ZvinaAGT.dvi


@
Applied General Topology

c© Universidad Politécnica de Valencia

Volume 7, No. 1, 2006

pp. 51-66

On i-topological spaces: generalization of
the concept of a topological space via ideals

Irina Zvina

Abstract. The aim of this paper is to generalize the structure

of a topological space, preserving its certain topological properties.

The main idea is to consider the union and intersection of sets modulo

“small” sets which are defined via ideals. Developing the concept of

an i-topological space and studying structures with compatible ideals,

we are concerned to clarify the necessary and sufficient conditions for

a new space to be homeomorphic, in some certain sense, to a topological

space.

2000 AMS Classification: 54A05, 54E99.

Keywords: compatible ideal, generalization, topological space.

1. Introduction

The use of the ideals in general topology historically developed along two
main lines. The first line is concerned with the study of the local properties
of topological spaces that may be extended to the global properties [7, 12].
The central concept in these investigations is the compatibility of an ideal with
a topology. For example, the well-known Banach Category Theorem is an im-
mediate corollary from the σ-Extension Theorem [7]. The works of the second
line use ideals to generalize the certain properties of topological spaces, such
as a compactness[9, 10, 11] and the separation axioms[1].

The aim of this paper is to generalize, via ideals, the concept of a topological
space itself, thus it differs from the works of two mentioned lines. However, we
make use of the ideas, developed in those works.

In the first section, we define an i-topological space, provide basic exam-
ples, introduce the notions of a subspace, an i-continuous mapping, a scattered
set and a nowhere dense set for such spaces and study properties of these no-
tions. We propose a sufficient condition for the existence of a supratopological
space [8] that is i-homeomorphic to a given i-topological space. In the second



52 I. Zvina

section, we introduce the concept of a unified i-topological space and discuss
two important examples.

In the last section, we study some properties of i-topological spaces with
compatible ideals. We propose a construction which shows that for every such
space there exists an i-homeomorphic supratopological space such that for ev-
ery two its i-open subsets U and V there is an element A in the ideal for
which (U ∩ V ) \ A is i-open. On the other hand, we show that the compatibil-
ity is necessary for the existence of a topological space that is i-homeomorphic
to a given i-topological space. We introduce a sufficient condition for the ex-
istence of such a space and leave as an open question whether there exists
an i-topological space with a compatible ideal which is not i-homeomorphic to
any topological space.

We refer a reader to the papers [3]-[6] by Dragan Janković and T.R. Hamlett
for the survey on the use of ideals in topological spaces. These works also
provide the historical background and contain many references to the related
papers.

2. General notions of i-topological spaces

2.1. i-Topological spaces. Subspaces. A nonempty family I of subsets
of a set X is called an ideal on X iff X /∈ I, A,B ∈ I implies A ∪ B ∈ I, and
A ∈ I and B ⊆ A imply B ∈ I. We define the relations ≤ and ≈ on P(X) as
follows:

(1) A ≤ B iff A \ B ∈ I,
(2) A ≈ B iff (A \ B) ∪ (B \ A) ∈ I,

where A,B ⊆ X. One can easily check that ≤ and ≈ are a preorder and
an equivalence on P(X), respectively.

Definition 2.1. Let X be a set and I be an ideal on X. Then an i-topology on X
is a family T of subsets of X that satisfies the following conditions:

(T1) ∅,X ∈ T ;
(T2) for any U ⊆ T there exists U ∈ T such that

⋃

U ≈ U;
(T3) for any V,W ∈ T there exists U ∈ T such that V ∩W ≈ U;
(T4) T ∩ I = {∅}.

The triple (X,T ,I) is called an i-topological space and the elements of T are
called i-open sets. We will use the notation T (x) = { U ∈ T | x ∈ U } for
any x ∈ X.

Let us give some basic examples.

Example 2.2. Suppose we are given a topological space (X,T ). Then (X,T ,{∅})
and (X,Tr,In) are i-topological spaces, where Tr is the family of all regular
open sets and In is the family of all nowhere dense sets.

For the next example we need the following lemma.



On i-topological spaces 53

Lemma 2.3. Let (X,T ) be a topological space and C ⊆ X. Then there ex-
ist A,B ⊆ X and an open set U such that C = A ∪ B, the set A is nowhere
dense and cl(B) = cl(U).

Proof. It is sufficient to take U = int(cl(C)), B = C∩U and A = C\B. Then A
is nowhere dense, since A = C \ B = C \ (C ∩ int(cl(C))) = C \ int(cl(C)).

To complete the proof, it remains to show that Ux ∩ B 6= ∅ for each x ∈ U
and any its open neighborhood Ux. Observe that Ux ∩ U is also an open
neighbourhood of x. Since x belongs to the closure of C, it follows that Ux ∩
U ∩ C 6= ∅ and hence Ux ∩ B 6= ∅. �

Example 2.4. Consider the real line R with the usual topology Tu and the fam-
ily In of all nowhere dense sets. Then the triple (R,D,In) is an i-topological
space, where D = { A ⊆ R | there exists U ∈ Tu such that cl(A) = cl(U) }.
Obviously, this can be generalized to any topological space. However, in our
paper we will only use to the example based on the real line with the usual
topology.

As we know the intersection of a family of ideals on the same set is an ideal.
We provide an example where a family T is an i-topology with respect to two
different ideals, but is not an i-topology with respect to their intersection.

Example 2.5. Let X be the union of X1 = R × {1} and X2 = R × {2}. We
construct a family T of subsets of X as follows:

T = {∅,X} ∪ T1 ∪ T2,

where

T1 = { {x} ∪ B | x ∈ X1 and B ⊆ X2 } ,

T2 = {A ∪ {y} | A ⊆ X1 and y ∈ X2 } .

Then T is an i-topology on X with respect to the ideals P(X1) and P(X2).
However, T is not an i-topology with respect to {∅}.

