J. Math. Anal. Appl.

412 (2014) 125–137

Contents lists available at ScienceDirect

Journal of Mathematical Analysis and

Applications
www.elsevier.com/locate/jmaa

A monotone version of the Sokolov property and monotone

retractability in function spaces
R. Rojas-Hernández a , V.V. Tkachuk b,∗,1,2
a
Facultad de Ciencias, Universidad Nacional Autónoma de México, Ciudad Universitaria, 04510, México D.F., Mexico
b
Departamento de Matemáticas, Universidad Autónoma Metropolitana, Av. San Rafael Atlixco, 186, Col. Vicentina, 09340, México D.F., Mexico

a r t i c l e i n f o a b s t r a c t

Article history: We introduce the monotone Sokolov property and show that it is dual to monotone
Received 31 March 2013 retractability in the sense that X is monotonically retractable if and only if C p ( X ) is
Available online 21 October 2013 monotonically Sokolov. Besides, a space X is monotonically Sokolov if and only if C p ( X )
Submitted by B. Cascales
is monotonically retractable. Monotone retractability and monotone Sokolov property are
Keywords:
shown to be preserved by R-quotient images and F σ -subspaces. Furthermore, every
Lindelöf Σ -space monotonically retractable space is Sokolov so it is collectionwise normal and has countable
Retraction extent. We also establish that if X and C p ( X ) are Lindelöf Σ -spaces then they are both
ω-Monotone operator monotonically retractable and have the monotone Sokolov property. An example is given
Monotonically retractable space of a space X such that C p ( X ) has the Lindelöf Σ -property but neither X nor C p ( X ) is
Simple space monotonically retractable. We also establish that every Lindelöf Σ -space with a unique
Sokolov space non-isolated point is monotonically retractable. On the other hand, each Lindelöf space
Monotonically Sokolov space
with a unique non-isolated point is monotonically Sokolov.
Gul’ko space
© 2013 Elsevier Inc. All rights reserved.
Normal space
Collectionwise normal space
Lindelöf space
Function space
Extent

1. Introduction

Spaces with a rich family of retractions often occur both in topology and in functional analysis. Amir and Lindenstrauss
used them in [1] to characterize Eberlein compact spaces. Gul’ko used families of retractions in [6] to prove that a compact
space X is Corson compact whenever C p ( X ) has the Lindelöf Σ -property. Tkachuk also applied families of retractions in [13]
to prove that, for any Tychonoff space X , if C p ( X ) is a Lindelöf Σ -space, then X can be condensed into a Σ -product of real
lines.
In the paper [10], Rojas-Hernández introduced the notion of a monotonically retractable space and proved that every
closed subset of a Σ -product of cosmic spaces has this property. Besides, if X is monotonically retractable then so is
C p C p ( X ) while the space C p ( X ) has to be a Lindelöf D-space. The techniques developed in the paper [10] made it possible
to solve several published questions on the D-property of function spaces and generalize some results of Okunev and
Buzyakova.
We continue the study initiated in [10] considering, together with monotone retractability, a monotone version of the
Sokolov property. It turns out that the monotone Sokolov property is two-way dual to monotone retractability in the sense

* Corresponding author.
E-mail addresses: satzchen@yahoo.com.mx (R. Rojas-Hernández), vova@xanum.uam.mx (V.V. Tkachuk).
1
Research supported by Promep 912011, Proyecto 12611768 (Mexico).
2
Research supported by CONACyT grant CB-2012-01-178103 (Mexico).

0022-247X/$ – see front matter © 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jmaa.2013.10.043
126 R. Rojas-Hernández, V.V. Tkachuk / J. Math. Anal. Appl. 412 (2014) 125–137

that X is monotonically retractable if and only if C p ( X ) is monotonically Sokolov and X is monotonically Sokolov if and
only if C p ( X ) is monotonically retractable.
Besides, monotone retractability implies the Sokolov property and hence every monotonically retractable space is col-
lectionwise normal and has countable extent. We show that if X and C p ( X ) are Lindelöf Σ -spaces, then they are both
monotonically retractable and monotonically Sokolov. Therefore X and C p ( X ) must both be Sokolov; this provides a positive
answer to Problem 3.8 of [12].
We give an example of a pseudocompact space X such that C p ( X ) has the Lindelöf Σ -property but neither X nor
C p ( X ) is monotonically retractable. It turns out that every Lindelöf Σ -space with a unique non-isolated point must be
monotonically retractable; however it seems to be a difficult problem to characterize monotone retractability even for the
spaces which have a unique non-isolated point.

2. Notation and terminology

All spaces are assumed to be Tychonoff. Given a space X , the family τ ( X ) is its topology and τ (x, X ) = {U ∈ τ ( X ): x ∈ U }
for any x ∈ X ; if A ⊂ X then τ ( A , X ) = {U ∈ τ ( X ): A ⊂ U }. If X is a set then exp( X ) = {Y : Y ⊂ X }. All ordinals are identified
with the set of their predecessors and are assumed to carry the interval topology. As usual, R is the set of reals; the set
ω\{0} is denoted by N. For any infinite cardinal κ the space L (κ ) is the lindelöfication of the discrete space of cardinality κ .
A space X is called simple if it has at most one non-isolated point. We denote by υ X the Hewitt extension of the space X .
For any spaces X and Y the set C ( X , Y ) consists of continuous functions from X to Y ; if it has the topology induced
from Y X then the respective space is denoted by C p ( X , Y ). We write C p ( X ) instead of C p ( X , R). For an arbitrary set A let
Σ( A ) = {x ∈ R A : the set {a ∈ A: x(a) = 0} is countable}. All possible sets Σ( A ) are called Σ -products of real lines.
If ϕ : X → Y is a map then ϕ n : X n → Y n is defined by the formula ϕ (x) = (ϕ (x1 ), . . . , ϕ (xn )) for any point x =
(x1 , . . . , xn ) ∈ X n and n ∈ N. Given points x1 , . . . , xn ∈ X and sets B 1 , . . . , B n ∈ τ (R), the set W = [x1 , . . . , xn ; B 1 , . . . , B n ] =
{ f ∈ C p ( X ): f (xi ) ∈ B i for all i  n} is open in C p ( X ) and supp( W ) = {x1 , . . . , xn }; more generally, if we have sets
A 1 , . . . , A n ⊂ X then [ A 1 , . . . , A n , B 1 , . . . , B n ] = { f ∈ C p ( X ): f ( A i ) ⊂ B i for all i  n}. For any set Y ⊂ X let πY : C p ( X ) →
C p (Y ) be the restriction map. If ϕ : X → Y is a continuous onto map then ϕ ∗ : C p (Y ) → C p ( X ) is the dual map of ϕ defined
by the formula ϕ ∗ ( f ) = f ◦ ϕ for any f ∈ C p (Y ). If X is a space and L ⊂ C p ( X ) then the map e L : X → C p ( L ) is defined by
the formula e L (x)( f ) = f (x) for all x ∈ X and f ∈ L.
Given a set A in a space X , say that a family N of subsets of X is an external network of A in X if for any x ∈ A and
U ∈ τ (x, X ) there exists N ∈ N such that x ∈ N ⊂ U . A family A is a network with respect to a family C of subsets of X if
for any C ∈ C and U ∈ τ (C , X ) there exists a set A ∈ A such that C ⊂ A ⊂ U . If N is a network in X with respect to the
family {{x}: x ∈ X } then N is called a network in X . The spaces with a countable network are called cosmic. If X is a space
and f : X → Y is a continuous map, then a family N ⊂ exp( X ) is a network of f if for any x ∈ X and U ∈ τ ( f (x), Y ) there
exists N ∈ N such that x ∈ N and f ( N ) ⊂ U . Say that X is a Lindelöf Σ -space (or has the Lindelöf Σ -property) if it has a
countable mod–k-network, i.e., a countable network with respect to a compact cover of X . Call X a Gul’ko space if C p ( X ) has
the Lindelöf Σ -property.
Given an infinite cardinal κ say that a space X is monotonically κ -monolithic if, to any set A ⊂ X with | A |  κ , we can
assign an external network O ( A ) to the set A in such a way that the following conditions are satisfied:

(a) |O ( A )|  max{| A |, ω};

(b) if A ⊂ B ⊂ X and | B |  κ then O ( A ) ⊂ O ( B );
(c) if λ κ is a cardinal
 and we have a family { A α : α < λ} ⊂ [ X ]κ such that α < β < λ implies A α ⊂ A β then
O( α <λ A α ) = α <λ O( A α ).

