You are on page 1of 10

Accepted Manuscript

Homeomorphisms of Hashimoto Topologies

Małgorzata Filipczak, Grażyna Horbaczewska

PII: S0166-8641(18)30248-7
Reference: TOPOL 6516

To appear in: Topology and its Applications

Received date: 3 April 2018

Revised date: 6 August 2018
Accepted date: 7 August 2018

Please cite this article in press as: M. Filipczak, G. Horbaczewska, Homeomorphisms of Hashimoto Topologies, Topol. Appl.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are
providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting
proof before it is published in its final form. Please note that during the production process errors may be discovered which could
affect the content, and all legal disclaimers that apply to the journal pertain.


Abstract. We investigate homeomorphisms of different types of Hashimoto

topologies based on the Euclidean topology on the real line and classic

1. Introduction
Let us start with a notion of Hashimoto topologies introduced indepen-
dently by Martin in [12] and by Hashimoto in [4].
Let (X, T ) be a T1 topological space and let I be an ideal of subsets of
X, containing all singletons and such that I ∩ T = {∅}. We say that such
an ideal is admissible. Then the family
{U \ P : U ∈ T , P ∈ I}
is a base of a topology.
Under additional assumptions that (X, T ) is a second-countable topolog-
ical space and I is a σ−ideal, the considered family is a topology, denoted
by TI .
This kind of topologies was considered by Lukeš, Malý, Zajíček [11] as
’ideal topologies’, by Jankovic and Hamlet [8] and by other authors (Lindner
[10], Hejduk [5], Terepeta [14], Bingham and Ostaszewski [3] ) as ’Hashimoto
Note that such topologies have some common properties.
Theorem 1. (compare [12],[4]) Let (X, T ) be a second-countable topological
space and let I be an admissible σ−ideal. Then
(1) (X, TI ) is T1 .
(2) The families of connected sets in (X, T ) and in (X, TI ) coincide.
(3) (X, TI ) does not satisfy the first axiom of countability at any point.
(4) (X, TI ) is not regular.
2010 Mathematics Subject Classification. 54A10, 54A05,54C05.
Key words and phrases. homeomorphism, σ-ideal, Hashimoto topology.

(5) Every compact set in (X, TI ) is finite.

(6) The family of TI -nowhere dense sets contains I and the family of
T -nowhere dense sets.
Let us mention a straightforward consequence of the property (3) from
the last theorem.
Corollary 2. If (X, T ) is second-countable and I is an admissible σ−ideal,
then (X, T ) is not homeomorphic to (X, TI ).
The most common Hashimoto topology is the one on the real line consist-
ing of sets U \ P , where U is open in the Euclidean topology and P belongs
to the family of all Lebesgue null sets (denoted here by N ).
In this paper we study possibilities of existence of homeomorphisms be-
tween Hashimoto topologies TI , where T is the Euclidean topology on R
(or on the interval [0, 1]) and I is a certain admissible σ−ideal of subsets
of R (or of [0, 1]). We are interested in classic σ−ideals. Some of results
obtained here are rather suprising.
As we stated in the last corollary, none of Hashimoto topologies is home-
omorphic to the Euclidean one, but we can observe even more.
Corollary 3. For any admissible σ−ideal I every continuous function f :
(R, T ) → (R, TI ) is constant.
Proof. If f is continuous, then the image f (I) of any closed interval I is
TI -connected and TI -compact. Then, by Theorem 1 (2) and (5), it is a

2. σ-ideals connected with probability measures

Assume that F is a distribution function, i.e. nondecreasing, continuous
from the right and such that 0 = limx→−∞ F (x) ≤ limx→+∞ F (x) = 1. It is
well known (see, for example, [2]) that there exists a unique regular measure
μF on the Borel subsets of [0,1] for which μF ((a, b]) = F (b) − F (a) .
Denote IF := {A ⊂ R : μF (A) = 0}.
If F were not continuous, IF wouldn’t contain all singletons, and conse-
quently it wouldn’t be admissible. We would have an analogous case (not
admissible σ-ideal) if F were constant on an interval. Hence from now on
we assume that F is continuous and strictly increasing.
Theorem 4. The topological spaces ([0, 1], TIF ) and ([0, 1], TN ) are homeo-

Proof. The function F is a homeomorphism ([0, 1], T ) → ([0, 1], T ). Since

F is a bijection to show that it is also a homeomorphism ([0, 1], TIF ) →
([0, 1], TN ) it is enough to prove that image of any set from IF is a Lebesgue
null set. Let A ∈ IF and fix ε > 0. Byregularity of μF there exists an open
set U such that μF (U ) <  and U = n∈N (an , bn ), where intervals (an , bn )
are pairwise disjoint. Then F (U ) is open and, denoting by λ the Lebesgue
measure, we have
⎛ ⎛ ⎞⎞ ⎛ ⎞
λ(F (U )) = λ ⎝F ⎝ (an , bn )⎠⎠ = λ ⎝ F ((an , bn ))⎠ =
n∈N n∈N

⎛ ⎞
= λ⎝ (F (an ), F (bn ))⎠ = (F (bn )−F (an )) = μF ((an , bn )) = μF (U ) < ε,
n∈N n∈N n∈N

so A ∈ N . 

