to the question, but the proof of this theorem contains a gap.
By the same method one also proves the following known result (see (4)).
Proposition (V. I. Ponomarev). An open perfect image of a normal \(Q\)-space is always a \(Q\)-space.
In proving this proposition the following assertion is used:
Lemma 1. The image \(f\Phi\) of a zero-set \(\Phi\) in a space \(X\) under an open perfect mapping \(f:X\to Y\) is a zero-set of the space \(Y\).
Theorem 2. Let \(X\) be a normal countably paracompact space. In order that \(X\) be a \(Q\)-space, it is necessary and sufficient that every maximal countably centered system of closed (not necessarily zero-) sets have nonempty intersection.
Proof. Necessity. Let \(\xi=\{F_\lambda\}\) be a maximal countably centered system of closed sets. Denote by \(\xi_0\) the collection of all zero-sets belonging to the system \(\xi\). Clearly, \(\xi_0\) is countably centered. We show that it is maximal.
Let \(Z\) be a zero-set which meets all elements of the system \(\xi_0=\{Z_\lambda\}\). If \(Z\notin\xi=\{F_\lambda\}\), then there is \(F_{\lambda_0}\in\xi\), \(F_{\lambda_0}\cap Z=\Lambda\). By the normality of \(X\) there exists a zero-set \(Z_0\) such that \(Z_0\supseteq F_{\lambda_0}\), \(Z_0\cap Z=\Lambda\). But since \(\xi\) is a maximal countably centered system, \(Z_0\in\xi\); consequently, \(Z_0\in\xi_0\), and then \(Z_0\cap Z\ne\Lambda\) (by the choice of \(Z\)). Thus \(Z\in\xi\), and consequently \(Z\in\xi_0\). Since \(X\) is a \(Q\)-space, we have
\[
\bigcap_{Z_\lambda\in\xi_0} Z_\lambda\ne\Lambda.
\]
Furthermore, any closed subset \(F\subseteq X\), in particular \(F_\lambda\in\xi\), is representable as an intersection of zero-sets, i.e. \(F_\lambda=\bigcap_\alpha Z_\alpha^\lambda\); \(Z_\alpha^\lambda\in\xi_0\). Then
\[
\bigcap_\lambda F_\lambda
=
\bigcap_\lambda\left(\bigcap_\alpha Z_\alpha^\lambda\right)
=
\bigcap_{Z\in\xi_0},
\]
which proves the necessity.
Sufficiency. Let \(\xi_0=\{Z_\lambda\}\) be a maximal countably centered-
system of zero-sets of the space \(X\). Extend it to a maximal simply centered system \(\xi=\{F_\lambda\}\) of closed sets. We shall prove the countable centeredness of the system \(\xi\). Let \(\{F_{\lambda_i}\}\) be some countable subsystem of the system \(\xi\). We must prove that
\[
\bigcap_{i=1}^{\infty} F_{\lambda_i}\ne \Lambda .
\]
Suppose the contrary, let
\[
\bigcap_{i=1}^{\infty} F_{\lambda_i}=\Lambda .
\]
Then, by normality and countable paracompactness (see \((^2)\)) of the space \(X\), there exist neighborhoods \(OF_{\lambda_i}\) of the closed sets \(F_{\lambda_i}\), for which also
\[
\bigcap_{i=1}^{\infty} OF_{\lambda_i}=\Lambda .
\]
Moreover, there exist zero-sets \(Z_i\) such that
\[
F_{\lambda_i}\subseteq Z_i\subseteq OF_{\lambda_i}.
\]
Since \(F_{\lambda_i}\in\xi\), it follows that \(Z_i\in\xi\), and consequently \(Z_i\in\xi_0\). Therefore, on the one hand,
\[
\bigcap_{i=1}^{\infty} Z_i\subset \bigcap_{i=1}^{\infty} OF_{\lambda_i}\ne \Lambda,
\]
and on the other hand, \(Z_i\in\xi_0\), i.e.
\[
\bigcap_{i=1}^{\infty} Z_i\ne \Lambda
\]
by the countable centeredness of the system \(\xi_0\). We have obtained a contradiction. Hence the system \(\xi=\{F_\lambda\}\) is countably centered. Then
\[
\bigcap_\lambda F_\lambda\ne \Lambda,
\]
and consequently, a fortiori,
\[
\bigcap_{Z\in\xi_0} Z\ne \Lambda,
\]
i.e. the space \(X\) is a \(Q\)-space, which proves sufficiency, and with it the whole theorem.
Remark. In proving necessity we did not use the countable paracompactness of \(X\).
Theorem 3. Let \(f:X\to Y\) be a continuous mapping of a normal space \(X\) onto a hereditarily finally compact space \(Y\), such that the preimage \(f^{-1}y\) of every point \(y\in Y\) is a \(Q\)-space. Then \(X\) is a \(Q\)-space.
Proof. Denote by \(\xi=\{Z_\lambda\}\) some arbitrary maximal countably centered system of zero-sets of the space \(X\). We must prove that
\[
\bigcap_\lambda Z_\lambda\ne \Lambda .
\]
Consider the system
\[
f\xi=\{fZ_\lambda\}
\]
of sets in \(Y\). The system \(f\xi\) is also countably centered. We shall show that
\[
\bigcap_\lambda fZ_\lambda\ne \Lambda .
\]
By the final compactness of the space \(Y\) we shall have
\[
\bigcap_\lambda [fZ_\lambda]_Y\ne \Lambda,
\]
where \([fZ_\lambda]_Y\) is the closure of the set \(fZ_\lambda\) in \(Y\). Moreover, observe that
\[
\bigcap_\lambda [fZ_\lambda]_Y
\]
consists of not more than one point.
Let
\[
\bigcap_\lambda [fZ_\lambda]=y_0 .
\]
Fix some index \(\lambda_0\) and consider the system
\[
\eta_{\lambda_0}=\{fZ_{\lambda_0}\cap fZ_\lambda\}.
\]
This system is countably centered and consists of sets belonging to the finally compact space
\[
fZ_{\lambda_0}\subseteq Y.
\]
Therefore
\[
A=\bigcap_\lambda \bigl([fZ_{\lambda_0}\cap fZ_\lambda]_{fZ_{\lambda_0}}\bigr)\ne \Lambda .
\]
But
\[
A\subseteq \bigcap_\lambda [fZ_{\lambda_0}\cap fZ_\lambda]_Y
\subseteq \bigcap_\lambda \bigl([fZ_{\lambda_0}]_Y\cap [fZ_\lambda]_Y\bigr)
\subseteq \bigcap_\lambda [fZ_\lambda]_Y=y_0 .
\]
It follows (by the arbitrariness of \(\lambda_0\)) that
\[
y_0=\bigcap_\lambda fZ_\lambda .
\]
Now consider the system
\[
\{f^{-1}y_0\cap Z_\lambda\}=\xi_{y_0}.
\]
This system is countably centered and consists of zero-sets in \(f^{-1}y_0\). We shall prove its maximality.
Let \(Z_0\subseteq f^{-1}y_0\) be a zero-set of some function \(\varphi\) defined on \(f^{-1}y_0\). By normality of the space \(X\) (Urysohn’s theorem) there exists an extension \(\widetilde{\varphi}\) of the function \(\varphi\) to the whole space \(X\).
Consider the zero-set \(Z(\widetilde{\varphi})\) of the function \(\widetilde{\varphi}\). We have
\[
Z(\widetilde{\varphi})\cap f^{-1}y_0=Z_0.
\]
Since \(y_0\) is a zero-set in \(Y\), \(f^{-1}y_0\) is a zero-set in \(X\). Therefore \(Z_0\) is a zero-set in all of \(X\) as the intersection of the zero-sets \(f^{-1}y_0\) and \(Z(\widetilde{\varphi})\). Let
\[
Z_0\cap (f^{-1}y_0\cap Z_\lambda)\ne \Lambda
\]
for every \(\lambda\). Then
\[
Z_0\cap Z_\lambda\ne \Lambda
\]
for all \(\lambda\), i.e. \(Z_0\in\xi=\{Z_\lambda\}\), and then \(Z_0\in\{f^{-1}y_0\cap Z_\lambda\}\), \(Z_0\subseteq\)
\(\subseteq f^{-1}y_0\), whereby the maximality of the system \(\xi_{y_0}\) is proved. Thus, \(\xi_{y_0}\) is a maximal countably centered system of zero-sets of the subspace \(f^{-1}y_0\). From the fact that \(f^{-1}y_0\) is a \(Q\)-space, it follows that
\[
\bigcap_\lambda (f^{-1}y_0 \cap Z_\lambda) \ne \Lambda,
\]
and then all the more
\[
\bigcap_\lambda Z_\lambda \ne \Lambda,
\]
which proves the theorem.
Remark. Theorem 3 strengthens B. A. Pasynkov’s theorem from \((^5)\). In B. A. Pasynkov’s theorem the space \(Y\) has a countable base.
Theorem 4. Let \(f:X\to Y\) be a closed continuous mapping of a normal space \(X\) onto a \(Q\)-space \(Y\). If every point \(y\in Y\) has type \(G_\delta\) in \(Y\) and its inverse image \(f^{-1}y\) is a \(Q\)-space, then \(X\) is also a \(Q\)-space.
The proof of Theorem 4 is simple, and we omit it.
Theorem 5. Let \(f:X\to Y\) be a closed continuous mapping of a countably paracompact normal space \(X\) onto a \(Q\)-space \(Y\), under which the inverse image \(f^{-1}y\) of every point \(y\in Y\) is functionally closed. Then the space \(X\) is also functionally closed.
Proof. Since the space \(X\) is normal and countably paracompact, by virtue of Theorem 2 it suffices for us to prove that every maximal countably centered system of closed (not necessarily zero-) sets has a nonempty intersection. Let \(\xi=\{F_\lambda\}\) be an arbitrary maximal countably centered system of closed sets of the space \(X\). Consider the system \(f\xi=\{fF_\lambda\}\) of closed subsets of \(Y\). The system \(f\xi\) is a maximal countably centered system of closed sets in \(Y\). Since \(Y\) is functionally closed, we have
\[
\bigcap_\lambda fF_\lambda \ne \Lambda.
\]
Let
\[
y_0\in \bigcap_\lambda fF_\lambda.
\]
Then \(f^{-1}y_0\cap F_\lambda\ne \lambda\) for all \(\lambda\). Hence it follows that \(f^{-1}y_0\in\xi\) by the maximality of the system \(\xi\). Therefore the system
\[
\{f^{-1}y_0\cap F_\lambda\}=\xi_{y_0}
\]
is a countably centered system of closed sets in \(f^{-1}y_0\). It is easy to prove that \(\xi_{y_0}\) is maximal with respect to these properties. Since \(f^{-1}y_0\) is functionally closed, we have
\[
\bigcap_\lambda (f^{-1}y_0\cap F_\lambda)\ne \Lambda,
\]
and then all the more
\[
\bigcap_\lambda F_\lambda\ne \Lambda.
\]
Theorem 5 is proved.
I express my gratitude to V. I. Ponomarev for his help in carrying out this work.
Moscow State University
named after M. V. Lomonosov
Received
10 VI 1966
REFERENCES
- E. Hewitt, Trans. Am. Math. Soc., 64, No. 1, 45 (1948).
- C. H. Dowker, Canad. J. Math., 3, No. 2, 219 (1951).
- Z. Frolik, Czechoslovak Math. J., vol. 13, No. 1, 127 (1963).
- V. I. Ponomarev, DAN, 126, 716 (1959).
- B. A. Pasynkov, Mat. Sb., 66, No. 1, 35 (1965).