Abstract.

We show that, given a semitopological group G, the semitopological groups Ti(G) and (Ti(G)), for each i{0,1,2}, and (Gsr) and (G)sr, are topologically isomorphic. Moreover, we prove that, given a paratopological group G, the paratopological groups Reg(G) and (Reg(G)) are topologically isomorphic.

keywords:
Hartman–Mycielski; semiregularization; Ti-reflection; semitopological group; paratopological group.
MSC:
54B05; 54B15; 54C05; 54H99.

1. Introduction

Given a topological (paratopological, semitopological) group G, we can construct other topological (paratopological, semitopological) groups associated with G, possibly with some improvements in their topological properties. In [2], S. Hartman and J. Mycielski presented a functorial construction that permits to embed a Hausdorff topological group into a path connected, locally path connected Hausdorff topological group. A brief outline of their construction is given below.

Let G be a topological group with identity e. Consider the set G of all functions f on J=[0,1) with values in G such that, for some sequence 0=a0<a1<<an=1, the function f is constant on [ak,ak+1) for each k=0,1,,n1.

Define a binary operation on G by (fg)(x)=f(x)g(x), for all f,gG and xJ. Each element fG has a unique inverse f1G defined by (f1)(x)=(f(x))1, for each xJ. Then (G,) is a group with identity e, where e(x)=e for each xJ.

Let V be an open neighborhood of e in G and ϵ a positive real number. Consider the subset

O(V,ϵ)={fG:μ({xJ:f(x)V})<ϵ}

of G, where μ is the usual Lebesgue measure on J. One can verify that the subsets O(V,ϵ) form a base of a Hausdorff topological group topology at the identity of G, thus making G into a pathwise connected and locally pathwise connected topological group. For details, the reader is referred to [2].

Let fG. Then there exist real numbers a0,a1,,an with 0=a0<a1<<an=1 such that f is constant on [ak,ak+1) for each k=0,1,,n1. Assuming that f(ak)f(ak+1) for each kn1, we put P(f)={a0,a1,,an1}, Δ(f)=min{ak+1ak:0k<n} and V(f)={f(a0),,f(an1)}. For the identity eG of the group G we put P(eG)={a0}.

It is easy to verify that in the Hartman–Mycielski construction, if G is a paratopological (semitopological) group, then G is also a paratopological (semitopological) group. For more information on the Hartman–Mycielski construction, see [1, Section 3.8] and [4]. In what follows, G refers to the Hartman–Mycielski group associated to G and 𝒩(e) to the family of all open neighborhoods of the identity in G.

In [8], M. Tkachenko defines what we will call the Ti-reflection, Ti(G), of an arbitrary semitopological group G for i{0,1,2,3,3.5}, as well as the regular reflection, Reg(G), of G. All these objects are new semitopological groups, and their names indicate that Ti(G) satisfies the Ti separation axiom for i{0,1,2,3,3.5}, whereas Reg(G) is regular, i.e., satisfies the T1 and T3 separation axioms.

In this paper we show that, in the category of semitopological groups, the functors and Ti for i{0,1,2} commute, that is, the semitopological groups Ti(G) and (Ti(G)) for i{0,1,2} are topologically isomorphic. Likewise, we prove that (Gsr) and (G)sr are topologically isomorphic. Finally, we show that in the category of paratopological groups, the functors Reg and commute, that is, the paratopological groups Reg(G) and (Reg(G)) are topologically isomorphic.

First, we present some simple but very important properties of the Hartman–Mycielski group G associated to a semitopological group G. We omit their proofs, since they are identical to those found in [1, Theorem 3.8.3], [1, Proposition 3.8.6], and [1, Proposition 3.8.7], respectively.

For each gG, denote by g the element of G defined by g(r)=g, for all rJ. Let iG:GG be the function defined by iG(g)=g for each gG.

Theorem 1.1.

Let G be a semitopological group. Then iG:GG is a topological monomorphism. If G is Hausdorff, then the image iG(G) is a closed subgroup of G.

Proposition 1.2.

Let φ:GH be a continuous homomorphism of semitopological groups. Then φ admits a natural extension to a continuous homomorphism φ:GH. If φ is open and onto, then so is φ.

The correspondences GG and φφ are functorial since the equality (ψφ)=ψφ holds for any continuous homomorphisms φ:GH and ψ:HK. This defines the covariant functor in the category of semitopological groups and continuous homomorphisms. It turns out that this functor preserves subgroups and quotient groups:

Proposition 1.3.