Note that if F is absolutely continuous, then IF = N and the last result

is trivial.
If F is singular, i.e. λ({x ∈ [0, 1] : F  (x) = 0}) = 1, then μF and λ have
disjoint supports (see [2]) and consequently, IF and N are incomparable.
In this case the last theorem gives a really interesting result. It is worth
mentioning that if we consider two Borel probability measures μ, ν with
strictly increasing continuous distributions and different σ-ideals of null sets,
both orthogonal to the Lebesgue measure, then corresponding Hashimoto
topologies are homeomorphic.
Classic examples of such measures are measures "generated by throwing
a coin". For a fixed p ∈ (0, 1) let X1 , X2 , ... be independent, identically
distributed random variables such that P {Xn = 0} = p and P {Xn = 1} =

1−p. Let X := Xn /2n . We denote by Fp (x) = P {X ≤ x} a distribution
function of X.
The function Fp is everywhere continuous, strictly increasing on [0, 1]

and (see
[2]). Therefore
for p, q ∈ (0, 1) the topological spaces
[0, 1], TIFp and [0, 1], TIFq are homeomorphic. It is worth underlying
that only one of the σ−ideals concerned here is shift invariant (if p = 1/2),
whereas the others are not. It makes this homeomorphism unexpected.

3. Shift invariant σ-ideals

In this section we will show that for classic shift invariant σ-ideals we get
Hashimoto topologies not homeomorphic to TN .
We begin with two lemmas.
Lemma 5. Assume that A is a family of subsets of R such that for any
nonempty open interval J there exists a set A ⊂ J belonging to A \ N .
Let h : ([0, 1] , T ) → ([0, 1] , T ) be an increasing homeomorphism.
If there exists a finite positive derivative of the function h at a point
x0 ∈ (0, 1), then there exists a set A ∈ A such that h (A) ∈ / N.
Proof. If h (x0 ) > 0 then there exist positive numbers a, b, δ such that for
any numbers x, y ∈ (x0 − δ, x0 + δ) , x > y we have
h (x) − h (y)
a< < b,
(1) a (x − y) < h (x) − h (y) < b (x − y) .
By the assumption there exists a set A ⊂ (x0 − δ, x0 + δ) belonging to A\N .
Since A ∈/ N then there exists ε0 > 0 such that for any sequence of intervals
{In } covering A we have
(λ(In )) ≥ ε0 .
Consider now h (A) and any sequence of  intervals {Jn } such that Jn ⊂
h ((x0 − δ, x0 + δ)) for every n ∈ N and ∞ n=0 Jn ⊃ h (A). For any n ∈ N
−1 −1
∞ set h (J n ) is connected, open and h (Jn ) ⊂ (x0 − δ, x0 + δ). Hence
n=0 (λ(h (Jn ))) ≥ ε0 . By (1) we have
aλ(h−1 (Jn )) < λ(Jn ).

(λ(Jn )) > aλ(h−1 (Jn )) = a (λ(h−1 (Jn ))) ≥ aε0 ,
n=0 n=0 n=0

i.e. h (A) ∈
/ N. 
We need also a classic result.
Lemma 6. ([13]) If f : [0, 1] → [0, 1] is a strictly increasing function, then
the set f ({x ∈ [0, 1] : f  (x) = 0}) has the Lebesgue measure zero.