Now, our aim is to define the notion of a subspace. Suppose we are given
an i-topological space (X,T ,I) and a subset M ⊆ X. We will use the following
notations:

T |M = { M ∩ U | U ∈ T and M ∩ U /∈ I } ∪ {∅,M},

if M /∈ I then

I|M = { M ∩ A | A ∈ I }

and if M ∈ I then

I|M = {∅}.

The triple (M,T |M,I|M ) is called a subspace of the i-topological space (X,T ,I).

Proposition 2.6. Every subspace of any i-topological space is an i-topological
space.



54 I. Zvina

Proof. Fix an i-topological space (X,T ,I) and its subspace (M,T |M,I|M ).
Let us prove that I|M is an ideal. If A,B ∈ I then (M ∩A)∪(M ∩B) = M ∩

(A∪B) ∈ I|M . Hence, we have finite additivity. Now, if A ∈ I and B ⊆ M ∩A
then B ⊆ A and by heredity B ∈ I. Hence, B = M ∩ B ∈ I|M and we have
heredity.

Let us prove that (M,T |M,I|M ) is an i-topological space. It follows from
the definition, that ∅,M ∈ T |M . Hence, we have (T1).

Suppose that C,D ⊆ X and C ≈ D. Then M ∩ C ≈I|M M ∩ D. Indeed,
there exist A,B ∈ I such that D = (C ∪A)\B, since C ≈ D. Hence, M ∩D =
M ∩ ((C ∪ A) \ B) = ((M ∩ C) ∪ (M ∩ A)) \ (M ∩ B) ≈I|M M ∩ C.

It follows from the definition of subspace, that for any family V = {Vs}s∈S
of the elements of T |M there exists a family U = {Us}s∈S of the elements of T
such that Vs = M ∩ Us for each Vs ∈ V. Take U from T such that

⋃

U ≈ U.
Then

⋃

V = M∩(
⋃

U ) and by the previous paragraph M∩(
⋃

U ) ≈I|M M∩U.
Therefore,

⋃

V ≈I|M M ∩ U.
If V and U are finite, we take W ∈ T such that W ≈

⋂

U and in the same
way show that

⋂

U ≈I|M M ∩ W . Thus, we have (T2) and (T3).
To complete the proof, it remains to show that T |M and I|M satisfy (T4).

Suppose that V ∈ T |M ∩ I|M . Then it follows from the definition of subspace,
that V = ∅, since I|M ⊆ I. �

2.2. i-Continuous mappings. Suppose we are given two i-topological spaces
(X,TX,IX) and (Y,TY ,IY ). We say that a mapping f : X → Y is i-continuous
if it satisfies the following conditions:

(N1) for every family {Vs}s∈S of i-open subsets of Y satisfying

Vs ∩ f(X) /∈ IY

for each s ∈ S there exists a family {Us}s∈S of nonempty
i-open subsets of X such that

Us ≈ f
−1(Vs)

holds for each s ∈ S and

⋃

s∈S

Us ≈ f
−1

(

⋃

s∈S

Vs

)

;

(N2) f−1IY ⊆ IX.

Notice that our definition generalizes the usual definition of continuous map-
ping. Indeed, if we consider topological spaces with the ideal {∅} and in (N1)
examine families which consist of only one i-open set, we obtain the usual
definition.

The mapping f is called i-homeomorphism if f is a bijection and both f
and f−1 are i-continuous. Two spaces are called i-homeomorphic if there exists
an i-homeomorphism of one space to the other. The following proposition is
a natural consequence from the definition of i-continuous mapping.



On i-topological spaces 55

Proposition 2.7. Let f : X → Y and g : Y → Z be i-continuous mappings
of i-topological spaces. Assume that for any i-open set in Y its intersection
with f(X) does not lie in the ideal of the space Y . Then the composition g ◦ f
is an i-continuous mapping.

Suppose we are given a family F = {(X,Ts,Is)}s∈S of i-topological spaces.
An i-topological space (X,TX,IX) is called minimal in the family F if it is
an element of F and the identity mapping id: (X,Ts,Is) → (X,TX,IX) is
i-continuous for each s ∈ S.

Let fs : X → Ys and (Ys,Ts,Is) be a mapping and an i-topological space for
each s ∈ S, respectively. Then (X,TX,IX) is called an initial i-topological space
generated by {fs}s∈S if it is minimal in the following family of i-topological
spaces { (X,T ,I) | fs : X → Ys is i-continuous for each s ∈ S }.

The next proposition shows that our definition of continuity generalizes
the usual definition in the sense of initial topology. We omit the proof.

Proposition 2.8. Let (Y,T ,I) be an i-topological space and f : X → Y be
a mapping. Consider families TX = f

−1T |f(X) and IX = f
−1I|f(X). Then

(X,TX,IX) is an initial i-topological space generated by {f}.

2.3. Scattered and nowhere dense sets. First, let us recall the definition
of the set operator ψ [3]. The various properties of this operator are to be found
in [3]. Suppose we are given a topological space (X,T ) and an ideal I on X.
Then ψ : P(X) → P(X) is defined as follows. For any A ⊆ X,

ψ(A) = { x ∈ X | there exists U ∈ T (x) such that U \ A ∈ I }.

We use this definition for i-topological spaces. One can easily check that ψ(A) =
⋃

{ U ∈ T | U ≤ A} holds for any A ⊆ X. If there is a chance for confusion,
we will use indexes to specify the ideal or i-topology in the way we do it in
Proposition2.10 and Lemma 2.12.

Now, let (X,T ,I) be an i-topological space and A be a subset of X. We
say that x ∈ X is an isolated point of A if there exists U ∈ T (x) such that A ∩
U \ {x} ∈ I. We say that A is scattered if every its subset contains an isolated
point. We say that A ⊆ X is nowhere dense if ψ(A ∪ B) = ∅ for each B ⊆ X
such that ψ(B) = ∅.