The space X is monotonically monolithic if it is monotonically κ -monolithic for any infinite cardinal κ .
Say that X is a Sokolov space (or has the Sokolov property) if for every sequence { F n : n ∈ N}, where each F n is a closed
subset of X n , there exists a continuous map f : X → X such that nw ( f ( X ))  ω and f n ( F n ) ⊂ F n for any n ∈ N. Recall that
t ( X )  κ if for any A ⊂ X and any point x ∈ A there exists a set B ⊂ A such that | B |  κ and x ∈ B.
The unexplained notions can be found in the books [3] and [14]; the survey of Hodel [7] can be consulted for definitions
of cardinal invariants.

3. Monotonically Sokolov spaces: an overview

Monotonically retractable spaces were introduced and studied in [10] where it was proved, among other things, that not
every Sokolov space is monotonically retractable. However, all examples of monotonically retractable spaces had the Sokolov
property. One of the motivations of the present paper was to show that this is not a coincidence.

3.1. Definition. (See [10].) A space X is monotonically retractable if we can assign to any countable set A ⊂ X a retraction
r A : X → X and a countable network N ( A ) of r A such that A ⊂ K ( A ) = r A ( X ) and the assignment N is ω -monotone.
 R. Rojas-Hernández, V.V. Tkachuk / J. Math. Anal. Appl. 412 (2014) 125–137 127

It turns out that the C p -dual of monotone retractability allows a natural description and has nice categorical properties.
We will call it monotone Sokolov property.

3.2. Definition. A space X is monotonically Sokolov if we can assign to any countable family F of closed subsets of X a
continuous retraction rF : X → X and a countable external network N (F ) for rF ( X ) in X such that rF ( F ) ⊂ F for each
F ∈ F and the assignment N is ω -monotone.

3.3. Observation. The referee noted that if we use countable sets instead of countable families of closed sets in Definition 3.2
(i.e., if we consider only the families F whose elements are singletons), then we obtain an interesting weakening of the
monotone Sokolov property which also looks similar to monotone retractability because external networks are used instead
of networks of retractions. For a moment, let us say that X is a WMS-space if it has the above-mentioned weakening of the
monotone Sokolov property.
Observe first that if X is a discrete space, then for any countable set A ⊂ X the family N A = {{x}: x ∈ A } is a countable
external network of A in X and it is easy to find a retraction r A : X → A. The assignment A → N A is trivially ω -monotone
so every discrete space has the WMS-property. Therefore, if X is a discrete space of cardinality ω1 then X is a WMS-space
which is not even Sokolov because every Sokolov space has countable extent. Therefore the class of WMS-spaces is strictly
larger than the class of monotonically Sokolov spaces.
The WMS-property does not coincide with monotone retractability either because every WMS-space is monotonically
ω-monolithic while a monotonically retractable space need not be monotonically ω-monolithic as can be seen from Exam-
ple 4.27.

We will prove in Section 4 that the class of monotonically Sokolov spaces has the so nice categorical properties that it
seems to be interesting in itself.

3.4. Theorem.

(a) Any monotonically Sokolov space is Lindelöf and Sokolov.

(b) Any monotonically Sokolov space is monotonically ω -monolithic.
(c) If X is a monotonically Sokolov space, then every F σ -subset of X is also monotonically Sokolov.
(d) Any countable product of monotonically Sokolov spaces is monotonically Sokolov.
(e) A compact monotonically Sokolov space is Corson compact.
(f) A Corson compact space must be Sokolov but not necessarily monotonically Sokolov.
(g) Every R-quotient image of a monotonically Sokolov space is monotonically Sokolov. In particular, any continuous image of a
monotonically Sokolov compact space is monotonically Sokolov.

One of the main results of the paper is the following fact that makes clear the relationship between monotone re-
tractability and monotone Sokolov property.

3.5. Theorem.

(a) A space X is monotonically retractable if and only if C p ( X ) is monotonically Sokolov.

(b) A space X is monotonically Sokolov if and only if C p ( X ) is monotonically retractable.

Of course, it is not a surprise that Theorem 3.5 makes it possible to prove new interesting facts about monotone re-
tractability. The following statement emphasizes some nice categorical properties of monotone retractability. The results in
(a)–(c) are cited from [10].

3.6. Theorem.

(a) Every Corson compact space is monotonically retractable.

(b) Any σ -product of monotonically retractable spaces must be monotonically retractable.
(c) Every Σ -product of monotonically retractable spaces must be monotonically retractable.
(d) Any monotonically retractable space X is a Sokolov space. In particular, X is ω -stable, collectionwise normal and ext( X )  ω .
However, X need not be Lindelöf.
(e) If a space X is monotonically retractable and Y is an F σ -set in X , then Y is monotonically retractable.
(f) Every R-quotient image of a monotonically retractable space X must be monotonically retractable. In particular, if X is a mono-
tonically retractable compact space then any continuous image of X is monotonically retractable.
(g) A monotonically retractable space must be ω -monolithic but it can fail to be monotonically ω -monolithic.
128 R. Rojas-Hernández, V.V. Tkachuk / J. Math. Anal. Appl. 412 (2014) 125–137

The last section of the paper is devoted to applications of monotonically retractable and monotonically Sokolov spaces.
The main result here is the following fact that solves (in a much stronger form) Problem 3.8 from [12].

3.7. Theorem. If X and C p ( X ) are Lindelöf Σ -spaces then they must be both monotonically retractable and monotonically Sokolov. In
particular, X and C p ( X ) have the Sokolov property.

4. Technicalities

Unfortunately, it is not possible to prove the facts of Section 3 in the same order as they were formulated. This section
will have its own numbering for its statements, but we will always indicate the moments when the proof of an assertion of
the overview is complete.

4.1. Definition. Suppose that a family A ⊂ exp( X ) is closed under countable increasing unions, B ⊂ exp(Y ) and we have a
map ϕ : A → B . We say that ϕ is ω -monotone if
(1) ϕ ( A ) is countable whenever A ∈ A is countable;
(2) if A ⊂ B and A , B ∈ A, then ϕ ( A ) ⊂ ϕ ( B );  
(3) if { A n : n ∈ ω} ⊂ A and A n ⊂ A n+1 for any n ∈ ω then we have the equality ϕ ( n∈ω A n ) = n∈ω ϕ ( A n ).

4.2. Observation. Suppose that X is a space and B is a base in R; given any A ⊂ exp( X ) let W (A) = {[ A 1 , . . . , A n ; B 1 , . . . ,
B n ]: n ∈ N, A 1 , . . . , A n ∈ A and B 1 , . . . , B n ∈ B }. It is an easy exercise to check that the assignment A → W (A) is
ω-monotone.

We will often use without reference the following easy fact.

4.3. Proposition. Given families A ⊂ exp( X ), B ⊂ exp(Y ) and C ⊂ exp( Z ) for some sets X , Y and Z ,

(a) if the maps p : A → B and q : B → C are ω -monotone then the map q ◦ p : A → C is also ω -monotone;
countable unions, | T |  ω and pt : A → B is ω -monotone for any t ∈ T then the map p : A → B
(b) if the family B is invariant under
defined by the equality p ( A ) = { pt ( A ): t ∈ T } for any A ∈ A is also ω -monotone.

The following statement takes care of Theorem 3.4(b).