Theorem 7. Let I be an admissible shift invariant and closed under multi-

plication by positive constants σ-ideal contained in N . Assume that for any
nonempty interval P there exists a set B ⊂ P belonging to N \ I. Then the
topological spaces (R, TN ) and (R, TI ) are not homeomorphic.
Proof. Assume, on the contrary, that there exists a homeomorphism h :
(R, TN ) → (R, TI ). We can assume that h is a homeomorphism ([0, 1] , TN ) →
([0, 1] , TI ).
By property (2) from Theorem 1, h is also a homeomorphism ([0, 1] , T ) →
([0, 1] , T ). Then h is invertible and h and h−1 are continuous. Hence h is
strictly monotonous (assume - increasing), so it has almost everywhere a
finite derivative. As a homeomorphism, h changes sets from N into sets from
I. Observe that the family A := {h−1 (I) : I ∈ N } fulfils the assumptions
of Lemma 5. Indeed, for any nonempty open interval J the set h(J), being
also a nonempty open interval, contains (by the assumption of the theorem)
a set B belonging to N \ I. Hence h−1 (B) is contained in J and belongs to
A \ N . By the definition of A there is no set A ∈ A such that h(A) ∈ / N.
Therefore, by Lemma 5, there is no point x for which h (x) is positive.
Then E := {x ∈ [0, 1] : h (x) = 0} has Lebesgue measure 1.
Moreover, by Lemma 6, λ (h (E)) = 0.
The set C := [0, 1] \ E, as a null set, is closed in ([0, 1], TN ). Moreover,
every subset of the set C is TN -closed. Since λ(h (C)) = 1 there exists a
nonmeasurable subset D of the set h(C). As TI -open and TI -closed sets
are measurable, the set D is not TI -closed. However h−1 (D) ⊂ C, so the
set D should be TI -closed - a contradiction.

We can give many examples of σ-ideals satisfying the assumptions of

the last theorem. For a given α ∈ (0, 1) it can be the σ-ideal of null sets
for α-dimensional Hausdorff measure or the σ-ideal of sets of σ-finite α-
dimensional Hausdorff measure or the σ-ideal of sets with Hausdorff dimen-
sion not greater than α ([6]). Other examples are a σ-ideal of microscopic
sets ([7]) and a σ-ideal of strong measure zero sets ([1]).
We would like to focus our attention on the smallest σ-ideal satisfying the
assumptions of Theorem 7 - the σ-ideal of countable sets, denoted here by
Iω . Of course (R, TN ) and (R, TIω ) are not homeomorphic, but it is possible
to prove a stronger result.

Theorem 8. Let I be an admissible σ-ideal containing Iω . Assume that

for any nonempty interval J there exists a set C ⊂ J belonging to I \ Iω .
Then every continuous function f : (R, TIω ) → (R, TI ) is constant.

Proof. Our proof starts with two remarks about the topology TIω .
Firstly, it is easily seen that every TIω -closed set is the union of a closed
set and a countable set. Hence, by Cantor-Bendixon Theorem ([9]), every
TIω -closed set may be written uniquely as a disjoint union of a perfect set
and a countable set.
Secondly, if every interval centered at a point x contains an uncountable
number of points from a certain set A, then x belongs to the TIω -closure
of A (denoted by clTIω (A)). Indeed, if x ∈ / clTIω (A), then there exists a set
B ∈ TIω containing x such that A ∩ B = ∅. Hence there exist an open set
C and a countable set D such that B = C \ D, so (A ∩ C) \ D = ∅. This
contradicts our assumption.
Now assume that there exists a continuous function f : (R, TIω ) →
(R, TI ), for which there exist points a, b ∈ R, a < b such that f (a)
= f (b).
By Theorem 1 (2), the sets [a, b] and f ([a, b]) are connected in TIω and
TI , respectively. Hence f ([a, b]), as a connected set, is an interval with a
nonempty interior.
Let C be a set contained in f ([a, b]) and belonging to I \ Iω . Then the
set C is closed in TI , so f −1 (C) as a closed set in TIω can be represented as
F ∪ P , where F is perfect and P ∈ Iω .
Put C1 := {y ∈ C : f −1 ({y}) ∩ P
= ∅}. Of course C1 ∈ Iω and C \ C1 ∈
I \ Iω .
Fix y ∈ C \ C1 . Then f −1 ({y}) ⊂ F and, since {y} and C \ {y} are
TI -closed, the sets f −1 ({y}) and f −1 (C \ {y}) are TIω -closed. Moreover
f −1 ({y}) ∩ f −1 (C \ {y}) = ∅.
Observe that for every x ∈ f −1 ({y}) there exists ε > 0 such that

(F \ f −1 ({y})) ∩ (x − ε, x + ε) ∈ Iω .

Indeed, in the opposite case by the remark from the beginning of the proof,
x ∈ clTIω (F \ f −1 ({y})), which is impossible, since F \ f −1 ({y}) ⊂ f −1 (C \
Therefore, since F is perfect, for every x ∈ f −1 ({y}) ⊂ F there exist
points ax , bx such that ax < x < bx and ax , bx are two-sided condensation
points of F and (F \f −1 ({y}))∩(ax , bx ) ∈ Iω . Obviously for every y ∈ C \C1
there exists an x ∈ f −1 ({y}), and so for every y ∈ C \ C1 there exist