In what follows we will use the notations s(I,T ) = { A ⊆ X | A is scattered }
and n(I,T ) = { A ⊆ X | A is nowhere dense }. Clearly, a subset is scattered or
nowhere dense with respect to a certain i-topology and ideal. However, in case
there is no chance for confusion, we will simply say scattered or nowhere dense
and write s(I) and n(I) instead of s(I,T ) and n(I,T ).

What is stated in the following proposition is an easy corollary from the
definition of a scattered set. The result of Proposition2.10 will be used later
in the next section.

Proposition 2.9. Let (X,T ,I) be an i-topological space. Then s({∅}) ⊆ s(I)
and I ⊆ s(I).



56 I. Zvina

Proposition 2.10. Let (X,T1,I1) and (X,T2,I2) be i-topological spaces. As-
sume that, for any subset A ⊆ X, it holds that ψ1(A) = ∅ iff ψ2(A) = ∅ holds.
Then n(I1,T1) = n(I2,T2).

Proof. Take B ∈ n(I1,T1) and A ⊆ X such that ψ2(A) = ∅. Then by our
assumption ψ1(A) = ∅ and hence ψ1(A ∪ B) = ∅, since B ∈ n(I1,T1). Thus,
under the assumption ψ2(A∪B) = ∅ and we have B ∈ n(I2,T2). In the similar
way we can show that B ∈ n(I2,T2) implies B ∈ n(I1,T1) for each B ⊆ X. �

The converse statement of the previous Lemma does not hold. Indeed, con-
sider the i-topological spaces (R,Tu,In) and (R,D,In) from Example 2.2 and
Example 2.4. It is not difficult to check that In = n(In,Tu) = n(In,D). How-
ever, ψTu(Q) = ∅ and ψD(Q) 6= ∅.

The next proposition shows that in an i-topological space the ideal can be
extended over the family of all nowhere dense sets such that together with
the given i-topology it will satisfy the conditions of an i-topological space.
Moreover, this process is finite.

Proposition 2.11. Let (X,T ,I) be an i-topological space. Then

(i) I ⊆ n(I);
(ii) n(I) is an ideal;
(iii) n(I) ∩ T = {∅};
(iv) n(n(I)) = n(I).

Proof. Statement (i) is obvious. Let us prove (ii). Suppose that A, B and C are
subsets of X such that A and B are nowhere dense and ψ(C) = ∅. Then ψ(B∪
C) = ∅ and ψ(A ∪ (B ∪ C)) = ψ((A ∪ B) ∪ C) = ∅. Hence, A ∪ B is nowhere
dense. On the other hand, a subset of a nowhere dense set is nowhere dense.
Hence, we have finite additivity and heredity.

Suppose that U ∈ n(I) ∩ T . Then ψ(U ∪ ∅) = ψ(U) = ∅, since ψ(∅) = ∅.
On the other hand, ψ(V ) 6= ∅ for each nonempty V ∈ T . Hence, U = ∅ and
we have (iii).

The last statement follows from Lemma 2.12. �

Lemma 2.12. Let (X,T ,I) be an i-topological space and A ⊆ X. Then ψI(A) =
∅ iff ψn(I)(A) = ∅ for each A ⊆ X.

Proof. Suppose that A is a subset of X, ψI(A) = ∅ and U is an i-open set
such that U ⊆ ψn(I)(A). Then there exists B ∈ n(I) such that U ⊆ A ∪ B.
Therefore, U ⊆ ψI(A∪B). On the other hand, ψI(A∪B) = ∅, since ψI(A) = ∅
and B is nowhere dense. Thus, U = ∅ and ψn(I)(A) = ∅.

Now, notice that ψI(A) ⊆ ψn(I)(A) follows from I ⊆ n(I) for any sub-
set A ⊆ X. Hence, ψn(I)(A) = ∅ implies ψI(A) = ∅ for any A ⊆ X. �

2.4. Existence of a supratopological space that is i-homeomorphic
to a given i-topological space. Let us recall that (X,T ), where X is a set
and T is a family of subsets of X, is called supratopological space if X ∈ T
and

⋃

U ∈ T for each U ⊆ T [8].



On i-topological spaces 57

Suppose we are given an i-topological space (X,T ,I). Then an opera-
tion α: P(T ) → T such that α(U) ≈

⋃

U holds for each U ⊆ T is called
associative if it satisfies

α({ α( Us ) | s ∈ S }) ≈ α
(

⋃

s∈S
Us

)

for each collection {Us}s∈S of families of i-open sets. To simplify our notations,
we write αs∈S(α(Us)) instead of α({ α( Us ) | s ∈ S }).

In what follows, we will use the notation T ∪ = {
⋃

U | U ⊆ T }.

Proposition 2.13. Let (X,T ,I) be an i-topological space. Then (X,T ∪,I)
is an i-topological space. If there exists an associative operation α: P(T ) → T
such that α(U) ≈

⋃

U for each U ⊆ T then the spaces (X,T ,I) and (X,T ∪,I)
are i-homeomorphic.

Proof. Obviously, (T1) and (T4) are satisfied for the triple (X,T ∪,I) and
(T2) is satisfied, since the family T ∪ is closed under the arbitrary unions.
Let us prove (T3). Suppose that U,V ∈ T ∪. Then there exist U,V ⊆ T
and U1,V1 ∈ T such that U =

⋃

U ≈ U1 and V =
⋃

V ≈ V1. Take W ∈ T
satisfying W ≈ U1 ∩ V1. Then W ≈ U ∩ V and (T3) is proved.

To prove that the spaces (X,T ,I) and (X,T ∪,I) are i-homeomorphic un-
der the assumption that there exists an associative operation α: P(T ) → T ,
such that α(U) ≈

⋃

U for each U ⊆ T , we have to show that the identity
mapping id: X → X satisfies (N1). Fix some U ⊆ T ∪. By the definition, there
exists a family VU ⊆ T , for each U ∈ U, such that U =

⋃

VU. Consider the
union V =

⋃

U∈U VU of all these families. Clearly, V ⊆ T . Then U ≈ α(VU ),
for each U ∈ U, and

⋃

U ≈
⋃

V ≈ α(V) ≈ αU∈U(α(VU )) ≈
⋃

U∈U α(VU ).
The proof is complete. �

3. Unified i-topological spaces

In this section, we will make use of the notion of a compatible ideal. Let
us recall the definition [6]. In a topological space (X,T ) an ideal I of subsets
of X is said to be compatible with T , denote I ∼ T , if it satisfies the following
condition for every subset A of X and every subfamily U of T

if A ⊆
⋃

U and U ∩ A ∈ I holds for every U ∈ U then A ∈ I.