4.4. Proposition. Suppose that X is a monotonically Sokolov space. Then X is monotonically ω -monolithic.

Proof. For any countable family F of closed subsets of the space X we can find a retraction rF : X → X and a
countable family N (F ) ⊂ exp( X ) that witness the monotone Sokolov property of X . Given a countable set A ⊂ X let
F ( A ) = {{x}: x ∈ A } and O( A ) = N (F ( A )). It is immediate that the assignment A → F ( A ) is ω -monotone so the oper-
ator O is also ω -monotone. Since rF ( A ) ( F ) ⊂ F for any F ∈ F ( A ), we have the equality rF ( A ) (x) = x for any x ∈ A. The
continuity of the map rF ( A ) implies that rF ( A ) (x) = x for any x ∈ A and therefore A ⊂ rF ( A ) ( X ). The family N (F ( A )) being
a countable external network for rF ( A ) ( X ) in X , it must be a countable external network for A in X so X is monotonically
ω-monolithic. 2

4.5. Corollary. If X is monotonically Sokolov and t ( X )  ω , then the space X is monotonically monolithic.

Proof. Apply Proposition 4.4 and note that any monotonically ω-monolithic space of countable tightness is monotonically
monolithic by Theorem 2.10 of the paper [15]. 2

Our next step is to settle Theorem 3.4(c). For a space X , we denote by CL( X ) thefamily of all closed subsets of X ; let
CL∗ ( X ) be the family of all closed subsets of some finite power of X , i.e., CL∗ ( X ) = {CL( X n ): n ∈ N}.

4.6. Proposition. If X is a monotonically Sokolov space, then any F σ -subset of X is monotonically Sokolov.

Proof. For any countable G ⊂ CL( X ) fix a continuous retraction rG : X → X and a countable collection N (G ) ⊂ exp( X ) that
witness the monotone Sokolov property of X . Take any F σ -subset Y of the space X and let H be a countable family of
closed subsets of X with Y = H.
For any G ⊂ CL(Y ) let F (G ) = H ∪ {G ∩ H: G ∈ G and H ∈ H}. Observe that F (G ) ⊂ CL( X ) and the assignment G →
F (G ) is ω -monotone. Consider the map sG = rF (G ) |Y and let O(G ) = { N ∩ Y : N ∈ N (F (G ))}. We will prove that the
map sG and the family O (G ) witness the monotone Sokolov property of the space Y .
 R. Rojas-Hernández, V.V. Tkachuk / J. Math. Anal. Appl. 412 (2014) 125–137 129

If y ∈ Y , then y ∈ H for some H ∈ H. Since H ∈ H ⊂ F (G ), by the choice of rF (G ) , we have rF (G ) ( H ) ⊂ H . Hence sG ( y ) =

rF (G ) ( y ) ∈ H ⊂ Y so sG : Y → Y is a continuous retraction. The family N (F (G )) being an external network for rF (G ) ( X )
in X , it follows that O (G ) is an external network for sG (Y ) in Y . The assignments F → F (G ) and F (G ) → N (F (G )) are
ω-monotone so the operator O must also be ω-monotone. Finally, we have the inclusion sG (G ) ⊂ G for each G ∈ G . Indeed,
take any y ∈ G for some G ∈ G . Since y ∈ Y = H, we can find H ∈ H with y ∈ H . Then y ∈ G ∩ H ∈ F (G ). It follows from
the choice of rF (G ) that rF (G ) (G ∩ H ) ⊂ G ∩ H . Thus sG ( y ) = rF (G ) ( y ) ∈ G ∩ H ⊂ G. 2

4.7. Theorem. Given a space X suppose that for any countable set A ⊂ X , we have a retraction r A : X → K ( A ) and a family N ( A ) that
witness monotone retractability of X . Then, to any countable family G of closed subsets of X we can assign a countable set P (G ) ⊂ X
such that r P (G ) (G ) ⊂ G for any G ∈ G and the assignment G → P (G ) is ω -monotone.

Proof. Let N ( X ) = {N ( A ): A is a countable subset of X }. For any closed set G ⊂ X and H ∈ N ( X ), if G ∩ H = ∅ then
pick a point p G , H ∈ G ∩ H . Given a countable family G of closed subsets of X and a countable subfamily H of N ( X ), let
E (G , H) = { p G , H : G ∈ G , H ∈ H, and G ∩ H = ∅}. If G is a non-empty closed subset of X , then pick a point p G ∈ G. For
a countable family G ⊂ CL( X ) let Q (G ) = { p G : G ∈ G \ {∅}}. The assignment G → Q (G ) is easily seen  to be ω -monotone.
Define P 0 (G ) = Q (G ) and P n+1 (G ) = P n (G ) ∪ E (G , N ( P n (G ))) for any n ∈ ω . Then the set P (G ) = { P n (G ): n ∈ ω} is as
promised.
It takes an easy induction and Proposition 4.3 to show that each of the assignments P n is ω -monotone and hence P is
also ω -monotone. Let us show that r P (G ) (G ) ⊂ G for each G ∈ G .
/ G for some G ∈ G and x ∈ G. Then U = X \G is a neighborhood of r P (G ) (x) which does
If this is not true, then r P (G ) (x) ∈
not intersect G. SinceN ( P (G )) is a network  of r P (G ) , there exists a set H ∈ N ( P (G )) such that x ∈ H and r P (G ) ( H ) ⊂ U .
Since N ( P (G )) = N ( { P n (G ): n ∈ ω}) = {N ( P n (G )): n ∈ ω}, we have H ∈ N ( P n (G )) for some n ∈ ω . It follows from
x ∈ G ∩ H that G ∩ H = ∅, so a point p G , H ∈ G ∩ H has been chosen and
    
p G , H ∈ E G , N P n (G ) ⊂ P n+1 (G ) ⊂ P (G ) ⊂ K P (G ) .
Therefore p G , H = r P (G ) ( p G , H ) ∈ r P (G ) ( H ) ⊂ U ⊂ X \ G. This contradiction shows that r P (G ) (G ) ⊂ G for any G ∈ G . 2

4.8. Corollary. For any space X , the following conditions are equivalent:

(a) X is monotonically retractable;

(b) for any countable family G ⊂ CL( X ) there exist a retraction sG : X → X and a countable network O (G ) of sG , such that
sG (G ) ⊂ G for each G ∈ G , and the assignment O is ω -monotone;
(c) for any countable family G ⊂ CL∗ ( X ) there exist a retraction t G : X → X and a countable network P (G ) of t G such that, for any
n ∈ N, we have the inclusion (t G )n (G ) ⊂ G for each G ∈ G ∩ CL( X n ), and the assignment P is ω -monotone.

Proof. It is immediate that (c) ⇒ (b). If (b) holds and A is a countable subset of X then consider the family G ( A ) =
{{x}: x ∈ A }, the retraction r A = sG ( A ) , the set K ( A ) = r A ( X ) and the family N ( A ) = O(G ( A )). It is easy to check that
r A , K ( A ) and N ( A ) witness that X is monotonically retractable so we have (b) ⇒ (a).
To prove that (a) ⇒ (c), for any countable set B ⊂ X fix a set L ( B ) ⊂ X , a continuous retraction s B : X → L ( B ) and a
countable network O ( B ) of s B such that B ⊂ L ( B ) and the assignment O is ω -monotone.
Fix a point a ∈ Y = X N and let pn : Y →  X be the projection onto the n-th factor for any n ∈ N. Given a countable set
A ⊂ Y , it is immediate that the set E ( A ) = { pn ( A ∪{a}): n ∈ N} ⊂ X is countable. For every point y ∈ Y let (s E ( A ) )N ( y )(n) =
N
s E ( A ) ( y (n)) for any n ∈ N. Then the  map r A = (s E ( A ) ) : Y → Y is a retraction; let K ( A ) = r A (Y ). Denote by N ( A ) the family
of all sets U ⊂ Y such that U = { N n : n ∈ N} where, for some finite set F ⊂ N, we have N n ∈ O ( E ( A )) for all n ∈ F and
N n = X if n ∈ / F . It is standard to verify that K ( A ), r A , and N ( A ) witness the fact that Y is monotonically retractable.
Now we can apply Theorem 4.7 to convince ourselves that for any countable family H of closed subsets of Y we can
choose a countable set P (H) ⊂ Y such that r P (H) ( H ) ⊂ H for any H ∈ H and the assignment P is ω -monotone.
Observe that the set Z n = { y ∈ Y : {k ∈ N: y (k) = a(k)} ⊂ {1, . . . , n}} ⊂ Y is closed in Y and canonically homeomorphic
to X n for any n ∈ N. Up to this canonical homeomorphism the map r A | Z n = (s E ( A ) )N | Z n can be identified with (s E ( A ) )n :
X n → X n for any n ∈ N and A ⊂ Y . If G is a countable subfamily of CL∗ ( X ), then we can apply the above-mentioned
canonical homeomorphisms to consider that every G ∈ G is a closed subset of Z n for some n ∈ N.
Let t G = s E ( P (G )) : X → X and P (G ) = O ( E ( P (G ))). Then the map t G is a retraction and P (G ) is a network of t G . Since
the assignments P , E and O are ω -monotone, the assignment P is also ω -monotone. Finally, for any G ∈ G ∩ CL( X n ) under
our identification of the set G with a subset of Z n , we have the inclusion (t G )n (G ) = (s E ( P (G )) )n (G ) = (s E ( P (G )) )N (G ) =
r P (G ) (G ) ⊂ G. 2