points ay , by such that ay , by are two-sided condensation points of F and

(F \ f −1 ({y})) ∩ (ay , by ) ∈ Iω .
Moreover, for different y1 , y2 ∈ C\C1 we get disjoint intervals (ay1 , by1 ), (ay2 , by2 ),
because if (ay1 , by1 ) ∩ (ay2 , by2 ) = (c, d)
= ∅, then (c, d) ∩ F
= ∅, which is
impossible, since (c, d) ∩ F ∈ Iω and F is perfect.
Therefore we have an uncountable number of disjoint intervals, a contra-
Corollary 9. The Hashimoto topology (R, TIω ) is not homeomorphic to
(R, TI ) if I is
- the σ-ideal of Lebesgue null sets,
- the σ-ideals of α-dimensional Hausdorff measure,
- the σ-ideals of sets of σ-finite α-dimensional Hausdorff measure,
- the σ-ideals of sets with Hausdorff dimension not greater than α for
α ∈ (0, 1)
- the σ-ideal of microscopic sets,
- the σ-ideal of strong measure zero sets (under CH),
- the σ-ideal of meager sets.

4. σ-ideals orthogonal to the σ-ideal of meager sets

We consider now σ-ideals orthogonal to the σ-ideal of meager sets, de-
noted by K. It means σ-ideals I for which there exist sets I ∈ I and J ∈ K
such that R = I ∪ J.
Theorem 10. If J is an admissible σ-ideal orthogonal to K, then the topo-
logical spaces (R, TJ ) and (R, TK ) are not homeomorphic.
Proof. Theorem 4 from [4] says that for an admissible ideal I if every
nowhere dense set belongs to I, then I is equal to the family of TI -nowhere
dense sets, so the σ-ideal K is equal to the family of TK -nowhere dense
sets and, consequently, to the family of TK -meager sets. Hence R is not
TK -meager.
However, by orthogonality of K and J the space R is a union of two sets:
A ∈ K and B ∈ J . By Theorem 1 (6) the family of TJ -meager sets contains
K ∪ J , so R is TJ -meager.
Therefore (R, TJ ) and (R, TK ) are not homeomorphic. 
Corollary 11. The Hashimoto topology (R, TK ) is not homeomorphic to
(R, TI ) if I is
- the σ-ideal of Lebesgue null sets,

- the σ-ideals of α-dimensional Hausdorff measure,

- the σ-ideals of sets of σ-finite α-dimensional Hausdorff measure,
- the σ-ideals of sets with Hausdorff dimension not greater than α for
α ∈ (0, 1)
- the σ-ideal of microscopic sets.

Proof. For orthogonality of these ideals to K compare [1], [7]. 

Observe that despite of the fact that the σ-ideal of countable sets is not
orthogonal to the σ-ideal of meager sets we concluded in Corollary 9 that
(R, TIω ) and (R, TK ) are not homeomorphic.

[1] T. Bartoszyński, H. Judah, Set theory: On the structure of the real line, A. K.
Peters, Ltd., Wellesley, MA, 1995.
[2] P. Billingsley, Probability and Measure, A Wiley-Interscience publication, 1995.
[3] N. H. Bingham, A. J. Ostaszewski, Beyond Lebesgue and Baire IV: density topologies
and a converse Steinhaus-Weil theorem, Topology Appl., available online 31 January
[4] H. Hashimoto, On the *-topology and its applications, Fund. Math. 91, 1976, 5-10.
[5] J. Hejduk, S. Lindner, On the Hashimoto topology with respect to an extension of
the Lebesgue measure, Tatra Mt. Math. Publ. 24 (2002), part II, 147–151.
[6] G. Horbaczewska On topologies connected with Hausdorff measures, Real Anal. Ex-
change 33 (1), 2007/2008, 151-158.
[7] G. Horbaczewska, A. Karasińska, E. Wagner-Bojakowska, Properties of the σ-ideal of
microscopic sets, Traditional and present-day topics in real analysis, Łódź University
Press, 2013.
[8] D. Jankovic, T.R. Hamlet, New topologies from old via ideals, Amer. Math. Monthly,
Vol. 97, No. 4, 1990, 295-310.
[9] A.S. Kechris, Classical Descriptive Set Theory, Berlin, New York, Springer-Verlag,
[10] S. Lindner, Topologies of Hashimoto type with respect to σ-ideal of countable sets,
Folia Math. 10 (2003), no. 1, 55–58.
[11] J. Lukeš, J. Malý, L. Zajíček, Fine topology methods in real analysis and potential
theory, Lecture Notes in Math. 1189, Springer, 1986.
[12] N.F.G. Martin, Generalized condensation points, Duke Math. Journal, 28(4), 1961,
[13] I.P. Natanson, Theory of Functions of a Real Variable, Dover Books on Mathematics,
[14] M.Terepeta, On Hashimoto type topologies, Tatra Mt. Math., Publ. 52, 2012, 19-28.

Faculty of Mathematics and Computer science,

University of Łódź,
Banacha 22, 90 238 Łódź, Poland
E-mail address: