This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

On some topological realizations of groups and homomorphisms

Pedro J. Chocano, Manuel A. Morón and Francisco R. Ruiz del Portal
Abstract

Let f:GHf:G\rightarrow H be a homomorphism of groups. We construct a topological space XfX_{f} such that its group of homeomorphisms is isomorphic to GG, its group of homotopy classes of self-homotopy equivalences is isomorphic to HH and the natural map between the group of homeomorphisms of XfX_{f} and the group of homotopy classes of self-homotopy equivalences of XfX_{f} is ff. In addition, we consider realization problems involving homology, homotopy groups and groups of automorphisms.

1 Introduction

2020 Mathematics Subject Classification: 20B25, 06A06, 06A11 05E18, 55P10, 55P99.Keywords: Automorphism group, homotopy equivalence, Alexandroff spaces, posets, homology groups, homotopy groups.This research is partially supported by Grants PGC2018-098321-B-100 and BES-2016-076669 from Ministerio de Ciencia, Innovación y Universidades (Spain).

Alexandroff spaces are topological spaces with the property that the arbitrary intersection of open sets is open. That sort of topological spaces was first studied by P.S. Alexandroff in [1], where it is shown that they can also be seen as partially ordered sets. This viewpoint can be used to express topological notions in combinatorial terms. A particular case of Alexandroff spaces are finite topological spaces. There are two foundational papers on this subject that were published independently in 1966 [15, 17]. In [17], R.E. Stong made an analysis of the homeomorphism classification of finite topological spaces using matrices and also introduced combinatorial techniques to study their homotopy type. In [15], M.C. McCord studied the singular homology groups and homotopy groups of Alexandroff spaces proving that for every Alexandroff space XX there exists a simplicial complex 𝒦(X)\mathcal{K}(X) sharing the same homotopy groups and singular homology groups. In fact, it is shown that there is a continuous map fX:|𝒦(X)|Xf_{X}:|\mathcal{K}(X)|\rightarrow X inducing isomorphism on all homotopy groups. The converse result is also obtained, that is, given a simplicial complex LL, there exists an Alexandroff space 𝒳(L)\mathcal{X}(L) having the same singular homology groups and homotopy groups of LL.

Finite topological spaces or Alexandroff spaces are a good tool to solve realization problems: given a category CC and a group GG, is there an object XX in CC such that the group of automorphisms of XX is isomorphic to GG? In [6, 18, 5], it is proved that for every finite group GG there exists a finite topological space XGX_{G} such that its group of homeomorphisms is isomorphic to GG. Recently, in [4], J.A. Barmak constructed a finite topological space XGX_{G} with 4|G|4|G| points and lower cardinality than the finite topological spaces obtained in [6, 18, 5]. On the other hand, L. Babai in [2] obtained a finite topological space XGX_{G} with 3|G|3|G| points realizing GG as a group of homeomorphisms. This topological space has the disadvantage that it requires to find a “good” set of generators of GG satisfying a list of non-trivial properties. A generalization of these results for non-finite groups was made in [7], where the realization problem for the homotopical category HTopHTop and the pointed homotopical category HTopHTop_{*} was also solved. To be precise, it was proved that for every group GG there exists an Alexandroff space X¯G\overline{X}_{G}^{*} such that its group of automorphisms in the topological category TopTop, HTopHTop and HTopHTop_{*} is isomorphic to GG.

The restriction of HTopHTop to topological spaces with the homotopy type of a CW-complex is denoted by HPolHPol. In HPolHPol and the pointed version HPolHPol_{*}, the realizability problem has a long history, see for instance [12]. Recently, C. Costoya and A. Viruel solved the realization problem in HPolHPol_{*} for finite groups in [9].

We introduce a bit of notation. Let XX be a topological space. Let us denote by Aut(X)Aut(X) the group of homeomorphisms of XX. Let (X)\mathcal{E}(X) denote the group of homotopy classes of self-homotopy equivalences of XX. Let (X)\mathcal{E}_{*}(X) denote the group of pointed homotopy classes of pointed self-homotopy equivalences of XX.

The following result answers a natural question posed by professor Jesús Antonio Álvarez López during a talk in the VIII Encuentro de Jóvenes Topólogos (A Coruña, 2019).

Lemma 1.1.

Let GG and HH be two groups. There exists a topological space XHGX_{H}^{G} such that Aut(XHG)Aut(X_{H}^{G}) is isomorphic to GG and (XHG)\mathcal{E}(X_{H}^{G}) is isomorphic to HH.

An immediate consequence of Lemma 1.1 is the following: for a general topological space there is no relation between its group of homeomorphisms and its group of homotopy classes of self-homotopy equivalences. Moreover, if GG and HH are two groups, then we can produce infinitely many non-homeomorphic topological spaces having GG as their group of homeomorphisms and having HH as their group of homotopy classes of self-homotopy equivalences.

For every topological space XX there is a natural homomorphism of groups τ:Aut(X)(X)\tau:Aut(X)\rightarrow\mathcal{E}(X) sending each homeomorphism ff to its homotopy class [f][f]. Given two groups GG and HH, we consider the topological space XGHX_{G}^{H} obtained in Lemma 1.1. The kernel of τ:Aut(XGH)(XGH)\tau:Aut(X_{G}^{H})\rightarrow\mathcal{E}(X_{G}^{H}) corresponds precisely to Aut(XHG)Aut(X_{H}^{G}). The image of τ\tau is the homotopy class of the identity map. However, modifying the construction of the topological space XHGX_{H}^{G}, we can obtain a stronger version of Lemma 1.1.

Theorem 1.2.

Let f:GHf:G\rightarrow H be a homomorphism of groups. There exists a topological space XfX_{f} such that Aut(Xf)=GAut(X_{f})=G, (Xf)=H\mathcal{E}(X_{f})=H and τ=f\tau=f.

Theorem 1.2 clearly generalizes Lemma 1.1. We prefer to prove Lemma 1.1 first for the sake of exposition. Omitting Lemma 1.1, the proof of Theorem 1.2 becomes less intuitive. In addition, subsequent results are obtained using the topological space given in Lemma 1.1.

Given a topological space XX and a non-negative integer number nn, we can consider the nn-th homology group Hn(X)H_{n}(X) of XX or the nn-th homotopy group πn(X)\pi_{n}(X) of XX. It is natural to consider more realization problems involving these groups, the group of homeomorphisms and the group of homotopy classes of self-homotopy equivalences.

Theorem 1.3.

Let GG and HH be finite groups and let XX be a topological space with the homotopy type of a compact CW-complex. There exists an Alexandroff space X¯HG\overline{X}_{H}^{G} such that Aut(X¯HG)Aut(\overline{X}_{H}^{G}) is isomorphic to GG, (X¯HG)\mathcal{E}(\overline{X}_{H}^{G}) is isomorphic to HH and XX is weak homotopy equivalent to X¯HG\overline{X}_{H}^{G}, which implies that Hn(X¯HG)H_{n}(\overline{X}_{H}^{G}) is isomorphic to Hn(X)H_{n}(X) and πn(X¯HG)\pi_{n}(\overline{X}_{H}^{G}) is isomorphic to πn(X)\pi_{n}(X) for every nn\in\mathbb{N}.

As an immediate consequence of Theorem 1.3, we can deduce the following corollaries.

Corollary 1.4.

Let HH and GG be finite groups and let {Fi}iI\{F_{i}\}_{i\in I} be a set of finitely generated Abelian groups, where II\subset\mathbb{N} is a finite set. There exists a topological space XX such that Aut(X)Aut(X) is isomorphic to GG, (X)\mathcal{E}(X) is isomorphic to HH and Hi(X)H_{i}(X) is isomorphic to FiF_{i} for every iIi\in I.

Corollary 1.5.

Let HH and GG be finite groups, nn\in\mathbb{N} and let TT be a finitely presented (Abelian) group (if n>1n>1). There exists a topological space XX such that Aut(X)Aut(X) is isomorphic GG, (X)\mathcal{E}(X) is isomorphic to HH and πn(X)\pi_{n}(X) is isomorphic to TT.