The following corollary settles the first part of Theorem 3.6(d).

4.9. Corollary. Any monotonically retractable space X is a Sokolov space. In particular, X is ω-stable, collectionwise normal and
ext( X )  ω .
130 R. Rojas-Hernández, V.V. Tkachuk / J. Math. Anal. Appl. 412 (2014) 125–137

Proof. The space X has the property described in Corollary 4.8(c) which clearly implies that X has the Sokolov property.
The rest follows from the fact that every Sokolov space is ω -stable, collectionwise normal and has countable extent (see
[12, Proposition 2.2]). 2

Now we can prove Theorem 3.6(e).

4.10. Proposition. If X is monotonically retractable and Y is an F σ -set in X , then Y is monotonically retractable.


Proof. Let F be a countable family of closed subsets of X such that Y = F . Apply Corollary 4.8 to assign to any countable
family G ⊂ CL( X ) a retraction sG : X → X and a countable network O (G ) of sG such that sG (G ) ⊂ G for any G ∈ G and the
assignment O is ω -monotone.
For any countable set A ⊂ Y let G ( A ) = F ∪ {{x}: x ∈ A }. It is clear that the assignment A → G ( A ) is ω -monotone;
observe first that sG ( A ) (Y ) ⊂ Y . Indeed, if y ∈ Y then y ∈ F for some F ∈ F ⊂ G ( A ) so sG ( A ) ( F ) ⊂ F which implies that
sG ( A ) ( y ) ∈ F ⊂ Y . Therefore r A = sG ( A ) |Y : Y → Y is a retraction. Since O (G ( A )) is a network of sG ( A ) , the family N ( A ) =
{ P ∩ Y : P ∈ O(G ( A ))} is a network of r A and it is immediate that the assignment N is ω -monotone.
If K ( A ) = r A (Y ) then for any x ∈ A it follows from r A ({x}) ⊂ {x} that r A (x) = x; this shows that A ⊂ K ( A ) and hence the
map r A , the set K ( A ) and the family N ( A ) witness that Y is monotonically retractable. 2

Recall that a family N ⊂ exp( X ) is an external network of a set A ⊂ X if for any x ∈ A and U ∈ τ (x, X ) there exists N ∈ N
such that x ∈ N ⊂ U . If we have a map f : X → Y then a family N ⊂ exp( X ) is a network of f if for any U ∈ τ (Y ) and
x ∈ f −1 (U ) there exists N ∈ N with x ∈ N ⊂ f −1 (U ). From now on, the symbol B R will denote some fixed countable base
in R.
If X is a space and A is a family of subsets of X then W (A) will be the family we defined in Observation 4.2, i.e.,
W (A) = {[ A 1 , . . . , An ; B 1 , . . . , B n ]: n ∈ N, A 1 , . . . , An ∈ A and B 1 , . . . , B n ∈ B R }.
The following facts proved in [11] illustrate the relationship between the above-mentioned notions in function spaces.

4.11. Proposition. Suppose that ϕ : X → Y is a continuous onto map and N is a network of ϕ . Then W (N ) is an external network of
ϕ ∗ (C p (Y )) in C p ( X ).

4.12. Proposition. If Y ⊂ X and N is an external network of Y in X , then the family W (N ) of subsets of C p ( X ) is a network of πY .

4.13. Lemma. If X is a space, K ⊂ X and r : X → K is a retraction, then r ∗ ◦ π K : C p ( X ) → C p ( X ) is a retraction onto its image.

We are now ready to prove that monotone retractability and monotone Sokolov properties are two-way dual with respect
to the C p -functor.

4.14. Theorem. If X is monotonically retractable, then C p ( X ) is monotonically Sokolov.

Proof. Apply Corollary 4.8, to assign to any countable family G ⊂ CL∗ ( X ) a continuous retraction t G : X → X and a count-
able network P (G ) of t G such that (t G )n (G ) ⊂ G for each G ∈ G ∩ CL( X n ) and the assignment P is ω -monotone.
Let F be a countable family of closed subsets of C p ( X ). Given any F ∈ F , for any n ∈ N and B 1 , . . . , B n ∈ B R consider
the subset
 
G ( F , B 1 , . . . , B n ) = (x1 , . . . , xn ) ∈ X n : [x1 , . . . , xn ; B 1 , . . . , B n ] ∩ F = ∅
of the space X n . Observe that every set G ( F , B 1 , . . . , B n ) is closed in X n and therefore G (F ) = {G ( F , B 1 , . . . , B n ): F ∈ F ,
B 1 , . . . , B n ∈ B R , and n ∈ N} is a countable subfamily of CL∗ ( X ). The assignment F → G (F ) is easily seen to be
ω-monotone. Consider the set M (F ) = t G (F ) ( X ); we will prove that the retraction rF = (t G (F ) )∗ ◦ πM (F ) and the fam-
ily N (F ) = W (P (G (F ))) witness that C p ( X ) is monotonically Sokolov.
By Lemma 4.13, the map rF : C p ( X ) → C p ( X ) is a retraction. The family P (G (F )) being a network of t G (F ) , we can
apply Proposition 4.11 to see that the family W (P (G (F ))) is an external network for (t G (F ) )∗ (C p ( M (F ))) in C p ( X ). This
is the same as saying that W (P (G (F ))) is an external network for the set (t G (F ) )∗ ◦ π M (F ) (C p ( X )) in C p ( X ) i.e., N (F ) is
an external network for rF (C p ( X )) in C p ( X ). Since the assignments W , P and G are ω -monotone, the assignment N is
ω-monotone. Now, to finish the proof, we only need to show that rF ( F ) ⊂ F for each F ∈ F .
Suppose that rF ( f ) ∈ / F for some F ∈ F and f ∈ F . The set F being closed in C p ( X ), we can choose x1 , . . . , xn ∈ X
and B 1 , . . . , B n ∈ B R such that the set U = [x1 , . . . , xn ; B 1 , . . . , B n ] contains rF ( f ) and U ∩ F = ∅. Therefore (x1 , . . . , xn ) ∈
G ( F , B 1 , . . . , B n ) ∈ G (F ) ∩ CL( X n ). Since (t G (F ) )n (G ) ⊂ G for any G ∈ G ( F ) ∩ CL( X n ), we have (t G (F ) )n (x1 , . . . , xn ) ∈
G ( F , B 1 , . . . , B n ). As a consequence, V ∩ F = ∅ if V = [t G (F ) (x1 ), . . . , t G (F ) (xn ); B 1 , . . . , B n ]. On the other hand, rF ( f ) ∈ U
implies rF ( f )(xk ) = f (t G (F ) (xk )) ∈ B k for k = 1, . . . , n, i.e., f ∈ V ∩ F ; this contradiction shows that rF ( F ) ⊂ F for each
F ∈ F. 2
 R. Rojas-Hernández, V.V. Tkachuk / J. Math. Anal. Appl. 412 (2014) 125–137 131

4.15. Lemma. Suppose that f : X → Y is a continuous map and B is a base for Y . Assume that, for the family F = { f −1 ( B ): B ∈ B },
there exists a continuous map r : X → X such that r ( F ) ⊂ F for any F ∈ F . Then f = ( f | r ( X )) ◦ r.

Proof. Take any point x ∈ X ; we must show that f (x) = f (r (x)). Given a set B ∈ B with f (x) ∈ B, observe that x ∈ F =
f −1 ( B ) ∈ F . By our hypothesis, we have r (x) ∈ F , so f (r (x)) ∈ B. Thus, f (r (x)) ∈ B for any B ∈ B with f (x) ∈ B and hence
f (x) = f (r (x)). 2

4.16. Corollary. Let X be a Sokolov space. Then for any function f ∈ C ( X ) there exists a continuous map r from X onto a cosmic
subspace of X , such that f = ( f | r ( X )) ◦ r.

4.17. Theorem. If X is monotonically Sokolov, then C p ( X ) is monotonically retractable.

Proof. We can assign to any countable F ⊂ CL( X ) a retraction rF : X → X and a countable external network N (F ) for rF ,
such that rF ( F ) ⊂ F for every F ∈ F and the assignment N is ω -monotone.
For any countable set E ⊂ C p ( X ) let F ( E ) = { f −1 ( B ): f ∈ E and B ∈ B R }. The assignment E → F ( E ) is easily seen to
be ω -monotone. Consider the set K (F ( E )) = rF ( E ) ( X ) and let s E = (rF ( E ) )∗ ◦ π K (F ( E )) . It turns out that the map s E , the set
L ( E ) = s E (C p ( X )) and the family O ( E ) = W (N (F ( E ))) witness that C p ( X ) is monotonically retractable.
It follows from Lemma 4.13 that s E = (rF ( E ) )∗ ◦ π K (F ( E )) : C p ( X ) → L ( E ) is a retraction. Since the family N (F ( E )) is
an external network for K (F ( E )) in X , Proposition 4.12 shows that the family W (N (F ( E ))) is a network of π K (F ( E )) . It
follows from continuity of (rF ( E ) )∗ that W (N (F ( E ))) is a network of (rF ( E ) )∗ ◦ π K (F ( E )) , i.e., the family O ( E ) is a network
of s E . All assignments W , N , and F being ω -monotone, the assignment O is ω -monotone as well.
To convince ourselves that E ⊂ L ( E ) take any f ∈ E and observe that { f −1 ( B ): B ∈ B R } ⊂ F ( E ). By the choice of rF ( E ) ,
we have rF ( E ) ( F ) ⊂ F for any F ∈ F ( E ). Therefore we can apply Lemma 4.15 to conclude that f = ( f | K (F ( E ))) ◦ rF ( E ) =
(rF ( E ) )∗ ◦ π K (F ( E )) ( f ) = s E ( f ) ∈ L ( E ). 2

The corollary that follows coincides with Theorem 3.5(a).

4.18. Corollary. A space X is monotonically retractable if and only if C p ( X ) is monotonically Sokolov.

Proof. Necessity was proved in Theorem 4.14. Now, if C p ( X ) is monotonically Sokolov then C p (C p ( X )) is monotonically
retractable by Theorem 4.17. Since X embeds in C p (C p ( X )) as a closed subspace, it must be monotonically retractable by
Proposition 4.10. 2

The next statement is the same as Theorem 3.5(b).

4.19. Corollary. A space X is monotonically Sokolov if and only if C p ( X ) is monotonically retractable.

Proof. Necessity was proved in Theorem 4.17. Now, if C p ( X ) is monotonically retractable then C p (C p ( X )) is monotoni-
cally Sokolov by Theorem 4.14. Since X embeds in C p (C p ( X )) as a closed subspace, it must be monotonically Sokolov by
Proposition 4.6. 2

We are ready now to prove the statement of Theorem 3.4(a).

4.20. Corollary. Any monotonically Sokolov space is Lindelöf and Sokolov.

Proof. If X is monotonically Sokolov, then C p ( X ) is monotonically retractable by Theorem 4.17 so it is Sokolov by Corol-
lary 4.9. Therefore X is a Sokolov space by [12, Theorem 2.1]. Besides, C p (C p ( X )) must be Lindelöf by Theorem 3.18 of [10]
so X is Lindelöf being homeomorphic to a closed subset of C p (C p ( X )). 2

Our next step is to verify Theorem 3.4(d).

4.21. Corollary. Any countable product of monotonically Sokolov spaces is monotonically Sokolov.

Proof. If X is monotonically Sokolov, then C p (C p ( X )) is monotonically Sokolov by Theorem 4.14 and Theorem 4.17. It follows
from [14, Problem 177] that the space Y = C p (C p ( X )) is homeomorphic to Y ω . The space X is homeomorphic to a closed
subspace of Y and therefore X ω is also homeomorphic to a closed subspace of Y Y ω so it follows from Proposition 4.6
that X ω is monotonically Sokolov. 
Now assume that a space X i is monotonically Sokolov for every i ∈ ω . Then the space X = { X i : i ∈ ω}  is also monotoni-
cally Sokolov because every C p ( X i ) must be monotonically retractable by Theorem 4.17 and hence C p ( X ) {C p ( X i ): i ∈ ω}
132 R. Rojas-Hernández, V.V. Tkachuk / J. Math. Anal. Appl. 412 (2014) 125–137


is also monotonically retractable by Corollary 3.15 of [10]. Thus X ω is monotonically Sokolov and hence so is { X i : i ∈ ω}
being homeomorphic to a closed subspace of X ω . 2

Recall that a space X is called monotonically ω -stable [9] if to each countable set L ⊂ C p ( X ) we can assign a network
O( L ) ⊂ exp( X ) of the reflection map e L : X → C p ( L ) in such a way that the operator O is ω -monotone.

4.22. Corollary. Any monotonically retractable space is monotonically ω -stable.

Proof. If X is monotonically retractable then C p ( X ) is monotonically Sokolov (see Theorem 4.14) and hence monotonically
ω-monolithic by Proposition 4.4. Therefore X is monotonically ω-stable by [9, Theorem 3.11]. 2

Now it is time to prove Theorem 3.4(e).

4.23. Corollary. A compact monotonically Sokolov space is Corson compact.

Proof. If X is a monotonically Sokolov compact space, then it is monotonically ω-monolithic by Proposition 4.4. Therefore
the space X is Corson compact by Corollary 2.2 of [5]. 2

The following statement combined with Corollary 4.18 and Corollary 4.9 shows that we verified the first part of Theo-
rem 3.4(f).

4.24. Corollary. If X is Corson compact then C p ( X ) is monotonically Sokolov.

Proof. It was proved in [10] that every Corson compact space is monotonically retractable. Theorem 4.14 does the rest. 2

We will check next that Theorem 3.4(g) and Theorem 3.6(f) are true.

4.25. Proposition.

(a) Any R-quotient image of a monotonically retractable space is monotonically retractable.

(b) Any R-quotient image of a monotonically Sokolov space is monotonically Sokolov.

Proof. If Y is an R-quotient image of X then C p (Y ) embeds in C p ( X ) as a closed subspace and hence both statements (a)
and (b) follow from Theorem 4.14 and Theorem 4.17 together with Proposition 4.6 and Proposition 4.10. 2

4.26. Corollary.

(a) If X is a compact monotonically retractable space, then any continuous image of X is monotonically retractable.
(b) If X is a compact monotonically Sokolov space, then any continuous image of X is monotonically Sokolov.