Let G be a semitopological group and H a subgroup of G.

  1. a)

    If φ is the identity embedding of H to G, then the natural homomorphic extension φ:HG is a topological monomorphism.

  2. b)

    If H is closed or invariant in G, then H is also closed or invariant in G.

  3. c)

    If H is a closed invariant subgroup of G, then the groups (G/H) and G/H are naturally topologically isomorphic.

Following [8], we say that a class 𝒞 of spaces is a PS-class if it contains arbitrary products of its elements, is hereditary with respect to taking subspaces, and contains a one-point space. It is clear that the class 𝒯i of Ti-spaces is a PS-class for each i{0,1,2,3,3.5}, and so are the classes of regular spaces (i.e., T1 and T3) and Tychonoff spaces (i.e., T1 and T3.5).

Given a semitopological group G and a number i{0,1,2,3,3.5}, the Ti-reflection of G, denoted by Ti(G), is a pair (H,φG,i), where H is a semitopological group satisfying the Ti separation axiom and φG,i is a continuous homomorphism from G onto H with the following property: For every continuous mapping f:GX, where X is a space satisfying the Ti separation axiom, there exists a continuous mapping h:HX such that f=hφG,i.

GφG,ifHhX

In [8], M. Tkachenko provides an internal characterization of the kernel of the canonical homomorphism φG,i for i{0,1}. I.  Sánchez, in [6] describes the kernel of the canonical homomorphism φG,2, for any semitopological group G.

The following example shows that the functors T1 and T3 do not commute in the category of paratopological groups.

Example 1.4 ([8, Example 3.10]).

Let be the additive group of integers. For every integer n0, let Un={0}{m:m>n}. It is easy to verify that the family {Un:n} forms a neighborhood base at zero for a paratopological group topology τ1 on , and that the paratopological group G=(,τ1) satisfies the T1 separation axiom. It is also clear that any two nonempty open subsets of G have a nonempty intersection. Since G is a T1-space, we have T1(G)=G and hence T3(T1(G))=T3(G). By [8, Lemma 3.9], the canonical homomorphism φG,3 of G onto T3(G) is a bijection. Therefore, T3(G) is an infinite group (endowed with the anti-discrete topology, see [7, Theorem 2.6]).

On the other hand, T1(T3(G)) is the trivial group, for [8, Theorem 3.4]. Consequently, the groups T1(T3(G)) and T3(T1(G)) are quite different.

2. Ti-reflection and the Hartman–Mycielski construction in semitopological groups for i{0,1,2}

In this section we deal with the Ti-reflection of G for i{0,1,2}, where G is a semitopological group. We will prove that Ti(G) and (Ti(G)) are topologically isomorphic.

The following Lemmas 2.1 and 2.3 are almost immediate from the definition of subsets O(U,ϵ) of G.

Lemma 2.1.

Let G be a semitopological group, U𝒩(e) and ϵ>0. Then [O(U,ϵ)]1=O(U1,ϵ).

Proof 2.2.

Let f[O(U,ϵ)]1. If g=f1, then gO(U,ϵ)μ({rJ:g(r)U})<ϵμ({rJ:f(r)U1})<ϵfO(U1,ϵ).

Lemma 2.3.

Let G be a semitopological group, and U and V be nonempty subsets of G. Then O(U,ϵ)O(V,δ)O(UV,ϵ+δ) for all ϵ,δ>0.

Proof 2.4.

If hO(U,ϵ)O(V,δ), then h=fg for some fO(U,ϵ) and gO(V,δ). To see that hO(UV,ϵ+δ) note that

{rJ:h(r)UV} ={rJ:f(r)g(r)UV}
{rJ:f(r)U}{rJ:g(r)V}.

This proves the lemma.

In Lemmas 2.5 and 2.7, we present some properties of the closures of basic open sets in G.

Lemma 2.5.

Let G be a semitopological group, U𝒩(e), and ϵ>0. Then the equality O(U¯,ϵ)=V𝒩(e)O(UV1,ϵ) holds.

Proof 2.6.

We have that U¯=V𝒩(e)UV1. Hence, O(U¯,ϵ)O(UV1,ϵ) for each V𝒩(e), whence it follows that O(U¯,ϵ)V𝒩(e)O(UV1,ϵ).

If fV𝒩(e)O(UV1,ϵ), then fO(UV1,ϵ) for each V𝒩(e). Let us verify that the Lebesgue measure of the set {rJ:f(r)U¯} is less than ϵ. Since fG, there exist real numbers a1,a2,,an with 0=a0<a1<<an<an+1=1, and elements x0,x1,,xnG, such that f(r)=xi for each r[ai,ai+1), where 0in.

If V(f)U¯, then fO(U¯,ϵ). If V(f)U¯ and f(ai)U¯ for some in, we can find an element Wi𝒩(e) such that f(ai)WiU=, so f(ai)UWi1. Let M={in:f(ai)U¯} and W=iMWi. Since f(ai)UW1 for each iM, we have

{rJ:f(r)U¯}={rJ:f(r)UW1}.

Therefore, μ({rJ:f(r)U¯})<ϵ, and we conclude that fO(U¯,ϵ). This completes the proof.

Lemma 2.7.

Let G be a semitopological group, U𝒩(e) and ϵ>0. Then O(U¯,ϵ)O(U,ϵ)¯.

Proof 2.8.

Let fO(U¯,ϵ), W𝒩(e) and δ>0. Suppose that P(f)={a0,,an}. For each in such that f(ai)U¯, we have f(ai)WU, that is, there exists wiW such that f(ai)wiU. Now, defining g(r)=wi for each r[ai,ai+1) if f(ai)U¯, and g(r)=e otherwise, we obtain an element gG such that fgfO(W,δ)O(U,ϵ). Therefore, fO(U,ϵ)¯, that is, O(U¯,ϵ)O(U,ϵ)¯.

In the following example we show that the inclusion in Lemma 2.7 can be proper.

Example 2.9.

Let G=p be the topological group of p-adic integers, for any integer p2. Suppose that the equality O(U,ϵ)¯=O(U¯,ϵ) holds in G for all U𝒩(e) and ϵ>0. Since G is zero-dimensional, our assumption implies that

O(V,ϵ)=O(V¯,ϵ)=O(V,ϵ)¯,

for each open subgroup V of G. This is a contradiction, since G is a connected semitopological group by [1, Theorem 3.8.3]. Therefore, the equality is not true.

It is known (see [7]) that the T2-reflection of an arbitrary topological group G is the quotient topological group G/N, where N is the closure of the identity element of G, that is, T2(G)=G/{e}¯. Furthermore, since in topological groups the separation axiom T0 is equivalent to complete regularity, it follows that for any topological group G, the Ti-reflections of G, for i=0,1,2,3,3.5, are equivalent.

Theorem 2.10.

Let G be a topological group. Then T2(G) and (T2(G)) are topologically isomorphic.

Proof 2.11.

Let N={e}¯. Since T2(G)=G/N, we have (T2(G))=(G/N) and, by Proposition 1.3, the groups (G/N) and G/N are topologically isomorphic. We also know that T2(G)=G/K, where K={e}¯. Hence, it suffices to show that N=K, where

K={fG:efO(U,ϵ) for all U𝒩(e) and ϵ>0}

and N={fG:V(f)N}. Let fN and assume that P(f)={a0,,an}. Since V(f)N we have f(ak)N for each kn, so ef(ak)U for all U𝒩(e) and kn. Therefore, for each k, there exists ukU such that f(ak)uk=e. We can then define gG with P(f)=P(g) and g(ak)=uk for k=0,,n. Since gO(U,ϵ) for all ϵ>0 and the equality fg=e holds, we see that efO(U,ϵ). Hence fK, that is, NK.

Conversely, if fN, then there exists xV(f) with xN. Since N is closed in G, there is V𝒩(e) such that exV. Taking the basic open set W=O(V,Δ(f)/2), we clearly have efW, that is, fK. Therefore, KN. This completes the proof.

The subsequent Lemmas 2.12, 2.13 and 2.14 follow from [6, Lemma 2.1], [6, Lemma 2.2] and [6, Theorem 2.3], respectively.

Lemma 2.12.

Let G be a semitopological group with identity e and 𝒩(e) be the family of all open neighborhoods of the identity in G. Then

K={O(U,ϵ)¯:U𝒩(e),ϵ>0}

is a closed invariant subgroup of G.

The subgroup K of G defined in Lemma 2.12 is used in the following two results.

Lemma 2.13.

Let G be a semitopological group and U,V𝒩(e). Then

O(U,ϵ)KO(V,δ)1=O(U,ϵ)O(V,δ)1,

for all positive real numbers ϵ,δ. Moreover, the equality O(U,ϵ)¯K=O(U,ϵ)¯ holds for all U𝒩(e) and ϵ>0.