Roughly speaking, these corollaries say that for a general topological space XX its group of automorphisms in TopTop or HTopHTop does not have any relation to its nn-th homology or homotopy group and vice versa. In contrast, for the category HPolHPol, the situation is completely different since (X)\mathcal{E}(X) contains normal subgroups that are nilpotent. For instance, given a topological space XX, we denote by #(X)\mathcal{E}_{\#}(X) ((X))(\mathcal{E}_{*}(X)) the set of self-homotopy equivalences that induce the identity map in homotopy (homology). It is trivial to check that #(X)\mathcal{E}_{\#}(X) ((X))(\mathcal{E}_{*}(X)) is a normal subgroup of (X)\mathcal{E}(X) and if XX is a finite CWCW-complex, then #(X)\mathcal{E}_{\#}(X) ((X))(\mathcal{E}_{*}(X)) is a nilpotent group. See [10] for more details. Using the construction obtained in Theorem 1.3, we can find topological spaces that do not satisfy the previous properties. From this we get that some of the techniques used to study the group of self-homotopy equivalences for CWCW-complexes cannot be adapted in a natural way to general spaces.

The organization of the paper is as follows. In Section 2 we introduce basic concepts and results from the literature. In Section 3 we provide an example of one of the main results in order to motivate the main ideas of the proof of Lemma 1.1. In Section 4 we prove Lemma 1.1 and give some remarks. In Section 5 we prove Theorem 1.2. In Section 6 we define a sequence of topological spaces whose homotopy and homology groups are all trivial, their group of automorphisms in TopTop and HTopHTop are also trivial, but they are not homeomorphic or homotopy equivalent to a point. Then, we prove Theorem 1.3 as well as Corollary 1.4 and Corollary 1.5. Finally, we give examples of topological spaces satisfying that the groups ()\mathcal{E}_{*}(\cdot) and #()\mathcal{E}_{\#}(\cdot) are not nilpotent in general.

2 Preliminaries

The following definitions and results can be found with more detail in [1, 3, 15, 17, 14] .

Definition 2.1.

Let XX and YY be topological spaces. A continuous function f:XYf:X\rightarrow Y is said to be a weak homotopy equivalence if it induces isomorphisms on all the homotopy groups.

Definition 2.2.

An Alexandroff space XX is a topological space satisfying that the arbitrary intersection of open sets is open.

If XX is an Alexandroff space, then for every xXx\in X there exists a minimal open neighbourhood UxU_{x} given by the intersection of every open set containing xx. FxF_{x} denotes the set given by the intersection of every closed set containing xx. Trivially, every finite topological space is an Alexandroff space. An Alexandroff space XX is locally finite if for every xXx\in X the set UxU_{x} is finite.

Let (X,)(X,\leq) be a partially ordered set or poset. If x,yXx,y\in X, then we write xyx\prec y (xyx\succ y) if and only if x<yx<y (x>yx>y) and there is no zXz\in X such that x<z<yx<z<y (x>z>yx>z>y). We will denote by max(X)max(X) the maximum of XX if it exists. We denote by Px=(Ex,Sx)P_{x}=(E_{x},S_{x}) the cardinal numbers Ex=|{yX|yx}|E_{x}=|\{y\in X|y\prec x\}| and Sx=|{yX|xy}|S_{x}=|\{y\in X|x\prec y\}|. A set SXS\subseteq X is called lower (upper) if for every xSx\in S and yxy\leq x (yxy\geq x) we have ySy\in S.

It is not difficult to verify the following two properties:

  • For a partially ordered set (X,)(X,\leq) the family of lower (upper) sets of \leq is a T0T_{0} topology on XX, that makes XX a T0T_{0} Alexandroff space.

  • For a T0T_{0} Alexandroff space, the relation xτyx\leq_{\tau}y if and only if UxUyU_{x}\subset U_{y} (UyUxU_{y}\subset U_{x}) is a partial order on XX.

In addition, for a set XX, the T0T_{0} Alexandroff space topologies on XX are in bijective correspondence with the partial orders on XX.

From now on, every Alexandroff space satisfies the T0T_{0} separation axiom. The following results can be found, for instance, in [3, 14].

Proposition 2.3.

If f:XYf:X\rightarrow Y is a map between Alexandroff spaces, then ff is continuous if and only if ff preserves the order.

From this and previous properties, the following result can be deduced.

Theorem 2.4.

The category of T0T_{0} Alexandroff spaces is isomorphic to the category of partially ordered sets.

Hence, partially ordered sets and T0T_{0} Alexandroff spaces can be treated as the same object.

Proposition 2.5.

Let f,g:XYf,g:X\rightarrow Y be continuous maps between Alexandroff spaces. If f(x)g(x)f(x)\leq g(x) (f(x)g(x)f(x)\geq g(x)) for every xXx\in X, then ff and gg are homotopic.

Remark 2.6.

If XX is a T0T_{0} Alexandroff space with a minimum (maximum) xx*, then XX is contractible to xx*. This follows from the previous proposition and the fact that the constant map c:XXc:X\rightarrow X given by c(x)=xc(x)=x* satisfies that c(x)xc(x)\leq x (c(x)xc(x)\geq x) for every xXx\in X.

Definition 2.7.

Given a finite poset (X,)(X,\leq), the height ht(X)ht(X) of XX is one less than the maximum number of elements in a chain of XX. The height of a point xx in a locally finite Alexandroff space is given by ht(Ux)ht(U_{x}). For a general Alexandroff space XX, the height of a point xXx\in X is defined as \infty if UxU_{x} contains a chain without a minimum and ht(Ux)ht(U_{x}) otherwise.

Example 2.8.

Let us consider the real numbers with the usual order. For every xx\in\mathbb{R} the height of xx is \infty because the chain <xn<<x2<x1<x...<x-n<...<x-2<x-1<x does not have a minimum. Moreover, Px=(0,0)P_{x}=(0,0) since there is no yy\in\mathbb{R} satisfying that xyx\prec y or xyx\succ y. Let us consider X={}X=\mathbb{N}\cup\{*\}, where we consider the partial order defined as follows: n<n<* for every nn\in\mathbb{N}. It is clear that UU_{*} is not a finite set but the height of * is 11. Furthermore, P=(||,0)P_{*}=(|\mathbb{N}|,0).

Proposition 2.9.

Let XX and YY be Alexandroff spaces. If f:XYf:X\rightarrow Y is a homeomorphism and xyx\prec y, then f(x)f(y)f(x)\prec f(y) (f(x)f(y)f(x)\succ f(y)). Furthermore, for every xXx\in X the height of xx is equal to the height of f(x)f(x) and Px=Pf(x)P_{x}=P_{f(x)}.

The following results provide a combinatorial way to study the homotopy and weak homotopy type of finite topological spaces.

Definition 2.10.

Let XX be an Alexandroff space. A point xx in XX is a down beat point (resp. up beat point) if Ux{x}U_{x}\setminus\{x\} has a maximum (resp. Fx{x}F_{x}\setminus\{x\} has a minimum). A finite T0T_{0} topological space XX is a minimal finite space if it has no beat points. A core of a finite topological space is a strong deformation retract which is a minimal finite space.

Proposition 2.11 ([17]).

Let XX be an Alexandroff space and let xXx\in X be a beat point. Then X{x}X\setminus\{x\} is a strong deformation retract of XX.

If XX is a finite T0T_{0} topological space, then XX has a core. We only need to remove beat points one by one to obtain a minimal finite space.

Theorem 2.12 ([17]).

If XX is a minimal finite space, then f:XXf:X\rightarrow X is a homeomorphism if and only if ff is a homotopy equivalence.

Corollary 2.13.

If XX is a minimal finite space, then Aut(X)Aut(X) is isomorphic to (X)\mathcal{E}(X).

Remark 2.14.

The result of Corollary 2.13 can be stated in a stronger way. Let XX be a minimal finite space. It is easy to check that Aut(X)=(X)Aut(X)=\mathcal{E}(X). For every [f][f] in (X)\mathcal{E}(X) there is only one element in the class [f][f]. We can identify every homeomorphism with its homotopy class.

Remark 2.15.

Corollary 2.13 can be generalized to Alexandroff spaces. To do this, consider the notion of being locally a core introduced in [13]. This notion generalizes the notion of minimal finite space, that is, every minimal finite space is locally a core. Let XX be locally a core. Then a continuous map f:XXf:X\rightarrow X is a homeomorphism if and only if ff is a homotopy equivalence. Moreover, we get that Aut(X)(X)Aut(X)\simeq\mathcal{E}(X).

Definition 2.16.

Let XX be a finite T0T_{0} topological space. A point xx in XX is a down weak beat point (resp. up weak beat point) if Ux{x}U_{x}\setminus\{x\} is contractible (resp. Fx{x}F_{x}\setminus\{x\} is contractible).

Proposition 2.17 ([3]).

Let XX be a finite T0T_{0} topological space and let xXx\in X be a weak beat point. Then the inclusion i:X{x}Xi:X\setminus\{x\}\rightarrow X is a weak homotopy equivalence.

Definition 2.18 ([3]).

Let XX be a finite T0T_{0} topological space and let YXY\subset X. It is said that XX collapses to YY by an elementary collapse if YY is obtained from XX by removing a weak beat point. Given two finite T0T_{0} topological spaces XX and YY, XX collapses to YY if there is a sequence X=X1,X2,,Xn=YX=X_{1},X_{2},...,X_{n}=Y of finite T0T_{0} topological spaces such that for each 1i<n1\leq i<n, XiX_{i} collapse to Xi+1X_{i+1} by an elementary collapse.

Lemma 2.19.

Let XX and YY be Alexandroff spaces and let f:XYf:X\rightarrow Y be a homeomorphism. Then xXx\in X is a down (up) weak beat point if and only if f(x)f(x) is a down (up) weak beat point.

Proof.

There is no loss of generality in assuming that xx is a down weak beat point. If xx is an up weak beat point, then the argument is similar. It is easy to see that f(Ux)=Uf(x)f(U_{x})=U_{f(x)}. Therefore, f(Ux{x})=Uf(x){f(x)}f(U_{x}\setminus\{x\})=U_{f(x)}\setminus\{f(x)\} and we get the desired result.

We recall the notion of a Hasse diagram for a locally finite Alexandroff space XX. The Hasse diagram H(X)H(X) of XX is a directed graph. The vertices of H(X)H(X) are the points of XX. There is an edge between two vertices xx and yy if and only if xyx\prec y and the orientation of the edge is from the lower element to the upper element. We omit the orientation of the subsequent Hasse diagrams and we assume an upward orientation.

Remark 2.20.

It is easy to identify beat points of a finite topological space XX by looking at its Hasse diagram. A vertex xx is a down beat point (resp. up beat point) if there is only one edge that enters (exits) it, i.e., Px=(a,b)P_{x}=(a,b), where a=1a=1 (b=1b=1).

The homotopy and singular homology groups of Alexandroff spaces were studied in [15].

Definition 2.21.

Let XX be an Alexandroff space. Its McCord complex or order complex 𝒦(X)\mathcal{K}(X) is the simplicial complex whose simplices are the non-empty chains of XX. Let LL be a simplicial complex. The face poset of LL, denoted by 𝒳(L)\mathcal{X}(L), is defined to be the poset of simplices of LL ordered by inclusion.

Remark 2.22.

A finite T0T_{0} topological space XX is said to be collapsible if it collapses to a point. If XX is a collapsible finite T0T_{0} topological space, then 𝒦(X)\mathcal{K}(X) is also collapsible.

The geometric realization of a simplicial complex KK is denoted by |K||K|.

Theorem 2.23.

[15] Given an Alexandroff space XX, there exists a weak homotopy equivalence f:|𝒦(X)|Xf:|\mathcal{K}(X)|\rightarrow X.

Theorem 2.24.

[15] Given a simplicial complex LL, there exists a weak homotopy equivalence f:|L|𝒳(L)f:|L|\rightarrow\mathcal{X}(L).

Finally, we recall some remarks and a definition. For a more complete treatment we refer to the reader to [3].

Definition 2.25.

The non-Hausdorff join XYX\circledast Y of two Alexandroff spaces XX and YY is the disjoint union XYX\sqcup Y keeping the given ordering within XX and YY and setting yxy\leq x for every xXx\in X and yYy\in Y.

Remark 2.26.

Given two Alexandroff spaces XX and YY, 𝒦(XY)=𝒦(X)𝒦(Y)\mathcal{K}(X\circledast Y)=\mathcal{K}(X)\ast\mathcal{K}(Y), where \ast denotes the usual join of simplicial complexes. If XX and YY are finite topological spaces and one of them is collapsible, then 𝒦(XY)\mathcal{K}(X\circledast Y) is collapsible.

Proposition 2.27.

Let XX and YY be two Alexandroff spaces. Then Aut(XY)=Aut(X)×Aut(Y)Aut(X\circledast Y)=Aut(X)\times Aut(Y).

3 Examples and motivation of the proof of Lemma 1.1

We present an example to illustrate the idea of the construction given in the proof of Lemma 1.1.

Example 3.1.

Let us consider the Klein four-group 22\mathbb{Z}_{2}\oplus\mathbb{Z}_{2}, where we denote g1=(0,0),g2=(1,0),g3=(0,1)g_{1}=(0,0),g_{2}=(1,0),g_{3}=(0,1) and g4=(1,1)g_{4}=(1,1), and the cyclic group of two elements 2\mathbb{Z}_{2}, where we denote h1=0h_{1}=0 and h2=1h_{2}=1. We also denote G=22G=\mathbb{Z}_{2}\oplus\mathbb{Z}_{2} and H=2H=\mathbb{Z}_{2} for simplicity. Moreover, let SG={g2,g3}S_{G}=\{g_{2},g_{3}\}, SH={h2}S_{H}=\{h_{2}\} be generating sets of G,HG,H respectively. We declare g2<g3g_{2}<g_{3}. Our goal is to find a finite T0T_{0} topological space XHGX_{H}^{G} such that Aut(XHG)GAut(X_{H}^{G})\simeq G and (XHG)H\mathcal{E}(X_{H}^{G})\simeq H.

By [5, 7], there exists a finite T0T_{0} topological space XGX^{G} satisfying that Aut(XG)Aut(X^{G}) is isomorphic to GG. In Figure 1 we have represented in blue the Hasse diagram of XGX^{G}. It is clear that adding to XGX^{G} a minimum *, i.e., XG=XG{}X_{*}^{G}=X^{G}\cup\{*\} with <x*<x for every xXGx\in X_{G}, we get that (XG)\mathcal{E}(X^{G}_{*}) is trivial since XGX_{*}^{G} is contractible. On the other hand, if fAut(XG)f\in Aut(X^{G}_{*}), then we have that f()=f(*)=* because * is a minimum. Thus we deduce that Aut(XG)Aut(X^{G}_{*}) is isomorphic to Aut(XG)Aut(X^{G}). Our next goal is to find a topological space XHX_{H}^{*} satisfying that Aut(XH)Aut(X_{H}^{*}) is trivial and (XH)\mathcal{E}(X_{H}^{*}) is isomorphic to HH. Again, by [7], there exists a finite T0T_{0} topological space XHX_{H} satisfying that (XH)Aut(XH)H\mathcal{E}(X_{H})\simeq Aut(X_{H})\simeq H. In Figure 1, the Hasse diagram of XHX_{H} corresponds to the red and black parts of the diagram on the right. We modify XHX_{H} in order to reduce the number of self-homeomorphisms without changing the number of self-homotopy equivalences. For this purpose we add some points to XHX_{H}. We consider XH=XH{wh1,wh2,a}X_{H}^{*}=X_{H}\cup\{w_{h_{1}},w_{h_{2}},a\}, where we have the following relations: A(h1,0)wh1A_{(h_{1},0)}\prec w_{h_{1}} and A(h2,0)wh2aA_{(h_{2},0)}\prec w_{h_{2}}\prec a. The Hasse diagram of XHX_{H}^{*} can be seen on the right on Figure 1, where the new points are pictured in orange. It is easy to check that (XH)(XH)H\mathcal{E}(X_{H}^{*})\simeq\mathcal{E}(X_{H})\simeq H. The new points are beat points so we can remove them without changing the homotopy type of XHX_{H}^{*}. Hence, we have that XHX_{H}^{*} and XHX_{H} have the same homotopy type. We prove that Aut(XH)Aut(X_{H}^{*}) is trivial. If fAut(XH)f\in Aut(X_{H}^{*}), then f(M)=Mf(M)=M and f(N)=Nf(N)=N, where MM and NN denote the set of maximal and minimal elements of XHX_{H}^{*} respectively. From this, using Proposition 2.9 and the fact that ff preserves heights, we deduce that ff is the identity.

Refer to caption
Figure 1: Hasse diagrams of XGX_{*}^{G} and XHX_{H}^{*}.

Combining XHX_{H}^{*} with XGX_{*}^{G} we obtain XHGX_{H}^{G}. We identify the point * of XHX_{H}^{*} and XGX_{*}^{G}. We also extend the partial order of the two previous posets using transitivity, that is, if xXHx\in X_{H}^{*} and yXGy\in X_{*}^{G}, then we have that x<yx<y if and only if x<<yx<*<y. It is not difficult to check that XHGX_{H}^{G} satisfies the properties required at the beginning. XHGX_{H}^{G} and XHX_{H} have the same homotopy type because we can collapse XGX_{*}^{G} to *. We thus get that (XHG)(XH)H\mathcal{E}(X_{H}^{G})\simeq\mathcal{E}(X_{H})\simeq H. Since * is the only point with height 11 and P=(2,4)P_{*}=(2,4), it follows that * is a fixed point for every fAut(XHG)f\in Aut(X_{H}^{G}). By the continuity of ff, it is easily seen that f(XG)=XGf(X^{G}_{*})=X^{G}_{*} and f(XH)=XHf(X_{H}^{*})=X_{H}^{*}. From this, we conclude that Aut(XHG)Aut(X_{H}^{G}) is isomorphic to GG.

Refer to caption
Figure 2: Hasse diagram of XHGX_{H}^{G}.

We can also consider what we call the dual case, that is, XGHX_{G}^{H}, where we have that Aut(XGH)HAut(X_{G}^{H})\simeq H and (XGH)G\mathcal{E}(X_{G}^{H})\simeq G. We can now proceed analogously to the previous arguments, i.e., we find XGX_{G}^{*} and XHX_{*}^{H}. In Figure 3 we have the Hasse diagrams of XHX_{*}^{H} and XGX_{G}^{*}.

Refer to caption
Figure 3: Hasse diagrams of XHX_{*}^{H} and XGX_{G}^{*}.

Again, combining XHX_{*}^{H} with XGX_{G}^{*} we get XGHX_{G}^{H}, that is, the finite topological space given by the Hasse diagram of Figure 4. It is trivial to verify that XGHX_{G}^{H} is not homeomorphic to XHGX_{H}^{G} because of their different cardinality. Furthermore, XHGX_{H}^{G} and XGHX_{G}^{H} are not homotopy equivalent since (XHG)\mathcal{E}(X_{H}^{G}) is not isomorphic to (XGH)\mathcal{E}(X_{G}^{H}). Another way to prove the last assertion is the following. After removing one by one the beat points of XHGX_{H}^{G} we get XHX_{H}; after removing one by one the beat points of XGHX_{G}^{H} we get XGX_{G}. However, XGX_{G} is not homeomorphic to XHX_{H} because of their different cardinality. By Theorem 2.12 we conclude that XHGX_{H}^{G} and XGHX_{G}^{H} are not homotopy equivalent. Moreover, studying their McCord complexes it can be shown that XHGX_{H}^{G} and XGHX_{G}^{H} are not weak homotopy equivalent.

Refer to caption
Figure 4: Hasse diagram of XGHX_{G}^{H}.

4 Proof of Lemma 1.1 and remarks

Proof of Lemma 1.1.

We follow the same strategy as in Example 3.1. Firstly, we find a topological space XHX_{H}^{*} such that (XH)H\mathcal{E}(X_{H}^{*})\simeq H and Aut(XH)Aut(X_{H}^{*}) is trivial. Secondly, we find a topological space XGX_{*}^{G} such that Aut(XG)GAut(X_{*}^{G})\simeq G and (XG)\mathcal{E}(X_{*}^{G}) is trivial. Finally, we combine properly both topological spaces to obtain a topological space XHGX_{H}^{G} satisfying that Aut(XHG)GAut(X_{H}^{G})\simeq G and (XHG)H\mathcal{E}(X_{H}^{G})\simeq H.

We first assume that GG and HH are non-trivial groups. The trivial case will be considered later.

Construction of XGX_{*}^{G} and properties. We consider a set of non-trivial generators SGS^{\prime}_{G} for GG. Without loss of generality we can consider a well-order on SGS^{\prime}_{G} satisfying that if max(SG)max(S^{\prime}_{G}) exists and |SG|>1|S^{\prime}_{G}|>1, then there exists αSG\alpha\in S^{\prime}_{G} satisfying αmax(SG)\alpha\prec max(S^{\prime}_{G}). If max(SG)max(S^{\prime}_{G}) exists and there is no αSG\alpha\in S^{\prime}_{G} satisfying αmax(SG)\alpha\prec max(S^{\prime}_{G}), then the well-order defined on SGS^{\prime}_{G} can be modified as follows: max(SG)<αmax(S^{\prime}_{G})<\alpha for every αSG{max(SG)}\alpha\in S^{\prime}_{G}\setminus\{max(S^{\prime}_{G})\} and the rest of the relations defined on SG{max(SG)}S^{\prime}_{G}\setminus\{max(S^{\prime}_{G})\} unaltered. It is obvious that the new partial order defined on SGS^{\prime}_{G} is indeed a well-order and max(SG)max(S^{\prime}_{G}) does not exist. We consider SG=SG{0,1}S_{G}=S^{\prime}_{G}\cup\{0,-1\}, where we assume that 1,0SG-1,0\notin S^{\prime}_{G}, and extend the well-order defined on SGS^{\prime}_{G} to SGS_{G} as follows: 1<0<α-1<0<\alpha for every αSG\alpha\in S^{\prime}_{G}. We consider

XG=(G×SG){},X^{G}_{*}=(G\times S_{G})\cup\{*\},

where we have the following relations:

  • (g,α)<(g,δ)(g,\alpha)<(g,\delta) if 1α<δ1\leq\alpha<\delta, where gGg\in G and α,δSG\alpha,\delta\in S_{G}.

  • (gα,1)(g,α)(g\alpha,-1)\prec(g,\alpha), where gGg\in G and αSG{1,0}\alpha\in S_{G}\setminus\{-1,0\}.

  • (g,1)*\prec(g,-1), where gGg\in G.

The rest of the relations can be deduced from the above relations using transitivity. It is easy to check that XGX_{*}^{G} is a partially ordered set.

We prove that Aut(XG)GAut(X_{*}^{G})\simeq G and (XG)\mathcal{E}(X_{*}^{G}) is the trivial group. We have that (XG)\mathcal{E}(X_{*}^{G}) is the trivial group because XGX_{*}^{G} is contractible to *, which is a minimum. Since * is a minimum, it follows that every self-homeomorphism must fix this point. From this we deduce that Aut(XG)Aut(XG{})Aut(X_{*}^{G})\simeq Aut(X_{*}^{G}\setminus\{*\}). In addition, XG{}X_{*}^{G}\setminus\{*\} is the same topological space considered in [7, Section 3] and denoted by XGX_{G}. Hence, we know that Aut(XG)Aut(XG)GAut(X_{*}^{G})\simeq Aut(X_{G})\simeq G, where φ:GAut(XG)\varphi:G\rightarrow Aut(X_{G}) is given by φ(s)(g,α)=(sg,α)\varphi(s)(g,\alpha)=(sg,\alpha) and is an isomorphism of groups.

Construction of XHX_{H}^{*} and properties. We consider a set of non-trivial generators SHS^{\prime}_{H} for HH. There is no loss of generality in assuming that there exists a well-order on SHS^{\prime}_{H} satisfying that if max(SH)max(S^{\prime}_{H}) exists and |SH|>1|S^{\prime}_{H}|>1, then there exists αSH\alpha\in S_{H} satisfying αmax(SH)\alpha\prec max(S^{\prime}_{H}). We repeat the same construction made before, that is, we consider SH=SH{0,1}S_{H}=S^{\prime}_{H}\cup\{0,-1\}, where we assume that 1,0SH-1,0\notin S^{\prime}_{H}, and extend the well-order defined on SHS^{\prime}_{H} to SHS_{H} as follows: 1<0<β-1<0<\beta for every βSH\beta\in S^{\prime}_{H}.

For every hHh\in H we take a well-ordered non-empty set WhW_{h} such that WhW_{h} is isomorphic to WtW_{t} if and only if h=th=t. For every hHh\in H let whWhw_{h}\in W_{h} denote the first element or minimum of WhW_{h}. We consider

XH=(H×SH)((h,β)G×SH0β<max(SH)(S(h,β)T(h,β))(hHWh)){},X_{H}^{*}=(H\times S_{H})\cup(\bigcup_{\begin{subarray}{c}{(h,\beta)\in G\times S_{H}}\\ 0\leq\beta<max(S_{H})\end{subarray}}(S_{(h,\beta)}\cup T_{(h,\beta)})\cup(\bigcup_{h\in H}W_{h}))\cup\{*\},

where

S(h,β)={A(h,β),B(h,β),C(h,β),D(h,β)},T(h,β)={E(h,β),F(h,β),G(h,β),H(h,β),I(h,β),J(h,β)},S_{(h,\beta)}=\{A_{(h,\beta)},B_{(h,\beta)},C_{(h,\beta)},D_{(h,\beta)}\},T_{(h,\beta)}=\{E_{(h,\beta)},F_{(h,\beta)},G_{(h,\beta)},H_{(h,\beta)},I_{(h,\beta)},J_{(h,\beta)}\},

and we have the following relations:

  1. 1.

    (h,β)<(h,γ)(h,\beta)<(h,\gamma) if 1α<γ-1\leq\alpha<\gamma, where hHh\in H and β,γSH\beta,\gamma\in S_{H}.

  2. 2.

    (hβ,1)(h,β)(h\beta,-1)\prec(h,\beta), where hHh\in H and βSH{1,0}\beta\in S_{H}\setminus\{-1,0\}.

  3. 3.

    A(h,β)C(h,β),D(h,β)A_{(h,\beta)}\succ C_{(h,\beta)},D_{(h,\beta)}; B(h,β)(h,β),C(h,β)B_{(h,\beta)}\succ(h,\beta),C_{(h,\beta)} and (h,β)D(h,β)(h,\beta)\succ D_{(h,\beta)}, where hHh\in H and βSH{1}\beta\in S_{H}\setminus\{-1\}.

  4. 4.

    E(h,β)(h,β),I(h,β)E_{(h,\beta)}\succ{(h,\beta)},I_{(h,\beta)}; F(h,β)H(h,β),J(h,β)F_{(h,\beta)}\succ H_{(h,\beta)},J_{(h,\beta)}; G(h,β)I(h,β),J(h,β)G_{(h,\beta)}\succ I_{(h,\beta)},J_{(h,\beta)} and (h,β)H(h,β)(h,\beta)\succ H_{(h,\beta)}, where hHh\in H and βSH{1}\beta\in S_{H}\setminus\{-1\}.

  5. 5.

    (h,1)*\succ(h,-1), where hHh\in H.

  6. 6.

    We extend the partial order defined on WhW_{h} to XHX_{H} declaring that A(h,0)whA_{(h,0)}\prec w_{h}, where hHh\in H.

The remaining relations can be deduced from the above using transitivity. It is routine to verify that XHX_{H}^{*} with the previous relations is a partially ordered set.

Refer to caption
Figure 5: Hasse diagram of S(h,0)T(h,0)WhS_{(h,0)}\cup T_{(h,0)}\cup W_{h}, where WhW_{h} is a finite well-ordered set.

We proceed to show that Aut(XH)Aut(X_{H}^{*}) is the trivial group and (XH)H\mathcal{E}(X_{H}^{*})\simeq H. It is clear that XHX_{H}^{*} and XH{Wh|hH}X_{H}^{*}\setminus\{W_{h}|h\in H\} have the same homotopy type. We define r:XHXH{Wh|hH}r:X_{H}^{*}\rightarrow X_{H}^{*}\setminus\{W_{h}|h\in H\} given by

r(x)={A(h,0)xWhxxXH{Wh|hH}.r(x)=\begin{cases}A_{(h,0)}&x\in W_{h}\\ x&x\in X_{H}^{*}\setminus\{W_{h}|h\in H\}.\end{cases}

It is trivial to show that rr is continuous and satisfies that r(x)id(x)r(x)\leq id(x) for every xXHx\in X_{H}^{*}, where id:XHXHid:X_{H}^{*}\rightarrow X_{H}^{*} denotes the identity map. This implies that XH{Wh|hH}X_{H}^{*}\setminus\{W_{h}|h\in H\} is a strong deformation retract of XHX_{H}^{*}. On the other hand, XH{Wh|hH}X_{H}^{*}\setminus\{W_{h}|h\in H\} is the same topological space considered in [7, Section 3] and denoted by X¯H\overline{X}^{*}_{H}. Therefore we know that (XH)(X¯H)Aut(X¯H)H\mathcal{E}(X_{H}^{*})\simeq\mathcal{E}(\overline{X}_{H}^{*})\simeq Aut(\overline{X}_{H}^{*})\simeq H, where ϕ:HAut(X¯H)\phi:H\rightarrow Aut(\overline{X}_{H}^{*}) given by ϕ(t)(h,β)=(th,β)\phi(t)(h,\beta)=(th,\beta) and ϕ(t)(S(h,β)T(h,β))=S(th,β)T(th,β)\phi(t)(S_{(h,\beta)}\cup T_{(h,\beta)})=S_{(th,\beta)}\cup T_{(th,\beta)} is an isomorphism of groups.

The task is now to prove that Aut(XH)Aut(X_{H}^{*}) is the trivial group. Let us take fAut(XH)f\in Aut(X_{H}^{*}). We consider A(h,0)A_{(h,0)} for some hHh\in H. Since FA(h,0){A(h,0)}F_{A_{(h,0)}}\setminus\{A_{(h,0)}\} has a minimum whw_{h}, it follows that A(h,0)A_{(h,0)} is an up beat point. By Lemma 2.19, we know that f(A(hi,0))f(A_{(h_{i},0)}) is also an up beat point. Therefore, f(A(hi,0))f(A_{(h_{i},0)}) is of the form A(t,0)A_{(t,0)} for some tHt\in H. By Proposition 2.9 we get that f(wh)=wtf(w_{h})=w_{t}. It follows from the continuity of ff that f(Wh)Wtf(W_{h})\subseteq W_{t}. Since ff is a homeomorphism, we have that f|Whf_{|W_{h}} is also a homeomorphism. Therefore we get that h=th=t; otherwise we would get a contradiction since WhW_{h} is homeomorphic to WtW_{t} if and only if h=th=t. Using Proposition 2.9 it is easy to verify that ff fixes S(h,0)S_{(h,0)} for every hHh\in H. On the other hand, [7, Remark 4.2] says that if a homeomorphism g:XH{Wh|hH}XH{Wh|hH}g:X_{H}^{*}\setminus\{W_{h}|h\in H\}\rightarrow X_{H}^{*}\setminus\{W_{h}|h\in H\} coincides at one point with the identity map, then gg is the identity map. Thus, we can deduce that ff is the identity map and Aut(XH)Aut(X_{H}^{*}) is the trivial group.

Construction of XHGX_{H}^{G}. We consider XHG=XHXGX_{H}^{G}=X_{H}^{*}\cup X_{*}^{G}, where we are identifying the point * of both topological spaces, i.e., the partial order of XHGX_{H}^{G} preserves the relations defined on XHX_{H}^{*} and XGX^{G}_{*}:

  • If xXHx\in X_{H}^{*} and yXGy\in X_{*}^{G}, then xx is smaller than yy if and only if xx\leq* and y*\leq y.

  • If x,yXHx,y\in X_{H}^{*}, then xx is smaller (greater) than yy if and only if xx is smaller (greater) than yy with the partial order defined on XHX_{H}^{*}.

  • If x,yXGx,y\in X_{*}^{G}, then xx is smaller (greater) than yy if and only if xx is smaller (greater) than yy with the partial order defined on XGX_{*}^{G}.

It is evident that (XHG)\mathcal{E}(X_{H}^{G}) is isomorphic to HH because XGX_{*}^{G} is contractible to * and XHX_{H}^{*} is homotopy equivalent to X¯H\overline{X}_{H}^{*}. It suffices to show that Aut(XHG)Aut(X_{H}^{G}) is isomorphic to GG. We verify that every fAut(XHG)f\in Aut(X_{H}^{G}) satisfies that f(x)XHf(x)\in X_{H}^{*} for every xXHx\in X_{H}^{*} and f(x)XGf(x)\in X_{*}^{G} for every xXGx\in X_{*}^{G}. Firstly, we show that * is a fixed point for every homeomorphism ff. We have ht()=1ht(*)=1. Since for every xXG{}x\in X_{*}^{G}\setminus\{*\} the height of xx is at least 22 or different from 11, it follows that f()XG{}XHGf(*)\notin X_{*}^{G}\setminus\{*\}\subset X_{H}^{G}. The only elements of XHX_{H}^{*} that have height one are of the form * or (h,0)(h,0) or A(s,α)A_{(s,\alpha)}, F(s,α)F_{(s,\alpha)}, E(s,α)E_{(s,\alpha)} for some h,sHh,s\in H and αSH{1}\alpha\in S_{H}\setminus\{-1\}. We can discard the maximal elements, otherwise, f1f^{-1} would send a maximal element to a non-maximal element. If f()=A(h,0)f(*)=A_{(h,0)}, then we get a contradiction since A(h,0)A_{(h,0)} is an up beat point. If f()=(h,0)f(*)=(h,0) for some hHh\in H, then we get that f1(E(h,0))f1(h,0)=f^{-1}(E_{(h,0)})\succ f^{-1}(h,0)=* by Proposition 2.9. By Lemma 2.19 we have that f1(E(h,0))(g,1)f^{-1}(E_{(h,0)})\neq(g,-1) for every gGg\in G because E(h,0)E_{(h,0)} is not a down beat point. Hence, the only possibility is f()=f(*)=*. Finally, by the continuity of ff, we get that f(x)XGXHGf(x)\in X_{*}^{G}\subset X_{H}^{G} for every xXGXHGx\in X_{*}^{G}\subset X_{H}^{G}. This implies that Aut(XHG)Aut(X_{H}^{G}) is isomorphic to GG.

We prove the remaining case. If GG is the trivial group, then it suffices to consider XHX_{H}^{*} to conclude. If HH is the trivial group, then XGX_{*}^{G} satisfies the desired properties. ∎

Remark 4.1.

If f,kAut(XHG)f,k\in Aut(X_{H}^{G}) are such that there exists xXG{}x\in X_{*}^{G}\setminus\{*\} satisfying f(x)=k(x)f(x)=k(x), then f=kf=k. This is a consequence of the isomorphism of groups φ\varphi given in the proof of Lemma 1.1. Similarly, if [f],[k](XHG)[f],[k]\in\mathcal{E}(X_{H}^{G}) are such that there exists xXHG({XG}{Wh|hH})x\in X_{H}^{G}\setminus(\{X_{*}^{G}\}\cup\{W_{h}|h\in H\}) satisfying that f(x)=k(x)f(x)=k(x), where f[f]f\in[f] and k[k]k\in[k], then f=kf=k and [f]=[k][f]=[k]. This is a consequence of the construction of XHGX_{H}^{G} and ϕ\phi given in the proof of Lemma 1.1.

Proposition 4.2.

Let GG and HH be groups. Then the Alexandroff space XHGX_{H}^{G} obtained in the proof of Lemma 1.1 has the weak homotopy type of the wedge sum of 3|H||SH|3|H||S_{H}| circles when HH is a finite group and the wedge sum of |||\mathbb{N}| circles when HH is a non-finite countable set.

Proof.

We have that XHGX_{H}^{G} has the same homotopy type of XHX_{H}^{*}. Repeating the same arguments used in [7, Proposition 6.1] the desired result follows. ∎

Remark 4.3.

Let GG and HH be finite groups and let XHGX_{H}^{G} be the finite topological space obtained in the proof of Lemma 1.1. We can remove {T(h,β)|hH,βSH{1,0}}\{T_{(h,\beta)}|h\in H,\beta\in S_{H}\setminus\{-1,0\}\} from XHGX_{H}^{G}. The resulting poset also satisfies that its group of homeomorphisms is isomorphic to GG and its group of homotopy classes of self-homotopy equivalences is isomorphic to HH. This finite topological space has |G|(|SG|+2)+|H|(|SH|+2)+4|SH||H|+|H|(|H|+1)2+1|G|(|S_{G}|+2)+|H|(|S_{H}|+2)+4|S_{H}||H|+\frac{|H|(|H|+1)}{2}+1 points. The first term corresponds to XG{}X_{*}^{G}\setminus\{*\}, the second term corresponds to XH{}X_{H}^{*}\setminus\{*\}, the third term corresponds to the sets S(h,β)S_{(h,\beta)}, the fourth term corresponds to the points of the sets WhW_{h} and the last term corresponds to the point *.

We can change the sets T(h,β)T_{(h,\beta)} from the proof of Lemma 1.1 by T(h,β)nT_{(h,\beta)}^{n} as in [7, Section 5], where T(h,β)n:={x1,x2,xn+3,y1,y2,,yn+3}T_{(h,\beta)}^{n}:=\{x_{1},x_{2},...x_{n+3},y_{1},y_{2},...,y_{n+3}\} with the following relations:

(h,β)<x1>y2<x3>y4<<xn+2>yn+3<xn+3>yn+2<xn+1<<x2>y1<(h,β),\displaystyle(h,\beta)<x_{1}>y_{2}<x_{3}>y_{4}<...<x_{n+2}>y_{n+3}<x_{n+3}>y_{n+2}<x_{n+1}<...<x_{2}>y_{1}<(h,\beta), (1)
(h,β)<x1>y2<x3>y4<>yn+2<xn+3>yn+3<xn+2>yn+1<<x2>y1<(h,β).\displaystyle(h,\beta)<x_{1}>y_{2}<x_{3}>y_{4}<...>y_{n+2}<x_{n+3}>y_{n+3}<x_{n+2}>y_{n+1}<...<x_{2}>y_{1}<(h,\beta). (2)

We consider (1) for nn odd and (2) for nn even. We denote this poset by XHnGX_{Hn}^{G}.

Corollary 4.4.

Given two finite groups GG and HH, there are infinitely many (non-homotopy-equivalent) topological spaces {XHnG}n\{X_{Hn}^{G}\}_{n\in\mathbb{N}} such that Aut(XHnG)Aut(X_{Hn}^{G}) is isomorphic to GG and (XHnG)\mathcal{E}(X_{Hn}^{G}) is isomorphic to HH for every nn\in\mathbb{N}.

Proof.

The proof is analogous to the proof of Lemma 1.1. By Theorem 2.12, we have that the topological spaces are not homotopy equivalent due to their different cardinality after removing all the beat points one by one. ∎

5 Examples, remarks and proof of Theorem 1.2

The idea of this section is to modify the topological space obtained in Lemma 1.1 to prove Theorem 1.2. Given a homomorphism of groups f:GHf:G\rightarrow H, we slightly modify the topological space XHX_{H}^{*} defined in the proof of Lemma 1.1 to get XfX_{f}. Adding new relations to XHX_{H}^{*} we can control the homomorphism of groups τ:Aut(Xf)(Xf)\tau:Aut(X_{f})\rightarrow\mathcal{E}(X_{f}) given by τ(f)=[f]\tau(f)=[f].

Example 5.1.

Let us consider the cyclic group of two elements 2\mathbb{Z}_{2} and the group of integer numbers \mathbb{Z}. We consider the homomorphism of groups f:2f:\mathbb{Z}\rightarrow\mathbb{Z}_{2} given by f(n)=nf(n)=n modmod 22. We consider the topological space X2X_{\mathbb{Z}_{2}}^{\mathbb{Z}} obtained in the proof of Lemma 1.1. We remove W0W_{0} and W1W_{1} from it. The resulting poset XfX_{f} corresponds to the Hasse diagram shown in black in Figure 6.

Refer to caption
Figure 6: Hasse diagram of XfX_{f}.

Now we add the following relations to XfX_{f}: (n,0)A(1,0)(n,0)\prec A_{(1,0)} if f(n)=1f(n)=1 and (n,0)A(0,0)(n,0)\prec A_{(0,0)} if f(n)=0f(n)=0. In Figure 6 we have represented these relations in blue. It is easy to verify that XfX_{f} satisfies that Aut(Xf)=Aut(X_{f})=\mathbb{Z}, (Xf)=2\mathcal{E}(X_{f})=\mathbb{Z}_{2} and f=τf=\tau.

Example 5.2.

Let f:222f:\mathbb{Z}_{2}\rightarrow\mathbb{Z}_{2}\oplus\mathbb{Z}_{2} be the homomorphism of groups given by f(0)=(0,0)f(0)=(0,0) and f(1)=(1,0)f(1)=(1,0). In Figure 7 we have the Hasse diagram of XfX_{f}, where we use the same notation introduced in Example 3.1. We have that Aut(Xf)=2Aut(X_{f})=\mathbb{Z}_{2}, (Xf)=22\mathcal{E}(X_{f})=\mathbb{Z}_{2}\oplus\mathbb{Z}_{2} and τ=f\tau=f.

Refer to caption
Figure 7: Hasse diagram of XfX_{f}.
Proof of Theorem 1.2.

Suppose GG and HH are not trivial groups since otherwise the result follows from Lemma 1.1. Let XHGX_{H}^{G} be the topological space obtained in the proof of Lemma 1.1. We consider Xf=XHG{Wh|hH}X_{f}=X_{H}^{G}\setminus\{W_{h}|h\in H\}. We keep the same relations defined on XfX_{f} as a subspace of XHGX_{H}^{G} and add the following relations:

  • A(h,0)(g,0)A_{(h,0)}\succ(g,0) if f(g)=hf(g)=h, where hHh\in H and gGg\in G.

  • A(h,0)A_{(h,0)}\succ* if hf(G)h\notin f(G).

It is easy to check that XfX_{f} is a partially ordered set with the above relations. The task is now to show that (Xf)H\mathcal{E}(X_{f})\simeq H. We consider r:XfXfr:X_{f}\rightarrow X_{f} given by

r(x)={xXGxxXf{XG}.r(x)=\begin{cases}*&x\in X_{*}^{G}\\ x&x\in X_{f}\setminus\{X_{*}^{G}\}.\end{cases}

We have that rr preserves the order so it is a continuous map. It is simple to verify that r(x)id(x)r(x)\leq id(x) for every xXfx\in X_{f}, where idid denotes the identity map. From this it follows that XfX_{f} is homotopy equivalent to r(Xf)=XHXfr(X_{f})=X_{H}^{*}\subset X_{f}. On the other hand, repeating the same arguments used in [7], it can be proved that XHX_{H}^{*} is locally a core [13] or a minimal finite space in case HH is a finite group, which implies that (Xf)(XH)=Aut(XH)\mathcal{E}(X_{f})\simeq\mathcal{E}(X_{H}^{*})=Aut(X_{H}^{*}). Since P=(|H|,|H|)P_{*}=(|H|,|H|) and ht()=1ht(*)=1, it follows that every homeomorphism T:XHXHT:X_{H}^{*}\rightarrow X_{H}^{*} fixes *. Hence, the group of homeomorphisms of XHX_{H}^{*} as a subspace of XfX_{f} is isomorphic to the group of homeomorphisms of the topological space XHX_{H}^{*} obtained in the proof of Lemma 1.1. Thus, (Xf)H\mathcal{E}(X_{f})\simeq H.

We proceed to show that Aut(Xf)GAut(X_{f})\simeq G. We consider the following auxiliary sets: Colh={(h,β)|βSH}{S(h,β)T(h,β)|βSH{1,max(SH)}}Col_{h}=\{(h,\beta)|\beta\in S_{H}\}\cup\{S_{(h,\beta)}\cup T_{(h,\beta)}|\beta\in S_{H}\setminus\{-1,max(S_{H})\}\}, where hHh\in H, and Colg={(g,α)|αSG}Col^{g}=\{(g,\alpha)|\alpha\in S_{G}\}, where gGg\in G. If xXGXfx\in X_{*}^{G}\subset X_{f}, then every homeomorphism T:XfXfT:X_{f}\rightarrow X_{f} satisfies that T(x)XGT(x)\in X_{*}^{G}. We prove the last assertion. We know that Xf{XG}X_{f}\setminus\{X_{*}^{G}\} does not contain beat points. On the other hand, for every gGg\in G we have that (g,1)(g,-1) is a beat point of height 22. Using Proposition 2.9 and the notion of continuity we deduce that for every TAut(Xf)T\in Aut(X_{f}) and (g,α)(g,\alpha), where gGg\in G and αSG\alpha\in S_{G}, we have T(g,α)=(g,α)T(g,\alpha)=(g^{\prime},\alpha) for some gGg^{\prime}\in G.

We consider φ:GAut(Xf)\varphi:G\rightarrow Aut(X_{f}) given by φ(g)(g,α)=(gg,α)\varphi(g)(g^{\prime},\alpha)=(gg^{\prime},\alpha) if gGg^{\prime}\in G and αSG\alpha\in S_{G}, φ(g)(h,β)=(f(g)h,β)\varphi(g)(h,\beta)=(f(g)h,\beta) if hHh\in H and βSH\beta\in S_{H}, φ(g)(S(h,β)T(h,β))=S(f(g)h,β)T(f(g)h,β)\varphi(g)(S_{(h,\beta)}\cup T_{(h,\beta)})=S_{(f(g)h,\beta)}\cup T_{(f(g)h,\beta)} defined in the natural way if hHh\in H and βSH{1,max(SH)}\beta\in S_{H}\setminus\{-1,max(S_{H})\} and φ(g)()=\varphi(g)(*)=*. We prove that φ\varphi is well-defined. We verify the continuity of φ(g)\varphi(g). Suppose (g,0)A(h,0)(g^{\prime},0)\prec A_{(h,0)} for some gGg^{\prime}\in G and hHh\in H. By hypothesis, f(g)=hf(g^{\prime})=h. Therefore,

φ(g)(g,0)=(gg,0)A(f(gg),0)=A(f(g)f(g),0)=A(f(g)h,0)=φ(g)A(h,0).\varphi(g)(g^{\prime},0)=(gg^{\prime},0)\prec A_{(f(gg^{\prime}),0)}=A_{(f(g)f(g^{\prime}),0)}=A_{(f(g)h,0)}=\varphi(g)A_{(h,0)}.

It is easy to check that φ(g)\varphi(g) preserves the remaining relations. The inverse of φ(g)\varphi(g) is given by φ(g1)\varphi(g^{-1}). Hence, φ\varphi is well-defined. By construction, φ\varphi is a monomorphism of groups. Suppose TAut(Xf)T\in Aut(X_{f}). Proposition 2.9, Remark 4.1 and the fact that T|XGAut(XG)T_{|X_{*}^{G}}\in Aut({X_{*}^{G}}) imply that every TAut(Xf)T\in Aut(X_{f}) satisfies T(Colh)=ColhT(Col_{h})=Col_{h^{\prime}} and T(Colg)=ColgT(Col^{g})=Col_{g^{\prime}} for some gGg^{\prime}\in G and hHh^{\prime}\in H, where gGg\in G and hHh\in H. We consider (g,0)A(h,0)(g,0)\prec A_{(h,0)} for some gGg\in G and hHh\in H. We get T(Colg)=ColgT(Col^{g})=Col^{g^{\prime}} for some gGg^{\prime}\in G and T(Colh)=ColhT(Col_{h})=Col_{h^{\prime}} for some hHh^{\prime}\in H. By Remark 4.1, the proof of Lemma 1.1 and the fact that T|XGAut(XG)T_{|X_{G}^{*}}\in Aut(X_{*}^{G}), there exists tGt\in G such that T(Cols)=ColtsT(Col^{s})=Col^{ts}, where sGs\in G. Hence, g=tgg^{\prime}=tg. By Proposition 2.9, T(A(h,0))=A(h,0)(tg,0)=T(g,0)T(A_{(h,0)})=A_{(h^{\prime},0)}\succ(tg,0)=T(g,0), we have h=f(tg)=f(t)f(g)=f(t)hh^{\prime}=f(tg)=f(t)f(g)=f(t)h. Thus, T=φ(t)T=\varphi(t) because of Remark 4.1 and the fact that T|Xf{XG}Aut(XH{})T_{|X_{f}\setminus\{X_{*}^{G}\}}\in Aut(X_{H}^{*}\setminus\{*\}). By construction, for every gGg\in G the equality τ(g)=f(g)\tau(g)=f(g) holds. Since every TAut(Xf)T\in Aut(X_{f}) can be seen as T=φ(g)T=\varphi(g) for some gg, it follows that τ(T)=f(g)\tau(T)=f(g), where f(g)=φ(g)|XH(Xf)f(g)=\varphi(g)_{|X_{H}^{*}}\in\mathcal{E}(X_{f}). ∎

Remark 5.3.

Theorem 1.2 generalizes Lemma 1.1 and the results of realization obtained in [7]. Let GG and HH be two groups. Using Theorem 1.2, we obtain a family of topological spaces {Xf}f:GH\{X_{f}\}_{f:G\rightarrow H} satisfying that Aut(Xf)GAut(X_{f})\simeq G and (Xf)H\mathcal{E}(X_{f})\simeq H.

Proposition 5.4.

Let GG and HH be groups. If g,f:GHg,f:G\rightarrow H are homomorphisms of groups, then XfX_{f} is homotopy equivalent to XgX_{g}.

Proof.

The result is an immediate consequence of the construction. We have that XfX_{f} is homotopy equivalent to XHX_{H}^{*} for every homomorphism of groups f:GHf:G\rightarrow H. Therefore, the homotopy type of the topological space obtained in the proof of Theorem 1.2 does not depend on the homomorphism chosen to construct it. Thus we deduce the desired result. ∎

Proposition 5.5.

Let GG and HH be groups and let f,g:GHf,g:G\rightarrow H be homomorphisms of groups. Then f=gf=g if and only if XfX_{f} is homeomorphic to XgX_{g}.

Proof.