We preserve this definition for i-topologies: in i-topological space, the ideal is
compatible with the i-topology if the above property is satisfied.

An i-topological space (X,T ,I) is called unified if U ≈ V implies U = V
for each i-open U and V . A triple (X,T1,I) is called a unification of (X,T ,I)
if X ∈ T1 and for any U ∈ T the set { V ∈ T1 | U ≈ V } contains exactly one
element.

We omit the proofs of the results from the following lemma. The statement
about the nowhere dense sets is an easy consequence from Proposition2.10.



58 I. Zvina

Proposition 3.1. Let (X,T ,I) be an i-topological space and (X,T1,I) be its
unification. Then the following statements hold:

(i) (X,T1,I) is a unified i-topological space;
(ii) n(I,T ) = n(I,T1).

If in additional I ∼ T then:

(iii) I ∼ T1;
(iv) (X,T ,I) and (X,T1,I) are i-homeomorphic.

The easiest way to obtain an i-homeomorphic unified i-topological space for
a given i-topological space (X,T ,I) with a compatible ideal is to choose one
element from each class of equivalence which are considered with the respect
to the equivalence relation ≈. Notice that we can construct such i-topology
without using the Axiom of Choice. Indeed, we take

⋃

{ V ∈ T | U ≈ V } for
any U ∈ T . The compatibility of the ideal implies that the union of any class
of equivalence is equivalent, in the sense of ≈, to each its member. In the last
section of this paper we will propose another way how to obtain such i-topology
without using the Axiom of Choice (see Proposition4.12).

Example 3.2. Consider the real line with the usual topology. Clearly, the par-
tial orders ⊆ and ≤, where ≤ is generated by the ideal of all nowhere dense sets,
are not equivalent. Indeed, A ≤ B does not imply A ⊆ B for each A,B ⊆ R.
Moreover, we will show that there does not exist family F of subsets of R such
that F contains at least one element from each equivalence class generated by ≤,
i.e. F ∩ [A ] = { B ⊆ R | A ≈ B } 6= ∅, and the condition A ≤ B implies A ⊆ B
for each A,B ∈ F.

In the process of proving of this statement we will show that in any unifi-
cation of the i-topological space (R,D,In) from Example 2.4 the ideal is not
compatible with the i-topology. First, let us prove the following lemmas.

Lemma 3.3. Consider the real line with the usual topology and the fam-

ily D from Example 2.4. For any nonempty A ⊆ R, whose closure is equal
to the closure of some nonempty open set, there exist non-empty B,C ∈ D
such that cl(A) = cl(B) = cl(C), B ∪ C ⊆ A and B ∩ C = ∅.

Proof. Take U ∈ Tu such that cl(A) = cl(U). To simplify the proof, we suppose
that U = (0,1). In all other cases, the proof will be similar. Let us define
the sets {Bn}n∈N and {Cn}n∈N as follows. Put B0 = ∅ and C0 = ∅. Then

Bn = Bn−1 ∪

(

⋃2n−1

k=1
{bnk}

)

and Cn = Cn−1 ∪

(

⋃2n−1

k=1
{cnk}

)

where

bnk ∈ I
n
k ∩ (A1 \ (Bn−1 ∪ Cn−1)), c

n
k ∈ I

n
k ∩ (A1 \ (Bn ∪ Cn−1)),

Ink =

(

k − 1

2n
,
k + 1

2n

)

, A1 = A \ {0,1} and k ∈ {1, . . . ,2
n−1}.

Clearly, A1 is dense in U. Thus, a complement in A1 of a union of two finite
sets is dense in U. The sets Bn and Cn are finite for any n ∈ N. Hence,



On i-topological spaces 59

the intersections in the middle line above are not empty for any valid n and k.
Therefore, we can take points bn+1

k
and cn+1

k
for any k ∈ {1, . . . ,2n−1}. We

define the sets B and C as follows: B =
⋃

n∈N Bn and C =
⋃

n∈N Cn.
Clearly, B ∪ C ⊆ A and B ∩ C = ∅. To complete the proof, it remains

to show that B and C are dense in U. Fix x ∈ U and ε > 0. We have to
show that there exist n,k ∈ N such that In

k
⊆ (x − ε,x + ε). Take n ∈ N such

that 1
2n−1

< ε. Now, we can choose the necessary k, since intervals In
k
cover U,

their lengths are equal to 1
2n−1

and length of two intervals is less than 2ε. �

Lemma 3.4. Consider the i-topological space (R,D,In) from Example 2.4.
Then in any its unification the ideal is not compatible with an i-topology.

Proof. Suppose that (R,D1,In) is a unification of (R,D,In). Take U = (0,1).
Define the sets {Bn}n∈N and {C

n}n∈N as follows. Put C
0 = U. Then take Bn

and Cn such that they are dense in U, Bn ∩ Cn = ∅ and Bn ∪ Cn ⊆ Cn−1 for
each n ∈ N. We can choose such Bn and Cn, since the previous Lemma holds.
Now, take B̂n ∈ D1 such that B̂

n ≈ Bn for each n ∈ N. Then B̂n∪Cn ≤ Cn−1.
Define A as follows:

A =
⋃

n∈N
An, where A0 = ∅, An =

⋃2n−1

k=1
{ank},

ank ∈ I
n
k ∩

(

B̂n \

(

⋃n−1

k=1
B̂k
))

and k ∈ {1, . . . ,2n−1}.

Then A is dense in U and hence A /∈ I. Clearly, A ⊆
⋃

n∈N B̂
n and A∩B̂n ∈ I

for each n ∈ N. Thus, In ≁ D1. �

Lemma 3.5. Consider the i-topological space (R,D,In) from Example 2.4.
Assume that (R,D1,In) is its unification and U ≤ V implies U ⊆ V for
each U,V ∈ D1. Then In ∼ D1.

Proof. Fix U ∈ D1 and V ⊆ D1 such that U ∩V ∈ In for each V ∈ V. It follows
from Lemma 2.3, that there exists W ∈ D1 such that U ∩W ∈ I and V ≤ W for
each V ∈ V. Hence,

⋃

V ⊆ W and (
⋃

V ) ∩ U ∈ In. Observe that In = n(In).
Then In ∼ D1 is an immediate corollary from Lemma 4.4. �

Comparing the results of Lemma 3.4 and Lemma 3.5, we conclude that in any
unification (R,D1,In) of the i-topological space (R,D,In) the condition U ≤ V
does not imply U ⊆ V for each U,V ∈ D1.

To complete the proof that there does not exist family F of subsets of R such
that F contains at least one element from each equivalence class generated by ≤
and the condition A ≤ B implies A ⊆ B for each A,B ∈ F it remains to observe
that for any F there exists a unification of (R,D,In) such that its i-topology
is a subfamily of F. Thus, by the result from the previous paragraph such
family F does not exist.

Example 3.6. Consider the real line R with the usual topology Tu. Clearly,
an i-topological space (R,Tu,I) is unified iff I = {∅}. Now, let us assume
that I is an ideal of subsets of R such that

⋃

I = R and there exists another



60 I. Zvina

topology T on the real line, which is a unified i-topology with respect to I,
and the spaces (R,Tu,I) and (R,T ,I) are i-homeomorphic. Then (R,T ,I)
possesses the following properties.

(i) U ≤ V implies U ⊆ V for each U,V ∈ T .
(ii) For any x ∈ R, there exist U,V ∈ T such that U ≈

(−∞,x), V ≈ (x,+∞), x ∈ U ∪ V and x /∈ U ∩ V .
(iii) Define two mappings f : R → T and g : R → {0,1} as

follows: for any x ∈ R, f(x) ≈ (−∞,x), and g(x) = 0,
if x /∈ f(x), or g(x) = 1, if x ∈ f(x). The both mappings
are defined correctly.

(iv) For any x ∈ R satisfying g(x) = 1, there exists δ(x) > 0
such that x ∈ f(y) for each y ∈ (x −δ(x),x) and x /∈ f(y)
for each y ∈ (−∞,x − δ(x)). We allow δ(x) to be equal
to +∞.

(v) At least one of the two sets A = { x ∈ R | g(x) = 1 }
and B = { x ∈ R | g(x) = 0 } is dense.

(vi) If A is dense then for any x ∈ R satisfying g(x) = 1 there
exist y ∈ (x−δ(x),x) and a decreasing sequence (yn)n∈N ⊆
(y,x) such that (yn)n∈N converges to y and (yn)n∈N ⊆
f(y).

However, we leave as an open question whether there exists such topological
space (R,T ,I).

4. i-Topological spaces with compatible ideals

In this section, we concentrate our attention on i-topological spaces with
compatible ideals. The corresponding definition is to be found at the beginning
of the second section.

4.1. Some general properties. The propositions of this subsection provide
some properties of i-topological spaces with compatible ideals that are well-
known for topological spaces[2].

Suppose we are given an i-topological space (X,T ,I). Then an opera-
tion β : P(T ) → T such that β(U) ≈

⋂

U for each finite U ⊆ T is called
distributive over the union operation if it satisfies

⋃

V ∈V
β({U,V }) ≈ U ∩

(

⋃

V

)

for each i-open set U and each collection of i-open sets V.

Proposition 4.1. Let (X,T ,I) be an i-topological space satisfying I ∼ T .
Then:

(i)
⋃

U ≈
⋃

V holds for any two families of i-open sets U =
{Us}s∈S and V = {Vs}s∈S which satisfy Us ≈ Vs for
each s ∈ S;

(ii) any operation α: P(T ) → T which satisfies α(U) ≈
⋃

U

for each U ⊆ T is associative;



On i-topological spaces 61

(iii) any operation β : P(T ) → T which satisfies β(U) ≈
⋂

U

for each finite U ⊆ T is distributive over the union oper-
ation.

Proof. The proofs of all statements are not difficult, so we demonstrate just
one of them. Let us prove (iii).

Suppose we are given an i-open set U and a family of i-open sets V. Clearly,
β({U,V }) ≤ U and β({U,V }) ≤

⋃

V for each V ∈ V. Since I is compatible
with T , it follows that

⋃

V ∈V β({U,V }) ≤ U and
⋃

V ∈V β({U,V }) ≤
⋃

V.
Therefore,

⋃

V ∈V β({U,V }) ≤ U ∩ (
⋃

V).
On the other hand, if we take A = (U ∩(

⋃

V))\(
⋃

V ∈V β({U,V })) then A ⊆
⋃

V and V ∩ A ∈ I for each V ∈ V. Since I is compatible with T , it follows
that A belongs to the ideal. Thus,

⋃

V ∈V β({U,V }) ≤ U ∩ (
⋃

V) and we
complete the proof. �

A family B ⊆ T is called a base for the i-topology T if for any U ∈ T there
exists U ⊆ B such that U ≈

⋃

U. A family C ⊆ T is called a subbase for
the i-topology T if there exists B ⊆ T such that B is a base for T and for
any V ∈ B there is a finite family V ⊆ C such that V ≈

⋂

V.
We omit the proof of the following proposition.

Proposition 4.2. Let (X,TX,IX) and (Y,TY ,IY ) be i-topological spaces with
compatible ideals and f : X → Y be a mapping such that fIX = IY . Then
the following conditions are equivalent:

(i) f is i-continuous;
(ii) inverse image of any member of a subbase C for TY such

that its intersection with f(X) does not lie in IY is equiv-
alent to some nonempty i-open set from TX;

(iii) inverse image of any member of a base B for TY such that
its intersection with f(X) does not lie in IY is equivalent
to some nonempty i-open set from TX;

(iv) inverse image of any i-open set from TY such that its in-
tersection with f(X) does not lie in IY is equivalent to
some nonempty i-open set from TX.