Lemma 2.14.

Let G be a semitopological group. Then, T2(G)=G/K, so K is the kernel of the canonical homomorphism φG,2:GT2(G). Moreover, φG,21(φG,2(O(U,ϵ)¯))=O(U,ϵ)¯, for all U𝒩(e) and ϵ>0.

In the following theorem we show that the functors T2 and commute in the category of semitopological groups.

Theorem 2.15.

Let G be an arbitrary semitopological group. Then T2(G) and (T2(G)) are topologically isomorphic.

Proof 2.16.

If G is a Hausdorff semitopological group, then T2(G)=G and T2(G)=G, hence T2(G)=(T2(G)). Suppose that G is not Hausdorff. By [6, Theorem 2.3], we have T2(G)=G/K, where K={U¯:U𝒩(e)}. Then T2(G)=(G/K) and by Proposition 1.3, the groups (G/K) and G/K are topologically isomorphic. Also, by Lemma 2.12, we have the equality T2(G)=G/H, where H={O(U,ϵ)¯:U𝒩(e),ϵ>0}. It is therefore suffices to prove that the groups K and H coincide. Note that

K={fG:V(f)U𝒩(e)U¯}.

If fK, then V(f)U𝒩(e)U¯, so Lemma 2.7 implies that

f{O(U¯,ϵ):U𝒩(e),ϵ>0}H.

Thus, KH.

Conversely, if fK, then we can find xV(f) and V𝒩(e) such that xV¯. Then there exists W𝒩(e) such that xVW1, which implies fO(VW1,Δ(f)/2). By Lemma 2.3, we have

fO(V,Δ(f)/4)O(W1,Δ(f)/4)=O(V,Δ(f)/4)O(W,Δ(f)/4)1.

It follows that fO(V,Δ(f)/4)¯, thus implying that fH. Hence, HK. This completes the proof.

Proposition 2.17.

Let H be a subgroup of a semitopological group G. Then H is closed in G if and only if H={fG:V(f)H} is a closed subgroup of G.

Proof 2.18.

By item b) of Proposition 1.3, it suffices to show that if H is closed in G, then H is closed in G. Suppose that H is closed in G and take xH¯. Then, for any U𝒩(e), we have xUH, which clearly implies that xO(U,ϵ)H for each ϵ>0. Hence, xH¯, and so xH. We conclude that xH and that H is closed in G.

In Lemma 2.19, we describe the T1-reflection of G, for an arbitrary semitopological group G.

Lemma 2.19.

Let G be a semitopological group, K be the smallest closed subgroup of G, and F the smallest closed subgroup of G. Then K=F, that is, K is the smallest closed subgroup of G. Moreover, T1(G)=G/K.

Proof 2.20.

Let I=[a,b), where 0a<b1. Define a mapping φI:GG by

φI(x)(r)={x,ifrI;e,ifrI.

Clearly, φI is a topological monomorphism. Let p:GG/F be the quotient homomorphism. Then pφI:GG/F is a continuous homomorphism. Since G/F is a T1-space, the kernel ker(pφI) of pφI is a closed subgroup of G. Therefore, Kker(pφI), which implies that φI(K)F.

For every fK, there exist real numbers a1,a2,,an with 0=a0<a1<<an<an+1=1 such that f(r)=xmK for each rJm=[am,am+1), where 0mn. Taking I=Jm, we have φJm(xm)F for all 0mn. Therefore,

f=m=0nφJm(xm)F.

This shows that KF. Since, by Proposition 2.17, K is a closed subgroup of G and F is the smallest closed subgroup of G, it follows that K=F. Finally, by [8, Theorem 3.4], we have the equality T1(G)=G/K.

From [8, Theorem 3.4], Proposition 1.3 and Lemma 2.19 we obtain the following theorem:

Theorem 2.21.

Let G be a semitopological group. Then T1(G) and (T1(G)) are topologically isomorphic.

The next two facts follow from [8, Theorem 3.1].

Lemma 2.22.

Let G be a semitopological group with identity e. Let also

P={O(U,ϵ):U𝒩(e),ϵ>0}G.

Then K=PP1 is an invariant subgroup of G. Furthermore, the equality KO(U,ϵ)=O(U,ϵ) holds for all U𝒩(e) and ϵ>0.

We use the subgroup K of G defined in Lemma 2.22 in the subsequent result.

Lemma 2.23.

Let G be a semitopological group. Then T0(G)=G/K.

Theorem 2.24.

Let G be an arbitrary semitopological group. Then the groups T0(G) and (T0(G)) are topologically isomorphic.

Proof 2.25.

By [8, Theorem 3.1], the equality T0(G)=G/K holds, where K=𝒩(e)(𝒩(e))1. Let π:GG/K be the quotient homomorphism. Then, by Proposition 1.2, π admits an extension to a continuous open homomorphism π:G(G/K). Note that

ker(π)={fG:π(f)=e}={gG:V(g)K}=K,

where e is the identity of the group (G/K). Then, by [1, Theorem 1.5.13], G/K is topologically isomorphic to (G/K). Therefore, the groups T0(G) and (T0(G)) are topologically isomorphic.

3. Semiregularization and the Hartman–Mycielski construction in semitopological groups

The operation of semiregularization of a topological space was defined by M. Katetov in [3]. A. Ravsky [5] considered this operation in the classes of semitopological and paratopological groups and proved that the semiregularization of a semitopological (paratopological) group is again a semitopological (paratopological) group. Furthermore, the semiregularization of a Hausdorff paratopological group is a regular space.

In this section we deal with semiregularizations of semitopological groups. Let G be a semitopological group with identity e. The family

βro={IntGU¯:U𝒩(e)}

serves as a base at the identity e for a topology 𝒯sr on G, which is coarser than the original topology of G. The group G with the topology 𝒯sr is called the semiregularization of G and is usually denoted by Gsr. We say that G is semiregular if G=Gsr.

Our aim is to prove that the semitopological groups (G)sr and (Gsr) are topologically isomorphic. This requires two auxiliary results.

Lemma 3.1.

Let G be a semitopological group, 0<ϵ<1, and U𝒩(e). Then

IntO(U¯,ϵ)=IntO(U,ϵ)¯.
Proof 3.2.

By Lemma 2.7, it suffices to prove that IntO(U,ϵ)¯IntO(U¯,ϵ). Take an arbitrary element fGO(U¯,ϵ). Then μ({rJ:f(r)U¯})ϵ. Let P(f)={a0,,an} and M={0in:f(ai)U¯}.

Case I. μ({rJ:f(r)U¯})>ϵ. For each mM, there exists Wm𝒩(e) such that f(am)WmU=. Let W=mMWm. Then f(am)WU= for each mM. We claim that fO(W,δ/2)O(U,ϵ)=, where δ=μ({rJ:f(r)U¯})ϵ>0. Indeed, suppose that there exists hfO(W,δ/2)O(U,ϵ). Since hfO(W,δ/2), we have μ({rJ:h(r)f(r)W})<δ/2. For each mM,

{r[am,am+1):h(r)f(am)W}{rJ:h(r)U},

whence it follows that

mM{r[am,am+1):h(r)f(am)W}{rJ:h(r)U}.

By the choice of δ, we have

μ(mM{r[am,am+1):h(r)f(am)W})ϵ+δ/2>ϵ,

thus μ({rJ:h(r)U})>ϵ, which contradicts the assumption. Therefore, fO(W,δ/2)O(U,ϵ)=, that is, fO(U,ϵ)¯.

Case II. μ({rJ:f(r)U¯})=ϵ. Let V𝒩(e), mM and δ>0. Choose akP(f) and bJ such that f(ak)U¯, ak<bak+1 and bak<δ. Define an element gG by the rule

g(r)={f(ak)1f(am)ifr[ak,b),eifr[ak,b).

Clearly, gO(V,δ). Then fg{hG:μ({rJ:h(r)U¯})>ϵ}, so by Case I, we have fgO(U,ϵ)¯. Thus fIntO(U,ϵ)¯.

We have shown that IntO(U,ϵ)¯O(U¯,ϵ), which implies the required inclusion IntO(U,ϵ)¯IntO(U¯,ϵ).

Lemma 3.3.

Let G be a semitopological group, 0<ϵ<1 and U𝒩(e). Then

O(IntU¯,ϵ)=IntO(U,ϵ)¯.
Proof 3.4.

Since IntU¯U¯ for all U𝒩(e), we have O(IntU¯,ϵ)O(U¯,ϵ). Then, by Lemma 3.1, we deduce that O(IntU¯,ϵ)IntO(U,ϵ)¯. Conversely, let fGO(IntU¯,ϵ), and suppose that P(f)={a0,a1,,an}. Then μ({rJ:f(r)IntU¯})ϵ. If μ({rJ:f(r)U¯})ϵ, then fO(U¯,ϵ), hence fIntO(U,ϵ)¯. Otherwise, let

M={0kn:f(ak)U¯IntU¯}.