It is trivial that every monotonically retractable space is ω -monolithic. The example that follows shows that such a space
is not necessarily monotonically ω -monolithic and proves Theorem 3.6(g). The same example proves the second part of
Theorem 3.4(f).

4.27. Example. In the paper [5] Gruenhage gave an example of a Corson compact space X that is not monotonically
ω-monolithic. Therefore X is a Corson compact space which is not monotonically Sokolov (see Proposition 4.4). Hence
C p ( X ) is not monotonically retractable by Corollary 4.19.

Next we are going to present an example that proves the second part of Theorem 3.6(d).

4.28. Example. If X = L (ω1 ) is the one-point lindelöfication of a discrete space of cardinality ω1 , then C p ( X ) is monotoni-
cally retractable but it is not Lindelöf and t ( X ) > ω .

Proof. The space C p ( X ) is homeomorphic to a Σ -product of real lines so it is not Lindelöf; however, it is monotonically
retractable by Theorem 3.11 of [10]. 2

Say that X is a simple space if X has at most one non-isolated point. Observe that the ordinal ω1 with its order topology
is Sokolov but not monotonically Sokolov by Proposition 4.4. It is known that a simple space is Sokolov if and only if it
 R. Rojas-Hernández, V.V. Tkachuk / J. Math. Anal. Appl. 412 (2014) 125–137 133

is Lindelöf (see [12, Theorem 2.1(i)]). It turns out that in simple spaces the Sokolov property is equivalent to it monotonic
version.

4.29. Proposition. For any simple space X , the following are equivalent:

(a) X is Lindelöf ;
(b) X is Sokolov;
(c) X is monotonically Sokolov.

Proof. If X is discrete then all conditions (a)–(c) are equivalent to countability of X so (a) ⇐⇒ (b) ⇐⇒ (c). Now, suppose
that p is a unique non-isolated point of a space X . We must only prove that the Lindelöf property of X implies that it is
monotonically Sokolov. Observe first that any closed subset of X which does not contain p is countable. It is easy to verify
that, for any set A ⊂ X the map s A : X → X given by s A (x) = x if x  ∈ A and s A (x) = p if x ∈
/ A is a continuous retraction.
Given a countable family F of closed subsets of X , the set E (F ) = { F ∈ F : p ∈ / F } is countable. Consider the retraction
rF = s E (F ) and the family N (F ) = {{x}: x ∈ E (F )} ∪ { p }. It is immediate that N (F ) is an external network for rF ( X ) in X ,
we have the inclusion rF ( F ) ⊂ F for each F ∈ F , and the assignment N is ω -monotone. Therefore X is monotonically
Sokolov. 2

4.30. Corollary. If X is a simple Lindelöf space then C p ( X ) is monotonically retractable.

4.31. Remark. A simple Lindelöf space X is not necessarily monotonically retractable; for example X = L (ω1 ) is a Lindelöf
simple space with t ( X ) > ω so it is not monotonically retractable.

5. Applications

In this section we show that the monotonically retractable and monotonically Sokolov spaces are useful for the study
of the Lindelöf Σ -property in function spaces. The main result here is Corollary 5.5 which gives a complete solution of
Problem 3.8 of the paper [12]. We will apply the method of conjugate sets developed by Gul’ko in [6] and generalized
in [13]. The referee pointed out that this method reminds the technique of norming pairs in Banach spaces used to construct
a projectional resolution of the identity.
Let X be a space and Y ⊂ C p ( X ). Given P ⊂ X and Q ⊂ Y we will say that sets M ⊂ P and L ⊂ Q are ( P , Q )-conjugate
if π M ( L ) = π M ( Q ) and e L ( M ) = e L ( P ); the sets M and L are ( P , Q )-preconjugate if π M ( L ) is dense in π M ( Q ) and e L ( M ) is
dense in e L ( P ). If no confusion is possible, then ( X , Y )-conjugate sets will be called conjugate and ( X , Y )-preconjugate sets
will be simply called preconjugate.

5.1. Proposition. (See [13].) Suppose that X is a space and a set Y ⊂ C p ( X ) generates the topology of X . Assume also that M ⊂ X
and L ⊂ Y are conjugate sets. Then the maps u = e L | M : M → e L ( M ) and v = π M | L : L → π M ( L ) are homeomorphisms; besides,
r = u −1 ◦ e L : X → M and q = v −1 ◦ π M : Y → L are continuous retractions such that q = r ∗ ◦ (π M |Y ). The maps r and q are called
the pair of retractions corresponding to the conjugate pair ( M , L ).

5.2. Proposition. (See [13].) Suppose that X is a Lindelöf Σ -space and a Lindelöf Σ -space Y ⊂ C p ( X ) generates the topology of X .
Assume additionally that countable families P ⊂ exp( X ) and Q ⊂ exp(Y ) are mod–k-networks of the spaces X and Y respec-
tively closed under finite intersections and finite unions. Assume that, for some sets M ⊂ X and L ⊂ Y the pair ( M ∩ P , L ∩ Q ) is
( P , Q )-preconjugate for any ( P , Q ) ∈ P × Q. Then the pair (cl X ( M ), clY ( L )) is conjugate.

5.3. Proposition. Suppose that X is a space and a set Y ⊂ C p ( X ) generates the topology of X . Assume also that P ⊂ exp( X ) and
Q ⊂ exp(Y ) are countable families. Then for any countable set A ⊂ X , it is possible to find countable sets M ( A ) ⊂ X and L ( A ) ⊂ Y such
that A ⊂ M ( A ), the pair ( M ( A ) ∩ P , L ( A ) ∩ Q ) is ( P , Q )-preconjugate for any P ∈ P and Q ∈ Q, and the assignments A → M ( A )
and A → L ( A ) are ω -monotone.

Proof. Let B be a countable base for the real line R. Choose an enumeration {( P n , Q n ): n ∈ ω} of the set P × Q
in which every pair ( P , Q ) ∈ P × Q occurs infinitely many times. Denote by Un the family of all non-empty sets
[x1 , . . . , xk ; B 1 , . . . , B k ] ∩ Q n , where x1 , . . . , xk ∈ X , B 1 , . . . , B k ∈ B and k ∈ N. Let Vn be the family of all non-empty sets
[ f 1 , . . . , f k ; B 1 , . . . , B k ] ∩ e Y ( P n ), where f 1 , . . . , f k ∈ Y , B 1 , . . . , B k ∈ B and k ∈ N. For any n ∈ ω and U ∈ Un pick a function
f U ∈ U ; for each V ∈ Vn fix a point x V ∈ P n such that e Y (x V ) ∈ V .
Take a countable set A ⊂ X . We will recursively construct M ( A ) and L ( A ). Let M 0 ( A ) = A and L 0 ( A ) = ∅. Assume that
n ∈ ω and we have countable sets M 0 ( A ), . . . , M n ( A ) ⊂ X and L 0 ( A ), . . . , L n ( A ) ⊂ Y . Let
 
L n+1 ( A ) = L n ( A ) ∪ f U : U ∈ Un and supp(U ) ⊂ M n ( A ) , and
 
M n+1 ( A ) = M n ( A ) ∪ x V : V ∈ Vn and supp( V ) ⊂ L n+1 ( A ) .
134 R. Rojas-Hernández, V.V. Tkachuk / J. Math. Anal. Appl. 412 (2014) 125–137

Notice that M n+1 ( A ) ⊂ X and L n+1 ( A ) ⊂ Y are countable sets since the sets M n ( A ) and L n ( A ) are countable. So our
inductive
 procedure can be continued
 to construct sequences { M i ( A ): i ∈ ω} and { L i ( A ): i ∈ ω}. We will prove that M ( A ) =
{ Mn ( A ): n ∈ ω} and L ( A ) = { Ln ( A ): n ∈ ω} are as promised.

Claim 1. The pair ( M ( A ) ∩ P , L ( A ) ∩ Q ) is ( P , Q )-preconjugate for any P ∈ P and Q ∈ Q.

Proof. Take an arbitrary pair ( P , Q ) ∈ P × Q and consider the sets M = M ( A ) ∩ P and L = L ( A ) ∩ Q ; we must prove that
the pair ( M , L ) is ( P , Q )-preconjugate.
Take any function f ∈ Q and W ∈ τ (π M ( f ), π M ( Q )). By the definition of the topology of pointwise convergence there
are x1 , . . . , xk ∈ M ⊂ M ( A ) and B 1 , . . . , B k ∈ B such that f ∈ U = [x1 , . . . , xk ; B 1 , . . . , B k ] ∩ Q ⊂ Y and we have π M ( f ) ∈
πM (U ) ⊂ W . It follows from the choice of our enumeration of P × Q and the fact that the sequence { M i ( A ): i ∈ ω} is
increasing, that there is n ∈ ω such that ( P n , Q n ) = ( P , Q ) and {x1 , . . . , xk } ⊂ M n ( A ). Since U ∈ Un , it follows from f U ∈
L n+1 ( A ) ∩ U ⊂ L ( A ) ∩ Q = L that π M ( f U ) ∈ π M (U ) ⊂ W . Therefore π M ( L ) is dense in π M ( Q ).
To show that e L ( M ) is dense in the space e L ( P ) fix any point x ∈ P and a set W ∈ τ (e L (x), e L ( P )). There exist functions
f 1 , . . . , f k ∈ L ⊂ L ( A ) and sets B 1 , . . . , B k ∈ B such that e Y (x) ∈ V = [ f 1 , . . . , f k ; B 1 , . . . , B k ] ∩ e Y ( P ) and e L (x) ∈ π L ( V ) ⊂ W .
It follows from the choice of our enumeration of P × Q and the fact that the sequence { L i ( A ): i ∈ ω} is increasing, that
there exists a number n ∈ ω for which ( P n , Q n ) = ( P , Q ) and { f 1 , . . . , f k } ⊂ L n ( A ). Then V ∈ Vn and it follows from x V ∈
M n ( A ) ∩ P n ⊂ M ( A ) ∩ P = M that e L (x V ) = π L (e Y (x V )) ∈ π L ( V ) ⊂ W . Therefore e L ( M ) is dense in the space e L ( P ). 2

Claim 2. The assignments A → M ( A ) and A → L ( A ) are ω -monotone.

Proof. It takes a straightforward induction to see that the sets M ( A ) and L ( A ) are countable for any countable set A ⊂ X
and the assignments A → M n ( A ) and A → L n ( A ) are ω -monotone
 for every number n ∈ ω . Applying
 Proposition 4.3
we can convince ourselves that the assignments A → M ( A ) = { M n ( A ): n ∈ ω} and A → L ( A ) = { L n ( A ): n ∈ ω} are
ω-monotone. 2

Finally, here the proof of Theorem 3.7 goes.

5.4. Theorem. Assume that X and C p ( X ) are Lindelöf Σ -spaces. Then they are monotonically retractable and monotonically Sokolov.

Proof. The set Y = C p ( X ) generates the topology of X . Choose countable families P ⊂ exp( X ) and Q ⊂ exp(Y ) closed under
finite intersections and finite unions which are mod–k-networks of X and Y respectively.
It follows from Proposition 5.3 that for any countable set A ⊂ X , there exist countable sets M ( A ) ⊂ X and L ( A ) ⊂ Y such
that A ⊂ M ( A ) while the pair ( M ( A ) ∩ P , L ( A ) ∩ Q ) is ( P , Q )-preconjugate for any ( P , Q ) ∈ P × Q, and the assignment
A → L ( A ) is ω -monotone. If F = cl X ( M ( A )) and G = clY ( L ( A )) then the pair ( F , G ) has to be conjugate by Proposition 5.2.
Apply Proposition 5.1 to see that the map u A = e G | F : F → e G ( F ) is a homeomorphism and the map r A = (u A )−1 ◦ e G :
X → F is a continuous retraction.
Since X is a Lindelöf Σ -space, it follows from Proposition 3.20 of [9], that X must be monotonically stable and, in
particular, it is monotonically ω -stable. Therefore, to each countable set L ⊂ Y we can assign a family O ( L ) ⊂ exp( X ) which
is a network of e L in such a way that the operator O is ω -monotone. Since (u A )−1 is a homeomorphism, the family
O( L ( A )) is a network of r A . Finally, let K ( A ) = cl X ( M ( A )) and N ( A ) = O( L ( A )). Then A ⊂ K ( A ) ⊂ X and r A : X → K ( A ) is
a continuous retraction. Besides, N ( A ) is a countable network of r A , and the assignment N is ω -monotone, that is, X is
monotonically retractable.
Since C p (C p ( X )) is also a Lindelöf Σ -space, we can apply the above reasoning to the spaces C p ( X ) and C p (C p ( X )) to
see that C p ( X ) is monotonically retractable as well and hence both X and C p ( X ) have the monotonic Sokolov property by
Theorem 4.14 and Theorem 4.17. 2

The following statement answers Problem 3.8 from [12].

5.5. Corollary. If X and C p ( X ) are Lindelöf Σ -spaces, then they both have the Sokolov property.

Proof. It follows from Theorem 5.4, and Corollary 4.20. 2

5.6. Corollary. If C p ( X ) is a Lindelöf Σ -space then the following conditions are equivalent:

(a) X is a Lindelöf Σ -space;

(b) X is monotonically retractable;
(c) X is monotonically Sokolov;
(d) C p ( X ) is monotonically retractable;
(e) C p ( X ) is monotonically Sokolov.
 R. Rojas-Hernández, V.V. Tkachuk / J. Math. Anal. Appl. 412 (2014) 125–137 135

Proof. It follows from Theorem 5.4 that we have the implications (a) ⇒ (b) and (a) ⇒ (c). Suppose that a space X is
monotonically retractable or monotonically Sokolov. It follows from [8, Theorem 3.5] that υ X must be a Lindelöf Σ -space.
Apply Corollary 4.9 (or Corollary 4.20 respectively), to see that ext ( X )  ω . Now, Baturov’s theorem [2, Theorem 1 ] implies
that X is Lindelöf and hence realcompact, i.e., X = υ X is a Lindelöf Σ -space and hence we proved that (b) ⇒ (a) and
(c) ⇒ (a), i.e., (a) ⇐⇒ (b) ⇐⇒ (c). Theorem 4.14 and Theorem 4.17 show that (b) ⇐⇒ (e) and (c) ⇐⇒ (d) so all
properties (a)–(e) are equivalent. 2

5.7. Corollary. If X is a Gul’ko compact space then C p ( X ) is monotonically retractable.

5.8. Proposition. If X is a Lindelöf Σ -space and C p ( X ) is monotonically retractable then C p ( X ) is Lindelöf and hence t ( X n )  ω for
every n ∈ N.

Proof. Corollary 4.9 shows that ext(C p ( X )) = ω so we can apply Baturov’s theorem [2, Theorem 1 ] to see that C p ( X ) is
Lindelöf and hence t ( X n )  ω for every n ∈ N (see [14, Problem 189]). 2

5.9. Corollary. The space C p (ω1 + 1) is neither monotonically retractable nor monotonically Sokolov.

Example 4.28 shows that monotone retractability of C p ( X ) does not imply that it is Lindelöf. However, the situation
changes if X is a Lindelöf Σ -space.

5.10. Proposition. Suppose that X is a Lindelöf Σ -space; if C p ( X ) is monotonically retractable then it is Lindelöf.

Proof. Since ext (C p ( X )) = ω by Corollary 4.9, we can apply Baturov’s theorem [2, Theorem 1 ] to see that C p ( X ) is Lin-
delöf. 2

5.11. Example. Let K be Reznichenko’s example of a Talagrand compact space such that, for some y ∈ K , the set X = K \{ y }
is pseudocompact and hence K = β X (see Section 8.4 of the book [4]). Then C p ( X ) is a Lindelöf Σ -space which is not
monotonically retractable.