4.2. Scattered and nowhere dense sets. The aim of this subsection is to
show some properties of families of scattered and nowhere dense sets in i-topological
spaces with compatible ideals.

Proposition 4.3. Let (X,T ,I) be an i-topological space such that
⋃

I = X
and I ∼ T . Then I = s(I).

Proof. The inclusion I ⊆ s(I) is obvious. Let us prove s(I) ⊆ I. Consider A ∈
s(I). Let B ⊆ A be a set of all points of A which are isolated from A. Then B ∈
I. By the definition of scattered set there exists a point x ∈ A \ B which is
isolated from A \ B. Since A ≈ A \ B, it follows that any point isolated
from A \ B is isolated from A, too. Then A \ B = ∅ and hence A ∈ I. �



62 I. Zvina

Lemma 4.4. Let (X,T ,I) be an i-topological space. Then the following state-
ments are equivalent:

(i) if A ⊆ X and there exists U ⊆ T such that A ⊆
⋃

U

and U ∩ A ∈ I for each U ∈ U then A is nowhere dense;
(ii) if V ∈ T and there exists U ⊆ T such that U ∩ V ∈ I for

each U ∈ U then (
⋃

U ) ∩ V is nowhere dense;
(iii) if A ⊆ X and there exists U ⊆ T such that A ⊆

⋃

U

and U ∩ A is nowhere dense for each U ∈ U then A is
nowhere dense.

Proof. To prove that (i) implies (ii), it is sufficient to observe that the condition
from (i) is satisfied for

⋃

U∈U(V ∩ U).
Let us prove that (ii) implies (iii). Suppose that B ⊆ X and ψ(B) = ∅.

Then ψ((A ∩ U) ∪ B) = ∅ for each U ∈ U. Take V ∈ T such that V ≤ A ∪ B.
Then U ∩ V ≤ (A ∩ U) ∪ B and hence U ∩ V ∈ I. It follows form (ii),
that (

⋃

U ) ∩ V is nowhere dense. Under the assumption that ψ(B) = ∅ we
conclude that V = ∅, since V ≤ ( (

⋃

U ) ∩ V ) ∪ B.
The fact that (iii) implies (i) is obvious, since every element of an ideal is

nowhere dense. �

The next corollary is an immediate consequence from the previous Lemma.

Corollary 4.5. Let (X,T ,I) be an i-topological space. Then n(I) ∼ T if one
of the following conditions is satisfied:

(1) I ∼ T ;
(2) U ∩ V ∈ I implies U ∩ V = ∅ for each U,V ∈ T ;
(3) there exists an operation β : P(T ) → T such that β(U) ≈

⋂

U for each finite U ⊆ T which is distributive over the union
operation.

4.3. Existence of a topological space that is i-homeomorphic to
a given i-topological space. In this subsection, we provide necessary and
sufficient conditions for the existence of a topological space that is i-homeomorphic
to a given i-topological space.

In what follows, we will need the next lemma.

Lemma 4.6. Let (X,TX,IX) and (Y,T ,I) be i-topological spaces, f : X → Y
be a surjective i-continuous mapping such that fIX = I. Assume that IX ∼
TX. Then I ∼ T .

Proof. Consider a subset A ⊆ Y and a family V = {Vs}s∈S of i-open subsets
of Y such that A ⊆

⋃

V and A ∩ Vs ∈ I for each Vs ∈ V. Then there ex-
ists a family U = {Us}s∈S of i-open subsets of X such that f

−1(
⋃

V) ≈
⋃

U

and f−1(Vs) ≈ Us for each Us ∈ U. Clearly, f
−1(A) ∩ Us lies in the ideal IX

for every s ∈ S. Then it follows from IX ∼ TX, that f
−1(A) ∩ (

⋃

U) ∈
IX. On the other hand, f

−1(A) ⊆ f−1(
⋃

V) and f−1(
⋃

V) ≈
⋃

U. Conse-
quently, f−1(A) ∈ IX and by the assumption A ∈ I. �



On i-topological spaces 63

Now, we refer the reader to the paper by Dragan Janković and T.R. Hamlett [6].
Let (X,T ) be a topological space and I be an ideal on X satisfying T ∩I = {∅}.

Then there exists an extension Ĩ of the ideal I such that T ∩ Ĩ = {∅} and

the ideal Ĩ is compatible with T . This is an immediate corollary from Theo-
rems 3.1. and 3.5. in [6]. We use this result to prove the following proposition.

Proposition 4.7. Let (X,T ,I) be an i-topological space. Assume that there
is a topological space, i-homeomorphic to (X,T ,I). Then there exists a com-

patible extension Ĩ of the ideal I such that Ĩ ∩ T = {∅}.

Proof. Let (Y,TY ,IY ) and h: X → Y be the corresponding topological space

and i-homeomorphism, respectively. Take Ĩ = h−1ĨY , where ĨY is a com-
patible extension of the ideal IY defined in [6]. The statements that Ĩ is

an ideal, I ⊆ Ĩ and Ĩ ∩ T = {∅} are the consequences from the facts that ĨY
is an ideal satisfying ĨY ∩ TY = {∅} and h is an i-homeomorphism. The fact

that Ĩ is compatible with T is an immediate corollary from Lemma 4.6. �

Given an i-topological space (X,T ,I), we say that a family A ⊆ P(X)
covers a set A ⊆ X if A ≤

⋃

A. The following corollary is a natural consequence
from the previous Proposition.

Corollary 4.8. Let (X,T ,I) be an i-topological space with a compatible ideal.
Assume that there is a nonempty i-open set U such that it can be covered with
i-open sets for which the intersection of U and each element of the cover lies
in the ideal. Then there does not exist a topological space, i-homeomorphic
to (X,T ,I).