It follows that

μ({rJ:f(r)IntU¯})=kM(ak+1ak)+μ({rJ:f(r)U¯})ϵ.

For each kM and each W𝒩(e), we have f(ak)WU and f(ak)W(GU¯). Take ykW such that f(ak)ykU¯, where kM. Choosing gG with P(g)=P(f) and g(ak)=yk for each kM, and g(r)=e otherwise, we have that gO(W,δ) for all δ>0 and

μ({rJ:(fg)(r)U¯})=1.

Thus fgfO(W,δ)O(U¯,ϵ), that is, fIntO(U¯,ϵ). Hence, IntO(U,ϵ)¯O(IntU¯,ϵ). This completes the proof.

The subsequent theorem is the main result of this section, it follows from the combination of Lemmas 3.1 and 3.3.

Theorem 3.5.

Let G be a semitopological group. Then the local bases

(βro)={O(IntU¯,ϵ):U𝒩(e),ϵ>0}

and

(β)ro={IntO(V,ϵ)¯:V𝒩(e),ϵ>0}

at e generate the same topology on G, that is, (G)sr and (Gsr) are topologically isomorphic.

In [7], M. Tkachenko proved that in paratopological groups, the T3-reflection and semiregularization coincide. Therefore, combining Theorems 3.5, 2.21 and [7, Theorem 3.8], we obtain the following result.

Theorem 3.6.

Let G be a paratopological group. Then the groups Reg(G) and Reg(G) are topologically isomorphic.

4. Conclusions

From this work we can conclude the following:

  1. (1)

    In the category of paratopological groups, the functors Ti and commute for i{0,1,2,3,3.5}.

  2. (2)

    In the category of semitopological groups, the functors Ti and commute for i{0,1,2}. Moreover, the semiregularization and the functor also commute in this category.

5. Open Problems

Since the semiregularization of a semitopological group does not, in general, coincide with its T3-reflection, it is natural to consider the following problems.

Problem 5.1.

Describe in internal terms the kernel of the canonical homomorphism φG,3 of a semitopological group G onto T3(G).

Problem 5.2.

Let G be a semitopological group. Are the semitopological groups T3(G) and (T3(G)) topologically isomorphic?

We recall that a class of spaces is a 𝑃𝑆-class if it is closed under arbitrary products, taking subspaces, and contains a singleton.

Let G be a topological group and 𝒞 be a 𝑃𝑆-class. By [7, Theorem 2.3], the 𝒞-reflection 𝒞(G) of G exists and is a topological group.

Problem 5.3.

Is (𝒞(G)) an element of the 𝑃𝑆-class 𝒞?

Problem 5.4.

Are the topological groups 𝒞(G) and (𝒞(G)) topologically isomorphic?

Acknowledgements.
I would like to thank Dr. Mikhail Tkachenko, for his comments and suggestions in writing this article.
Funding.
This research has not received external funding.
Author contributions.
Conceptualization, methodology and investigation B.R.; writing–original draft preparation, B.R.; writing–review and editing, B.R. All authors have read and agreed to the published version of the manuscript.

References

  • [1] A. V. Arhangel’skii and M. G. Tkachenko, Topological Groups and Related Structures, Atlantis Studies in Mathematics, Vol. I, Atlantis Press/World Scientic, Paris-Amsterdam (2008).
  • [2] S. Hartman and J. Mycielski, On embedding of topological groups into connected topological groups, Colloq. Math. 5 (1958) 167–169.
  • [3] M. Katětov, A note on semiregular and nearly regular spaces, Časopis pro pěstování mat. fys. 72, no. 3 (1947), 97–99.
  • [4] M. López and I. Sánchez, Lindelöfness and Čech-completeness in the construction of Hartman–Mycielski, Houston J. Math. 47, no. 1 (2021), 263–270.
  • [5] O. V. Ravsky, Paratopological groups, II, Mat. Stud. 17, no. 1 (2002) 93–101.
  • [6] I. Sánchez, A note on the T2-reflection of semitopological groups, Topology Appl. 182 (2015), 117–121.
  • [7] M. Tkachenko, Axioms of separation in paratopological groups and reflection functors, Topology Appl. 179 (2015), 200–214.
  • [8] M. Tkachenko, Axioms of separation in semitopological groups and reflection functors, Topology Appl. 161 (2014), 364–376.