Proof. If π : C p ( K ) → C p ( X ) is the restriction map then π (C p ( K )) = C p ( X ) so the space C p ( X ) has the Lindelöf Σ -property.
Now, the fact that X is not Lindelöf together with Corollary 5.6 show that C p ( X ) is not monotonically retractable. 2

It turns out that it is not so easy to characterize even simple monotonically retractable spaces. So far we have the
following result.

5.12. Theorem. If X is a simple Lindelöf Σ -space then both X and C p ( X ) are monotonically retractable and monotonically Sokolov.

Proof. If the space X is discrete then it is countable and hence our theorem holds trivially; therefore we can assume that
X is uncountable and non-discrete. Observe that X is monotonically Sokolov by Proposition 4.29 so C p ( X ) is monotonically
retractable by Theorem 4.17. Therefore we only need to prove that X is monotonically retractable.
Denote by p the unique non-isolated point of X and let D = X \{ p }. Let { F n : n ∈ ω} be a network of X with respect to a
compact cover C .
Given a countable set A ⊂ X let r A (x) = x for any x ∈ A and r A (x) = p for any x ∈ X \ A. Observe that r A ( p ) = p for any
countable A ⊂ X . If we let K ( A ) = A ∪ { p } then r A : X → K ( A ) is a retraction. Note that the continuity of r A at the point p
follows from the fact that r A (U ) ⊂ U for any U ∈ τ ( p , X ); since all points of X \{ p } are isolated, the map r A is continuous
at all points of X .
Now, let N ( A ) = {{x}: x ∈ A ∪ { p }} ∪ { F n \ B: B is a finite subset of A and n ∈ ω}. It is straightforward that the as-
signment N is ω -monotone. To verify that N ( A ) is a network of r A , take a point x ∈ X and a set U ∈ τ (r A (x), K ( A )). If
x ∈ A ∪ { p } then V = {x} is an element of N ( A ) and r A ( V ) ⊂ U . If x ∈ / A ∪ { p } then r A (x) = p; choose a set C ∈ C with
x ∈ C . The set B = C \U is compact and discrete so it must be finite. Since B ∪ U is a neighborhood of C , there exists n ∈ ω
such that C ⊂ F n ⊂ B ∪ U . Thus x ∈ G = F n \( B ∩ A ) ∈ N ( A ) and r A (G ) ⊂ U so N ( A ) is a network of r A and hence X is
monotonically retractable. 2

6. Open problems

Since monotone Sokolov property and monotone retratability are new concepts, there are more unsolved problems about
them than the solved ones. However, the knowledge we have at the present moment shows that they are interesting and
useful notions. We hope that the following list of open problems illustrates that.
It was proved in [10] that every Corson compact space is monotonically retractable so we start our list with the central
question of this topic (which is also formulated in [10]).
136 R. Rojas-Hernández, V.V. Tkachuk / J. Math. Anal. Appl. 412 (2014) 125–137

6.1. Question. Suppose that X is a monotonically retractable compact space. Must X be Corson compact?

Recall that X is a Collins–Roscoe space (or has the Collins–Roscoe

 property) if for each x ∈ X , we can find a countable family
Px ⊂ exp( X ) such that for every set A ⊂ X , the collection {Px : x ∈ A } is an external network for A. Every Collins–Roscoe
compact space is Corson compact but not necessarily Gul’ko compact while every Gul’ko compact has the Collins–Roscoe
property.

6.2. Question. Suppose that X is a monotonically Sokolov compact space. Must X have the Collins–Roscoe property?

Every Collins–Rocoe space is monotonically monolithic and it is not known whether these properties coincide in com-
pact spaces. Consequently, it is also important to know how the monotone Sokolov property is placed with respect to
monotonically monolithic compact spaces.

6.3. Question. Suppose that X is a monotonically monolithic compact space. Must X be monotonically Sokolov?

It is known that there are examples of Collins–Roscoe compact spaces which are not Gul’ko, so it is natural to ask
whether Collins–Roscoe property implies the monotone Sokolov property.

6.4. Question. Suppose that X is a compact Collins–Roscoe space. Must X be monotonically Sokolov?

It is also important to try to determine the placement of monotone Sokolov compact spaces with respect to Gul’ko
compact ones. So far, we do not even know whether these two classes coincide.

6.5. Question. Suppose that X is a monotonically Sokolov compact space. Must X be Gul’ko compact?

However, we could suspect that monotonically Sokolov compact spaces might share some properties with Gul’ko compact
ones. For example, it is a classical theorem that every Gul’ko compact space has a dense metrizable subspace so a positive
answer to the following question would strengthen it.

6.6. Question. Suppose that X is a monotonically Sokolov compact space. Must X have a dense metrizable subspace?

If the previous question turns out to be too difficult then its weaker version, that follows, might be solvable.

6.7. Question. Suppose that X is a monotonically Sokolov compact space and c ( X )  ω . Must X be metrizable?

It was an open question for several years whether Sokolov spaces of countable character have cardinality at most c.
The second author gave a counterexample recently. The respective question has an easy answer for monotonically Sokolov
spaces because they are Lindelöf. However, it is a famous open problem of Arhangel’skii whether every Lindelöf space of
countable pseudocharacter has cardinality  c.

6.8. Question. Suppose that X is a monotonically Sokolov space and ψ( X )  ω . Is it true that | X |  c?

The following question is motivated by Theorem 5.12.

6.9. Question. Suppose that X is a simple monotonically retractable space. Must X have the Lindelöf Σ -property?

Since monotone Sokolov property is a very strong restriction on a space, it is natural to ask what happens if we have it
hereditarily. The following question is inspired by the fact that hereditarily Lindelöf Σ -spaces have countable network.

6.10. Question. Suppose that X is hereditarily monotonically Sokolov. Must X have a countable network?

References

[1] D. Amir, J. Lindenstrauss, The structure of weakly compact sets in Banach spaces, Ann. of Math. 88 (1) (1968) 35–46.
[2] D.P. Baturov, On subspaces of function spaces, Vestnik Moskov. Univ. Ser. I Mat. Mekh. 42 (4) (1987) 66–69 (in Russian).
[3] R. Engelking, General Topology, PWN, Warszawa, 1977.
[4] M. Fabian, Gateaux Differentiability of Convex Functions and Topology, Weak Asplund Spaces, Wiley, New York, 1997.
[5] G. Gruenhage, Monotonically monolithic spaces, Corson compacts, and D-spaces, Topology Appl. 159 (2012) 1559–1564.
[6] S.P. Gul’ko, On the structure of spaces of continuous functions and their complete paracompactness, Russian Math. Surveys 34 (6) (1979) 36–44.
 R. Rojas-Hernández, V.V. Tkachuk / J. Math. Anal. Appl. 412 (2014) 125–137 137

[7] R.E. Hodel, Cardinal Functions I, in: K. Kunen, J.E. Vaughan (Eds.), Handbook of Set-Theoretic Topology, North Holland, Amsterdam, 1984, pp. 1–61.
[8] O.G. Okunev, On Lindelöf Σ -spaces of continuous functions in the pointwise topology, Topology Appl. 49 (2) (1993) 149–166.
[9] R. Rojas-Hernández, A note on monotone monolithicity, submitted for publication.
[10] R. Rojas-Hernández, Function spaces and D-property, Topology Proc. 43 (2014) 301–317.
[11] R. Rojas-Hernández, A. Tamariz-Mascarúa, D-property, monotone monolithicity and function spaces, Topology Appl. 159 (2012) 3379–3391.
[12] V.V. Tkachuk, A nice class extracted from C p -theory, Comment. Math. Univ. Carolinae 46 (3) (2005) 503–513.
[13] V.V. Tkachuk, Condensing function spaces into Σ -products of real lines, Houston J. Math. 33 (1) (2007) 209–228.
[14] V.V. Tkachuk, A C p -Theory Problem Book. Topological and Function Spaces, Springer, New York, 2011.
[15] V.V. Tkachuk, Lifting the Collins–Roscoe property by condensations, Topology Proc. 42 (2013) 1–15.