Consider (R,D,In) from Example 2.4. The family every element of which
is of the form (R \ Q) ∪ {x}, x ∈ Q, consists of open sets. Clearly, this family
covers Q and every intersection of Q and an element of the family belongs to
the ideal. Thus, by the previous corollary (R,D,In) is not i-homeomorphic to
any topological space.

Now, we provide one particular construction for i-topological spaces with
compatible ideals and prove a sufficient condition for the existence of a topo-
logical space that is i-homeomorphic to a given i-topological space. First, let
us recall two related notions [5]. For a topological space (X,T ) and an ideal I
on X, an operator ∗ : P(X) → P(X) is said to be a local function of I with re-
spect to T iff A∗ = { x ∈ X | U ∩ A /∈ I for each U ∈ T (x) } for any A ⊆ X.
The various properties of this operator are to be found in [5]. An extension T ∗

of the topology T is defined as follows: T ∗ = { U \ A | U ∈ T and A ∈ I }. We
preserve these definitions for i-topological spaces.

We will exploit an operator ˜: P(X) → P(X) satisfying à = ψ(A) ∩ A∗ for

each A ⊆ X and use the notation T̃ = { Ũ | U ∈ T }. Notice that our use of

tilde differs from the notation Ĩ for ideals in [6]. We hope that there will be no
chance for confusion.

Lemma 4.9. Let (X,T ,I) be an i-topological space satisfying I ∼ T . Then

the following statements hold for any i-open U,V,W and U ⊆ T̃ :



64 I. Zvina

(i) U ≈ Ũ;

(ii) if Ũ ≤ Ṽ then Ũ ⊆ Ṽ ;

(iii) if Ũ ≈ Ṽ then Ũ = Ṽ ;

(iv) if Ũ ∩ Ṽ ∈ I then Ũ ∩ Ṽ = ∅;

(v) if Ũ ≈ Ṽ ∩ W̃ then Ũ ⊆ Ṽ ∩ W̃;

(vi) if Ũ ≈
⋃

U then
⋃

U ⊆ Ũ.

Proof. Statement (i) is obvious, since U ∈ ψ(U), ψ(U) \ U ∈ I and U \ U∗ ∈ I
for each i-open U.

First, we show that Ũ ⊆ Ṽ follows from U ≤ V . Clearly, Wx ∩ U /∈ I
implies Wx ∩ V /∈ I for each x ∈ U

∗ and any its i-open neighbourhood Wx.
Hence, U∗ ⊆ V ∗. On the other hand, for any x ∈ ψ(U) there exists its i-open
neighbourhood Wx such that Wx ≤ U. Then Wx ≤ V for each such Wx and
hence ψ(U) ⊆ ψ(V ). Thus, we have Ũ ⊆ Ṽ .

It follows from (i), that Ũ ≤ Ṽ implies U ≤ V . Thus, by the previous
paragraph we have (ii) and hence (iii).

Let us prove (iv). Suppose that Ũ ∩ Ṽ ∈ I and x ∈ Ũ ∩ Ṽ . Then there
is an i-open neighbourhood Wx of x such that Wx ≤ U. It follows from (i),

that Wx ≤ Ũ. Hence, Wx ∩Ṽ ∈ I and Wx ∩V ∈ I. Then by the definition x /∈
V ∗ and hence x /∈ Ṽ . The last statement contradicts our assumption, thus Ũ ∩
Ṽ = ∅.

Now, let us prove (v). Suppose that Ũ ≈ Ṽ ∩ W̃ . Then it follows from (i),
that U ≈ V ∩ W and hence U ≤ V and U ≤ W . By the second paragraph of
this proof we conclude that Ũ ⊆ Ṽ and Ũ ⊆ W̃ . Therefore, Ũ ⊆ Ṽ ∩ W̃ .

Finally, let us prove (vi). Suppose Ũ ≈
⋃

U. Then Ṽ ≤ Ũ for each Ṽ ∈ U.

It follows from (ii), that Ṽ ⊆ Ũ for each Ṽ ∈ U. Thus,
⋃

U ⊆ Ũ. �

Proposition 4.10. Let (X,T ,I) be an i-topological space satisfying I ∼ T .

Then
(

X, T̃ ,I
)

is a unified i-topological space such that I ∼ T̃ and the spaces

(X,T ,I) and
(

X, T̃ ,I
)

are i-homeomorphic.

Proof. The facts that (T1), (T3) and (T4) hold for
(

X, T̃ ,I
)

are obvious.

Let us prove (T2). Suppose that U ⊆ T and Ũ = { Ũ | U ∈ U }. Then (
⋃

U )\
(

⋃

Ũ

)

belongs to the ideal I, since U \ Ũ ∈ I for each U ∈ U and I is com-

patible with T . On the other hand, for any x ∈ Ũ \ U there is an i-open

neighbourhood Ux such that Ux ≤ U. Take Ux = { Ux | x ∈ Ũ and U ∈ U }

and A =
(

⋃

Ũ

)

\ (
⋃

U ). It follows that Ux ∩ A ∈ I for each Ux ∈ Ux and,

since I is compatible with T , (
⋃

Ux ) ∩ A belongs to I. Thus, we have (T2).
The fact that (N2) holds for the identity mapping id: X → X is obvious and

(N1) follows from the previous paragraph of this proof. Then it follows from

Lemmas 4.9 and 4.6, respectively, that
(

X, T̃ ,I
)

is unified and the ideal I is

compatible with T̃ . �



On i-topological spaces 65

Notice that
˜̃
T = T̃ holds for each i-topological space (X,T ,I) with a com-

patible ideal. This is an immediate consequence from the definition of T̃ ,
the previous Proposition and Lemma 4.9.

Proposition 4.11. Let (X,T ,I) be an i-topological space satisfying I ∼ T .
Then (X,T ∗,I) is an i-topological space such that I ∼ T ∗ and the spaces (X,T ,I)
and (X,T ∗,I) are i-homeomorphic.

Proof. Suppose that U = {Us}s∈S ⊆ T
∗. Then there are an i-open set U and

a collection A = {As}s∈S of the elements of the ideal such that Us ∪ As ∈ T
for each s ∈ S and U ≈

⋃

s∈S(Us ∪ As). It follows from I ∼ T , that
⋃

A \
⋃

U

belongs to I. Hence,
⋃

U ≈
⋃

s∈S(Us ∪ As) ≈ U. Observe that U = U \ ∅ is
an element from T ∗. Thus, we have (T2). All other facts which it remains to
show to complete the proof are obvious. �

The next proposition is a natural corollary from Propositions 4.12 and 4.11.

In what follows, all results hold if we replace
(

T̃
)∗

with
(

T̃
)∪

. Notice that

by Lemma 4.9
(

T̃
)∪

⊆
(

T̃
)∗

.

Proposition 4.12. Let (X,T ,I) be an i-topological satisfying I ∼ T . Assume

that P =
(

T̃
)∗

. Then the following statements hold:

(i) (X,P,I) is an i-topological space with a compatible ideal;
(ii) (X,T ,I) and (X,P,I) are i-homeomorphic;
(iii) U ∩ V ∈ I implies U ∩ V = ∅ for each U,V ∈ P;
(iv)

⋃

U ∈ P for each U ⊆ P;
(v) for any U,V ∈ P there is A ∈ I such that (U ∩V )\A ∈ P.

The following proposition gives us a sufficient condition for the existence of
a topological space i-homeomorphic to a given i-topological space. Notice that
this condition implies compatibility of an ideal and it holds for any topological
space.

Proposition 4.13. Let (X,T ,I) be an i-topological space. Assume that I =
n(I) and for any x ∈ X and any U,V ∈ T (x) there is W ∈ T (x) such that W ≤
U ∩ V . Then there exists a topological space, i-homeomorphic to (X,T ,I).

Proof. It follows from Corollary4.5, that I ∼ T .
First, let us show that Ũ1 ∩ Ũ2 ∈ T̃ for each Ũ1, Ũ2 ∈ T̃ . Take x ∈ Ũ1 ∩ Ũ2.

Then there are two i-open neighbourhoods W1 and W2 of x satisfying W1 ≤ U1
and W2 ≤ U2. By the assumption there is the third i-open neighbourhood W
of x such that W ≤ W1 ∩ W1. Then W ≤ V , where V is an i-open set such
that Ṽ ≈ Ũ1 ∩ Ũ2. Thus, x ∈ ψ(V ).

On the other hand, Wx ∩ W 6∈ I and hence Wx ∩ V /∈ I for each i-open
neighbourhood Wx of x, since x ∈ U

∗
1 . Hence, x ∈ V

∗ and we have x ∈ Ṽ .

The point x is an arbitrary point from Ũ1 ∩ Ũ2, thus Ũ1 ∩ Ũ2 ⊆ Ṽ and by
Lemma 4.9 Ṽ = Ũ1 ∩ Ũ2.



66 I. Zvina

We observe that U ∩ V = W implies (U \ A) ∩ (V \ A) = W \ A for each

subsets U,V,W and A. Suppose that P =
(

T̃
)∗

. Then for any U,V ∈ P there

exists W ∈ P such that U ∩ V = W . Finally, it follows from Proposition4.12
that (X,P,I) is a topological space, i-homeomorphic to (X,T ,I). �

We leave as an open question whether there exists an i-topological space
with a compatible ideal that is not i-homeomorphic to any topological space.

References

[1] F. G. Arenas, J. Dontchev, M. L. Puertas, Idealization of some weak separation axioms,
Acta Math. Hung. 89 (1-2) (2000), 47–53.

[2] R. Engelking, General Topology, Warszawa, 1977.
[3] T. R. Hamlett and D. Janković, Ideals in Topological Spaces and the Set Operator ψ,

Bollettino U.M.I. 7 (1990), 863–874.
[4] T. R. Hamlett and D. Janković, Ideals in General Topology, General Topology and

Applications, (Middletown, CT, 1988), 115 – 125; SE: Lecture Notes in Pure & Appl.
Math., 123 1990, Dekker, New York.

[5] D. Janković and T. R. Hamlett, New topologies from old via ideals, Amer. Math. Monthly
97 (1990), 295–310.

[6] D. Janković and T. R. Hamlett, Compatible Extensions of Ideals, Bollettino U.M.I. 7
(1992), 453–465.

[7] D. Janković, T. R. Hamlett and Ch. Konstadilaki, Local-to-global topological properties,
Mathematica Japonica 52 (1) (2000), 79–81.

[8] A. S. Mashhour, A. A. Allam, F. S. Mahmoud, F. H. Khedr, On supratopological spaces,
Indian J. Pure Appl. Math. 14 (1983), 502–510.

[9] R. L. Newcomb, Topologies which are compact modulo an ideal, Ph.D. Dissertation,
Univ. of Cal. and Santa Barbara, 13 (1) 1972, 193–197.

[10] D. V. Rančin, Compactness modulo an ideal, Soviet Math. Dokl. 13 (1) (1972), 193–197.
[11] S. Solovjovs, Topological spaces with a countable compactness defect (in Latvian), Bach-

elor Thesis, Univ. of Latvia, Riga, 1999.
[12] R. Vaidyanathaswamy, The localization theory in set-topology, Proc. Indian Acad. Sci.

Math Sci. 20 (1945), 51–61.

Received May 2004

Accepted April 2005

Irina Zvina (IrinaZvina@inbox.lv)
Institute of Mathematics of Latvian Academy of Sciences and University of
Latvia, Akademijas laukums 1, Riga, LV-1050, Latvia.