One of the implications is trivial. It suffices to show that if XfX_{f} is homeomorphic to XgX_{g}, then f=gf=g. Since XfX_{f} is homeomorphic to XgX_{g}, it follows that there exists a homeomorphism T:XfXgT^{\prime}:X_{f}\rightarrow X_{g}. From the construction of XfX_{f} and XgX_{g} in the proof of Theorem 1.2 it can be easily deduced that T|XGAut(XG)GT^{\prime}_{|X^{G}_{*}}\in Aut(X^{G}_{*})\simeq G and T|XHAut(XH)HT^{\prime}_{|X_{H}^{*}}\in Aut(X_{H}^{*})\simeq H. This is due to the fact that XGX^{G}_{*} contains beat points while XHX_{H}^{*} does not have beat points. Therefore, T|XGT^{\prime}_{|X^{G}_{*}} can be related to the action of an element TGT\in G and T|XHT^{\prime}_{|X_{H}^{*}} can be related to the action of an element T¯H\overline{T}\in H. We have (e,1)A(f(e),0)(e,1)\prec A_{(f(e),0)}, where ee denotes the identity element in GG, and we also have

T(e,1)=(T,1)A(T¯f(e),0)=T(A(f(e),0)),T^{\prime}(e,1)=(T,1)\prec A_{(\overline{T}f(e),0)}=T^{\prime}(A_{(f(e),0)}),

which implies that g(T)=T¯f(e)g(T)=\overline{T}f(e). Thus, g(T)=T¯g(T)=\overline{T} because ff is a homomorphism of groups. In addition, for every hGh\in G, we know that there exists a relation in XfX_{f} of the following form (h,1)A(f(h),0)(h,1)\prec A_{(f(h),0)}. We have

T(h,1)=(Th,1)A(T¯f(h),0)=T(A(f(h),0)).T^{\prime}(h,1)=(Th,1)\prec A_{(\overline{T}f(h),0)}=T^{\prime}(A_{(f(h),0)}).

By the construction of XgX_{g} we get g(Th)=g(T)g(h)=T¯f(h)g(Th)=g(T)g(h)=\overline{T}f(h). Earlier we prove that g(T)=T¯g(T)=\overline{T}, which implies that g(h)=f(h)g(h)=f(h) for every hGh\in G. ∎

6 Groups of homology, homotopy and automorphisms

We first prove that the groups studied previously do not determine neither the homotopy type nor the topological type of a topological space XX in general. To do this we provide an example. However, if the topological space XX satisfies some properties, namely, XX is compact and a locally Euclidean manifold with or without boundary, then its group of homeomorphisms determines its topological type of it, see [19] for more details.

Example 6.1.

Let us consider the Alexandroff space W2W_{2} given by the Hasse diagram of Figure 8. It is the union of L1={xi}i=1,,9L_{1}=\{x_{i}\}_{i=1,...,9} and L2={xj}j=9,17L_{2}=\{x_{j}\}_{j=9,...17}, where we are identifying the point x9x_{9} of LiL_{i} for i=1,2i=1,2. For simplicity, W1W_{1} denotes L1L_{1}. The topological space L1L_{1} was introduced in [16, Figure 2] and has the weak homotopy type of a point. It is proved in [8] that W1W_{1} is the smallest finite topological space having the same weak homotopy type of a point but not contractible. It is clear that W2W_{2} does not have beat points, so Aut(W2)Aut(W_{2}) is isomorphic to (W2)\mathcal{E}(W_{2}). On the other hand, it is easy to show that Aut(W2)Aut(W_{2}) is the trivial group. Since Px=(3,5)P_{x}=(3,5) only for x{x5,x13}x\in\{x_{5},x_{13}\} and the heights of these points are different, it follows that every homeomorphism must fix x5x_{5} and x13x_{13}. Proposition 2.9 leads to the desired result. Furthermore, the homotopy and singular homology groups of W2W_{2} are trivial. We can study the weak homotopy type of W2W_{2} studying the McCord complex 𝒦(W2)\mathcal{K}(W_{2}) or removing beat and weak beat points. We have that x16x_{16} is a weak beat point. After removing this point, x12x_{12} and x14x_{14} are up beat points. If we remove them, then it is easy to check that the remaining space is homotopy equivalent to the space given by the points {xi}i=1,,9\{x_{i}\}_{i=1,...,9}. We continue in this fashion. We have x8x_{8} is a weak beat point. After removing this point, x7x_{7} and x9x_{9} are down beat points. Thus, XX has the same weak homotopy type of a point. Therefore we have that W2W_{2} is a topological space satisfying that Aut(W2)(W2)πn(W2)Hn(W2)0Aut(W_{2})\simeq\mathcal{E}(W_{2})\simeq\pi_{n}(W_{2})\simeq H_{n}(W_{2})\simeq 0 for every n>0n>0 but it is not homeomorphic nor homotopy equivalent to a point. We can generalize this topological space by taking more copies of the topological space introduced in [16]. For instance, we can define W3W_{3} just as L3W2L_{3}\cup W_{2}, where L3={xi}i=17,,25L_{3}=\{x_{i}\}_{i=17,...,25} and we are identifying the point x17x_{17} of W3W_{3} and W2W_{2}. It is easy to prove that WnW_{n} has the weak homotopy type of a point for every nn\in\mathbb{N} because xiWnx_{i}\in W_{n} with i0(mod 8)i\equiv 0\ (\textrm{mod}\ 8) is a weak beat point.

Refer to caption
Figure 8: Hasse diagram of W2W_{2}.

One possible consequence of the previous construction is the following result.

Proposition 6.2.

Let GG and HH be finite groups. There exists a topological space XX such that Aut(X)Aut(X) is isomorphic to GG, (X)\mathcal{E}(X) is isomorphic to HH and XX is weak homotopy equivalent to a point.

Proof.

We consider the topological space XHGX_{H}^{G} obtained in the proof of Lemma 1.1 and the finite topological space W2W_{2} given in Example 6.1. Let XX denote XHGW2X_{H}^{G}\circledast W_{2}. By Proposition 2.27, Example 6.1 and the proof of Lemma 1.1 we have that Aut(X)Aut(XHG)×Aut(W2)Aut(XHG)GAut(X)\simeq Aut(X_{H}^{G})\times Aut(W_{2})\simeq Aut(X_{H}^{G})\simeq G. We get that XX is homotopy equivalent to XHW2X_{H}^{*}\circledast W_{2} by removing its beat points one by one. Since XHW2X_{H}^{*}\circledast W_{2} does not contain beat points by Corollary 2.13, it follows that (XHW2)Aut(XHW2)Aut(XH)H\mathcal{E}(X_{H}^{*}\circledast W_{2})\simeq Aut(X_{H}^{*}\circledast W_{2})\simeq Aut(X_{H}^{*})\simeq H. In addition, since W2W_{2} is collapsible, it follows that XX has the weak homotopy type of a point by Remark 2.26

We motivate the proof of Theorem 1.3 with one example.

Example 6.3.

Let us consider G=3G=\mathbb{Z}_{3} and H=2H=\mathbb{Z}_{2}. We consider the minimal finite model of the 2-dimensional sphere XX, that is, X={A,B,C,D,E,F}X=\{A,B,C,D,E,F\}, where A,B>C,D,E,FA,B>C,D,E,F and C,D>E,FC,D>E,F. Then |𝒦(X)||\mathcal{K}(X)| is homeomorphic to S2S^{2}.

We want to find a finite T0T_{0} topological space X¯HG\overline{X}_{H}^{G} such that Hn(X¯HG)H_{n}(\overline{X}_{H}^{G}) and πn(X¯HG)\pi_{n}(\overline{X}_{H}^{G}) are isomorphic to Hn(S2)H_{n}(S^{2}) and πn(S2)\pi_{n}(S^{2}) respectively for every non-negative integer nn, Aut(X¯HG)Aut(\overline{X}_{H}^{G}) is isomorphic to 3\mathbb{Z}_{3} and (X¯HG)\mathcal{E}(\overline{X}_{H}^{G}) is isomorphic to 2\mathbb{Z}_{2}. The idea is to modify XX to obtain a new space satisfying that its group of homeomorphisms is trivial. We enumerate the points of XX. For each iXi\in X with i=1,,|X|i=1,...,|X|, we add WiW_{i} to XX as in Example 6.1. The Hasse diagram of the new topological space, denoted by XX^{\prime}, can be seen in Figure 9. The Hasse diagram of XX is painted black whereas the new part is blue and purple. In purple we have the weak beat points that are not beat points. It is clear that XX^{\prime} does not have beat points so Aut(X)Aut(X^{\prime}) is isomorphic to (X)\mathcal{E}(X^{\prime}). A homeomorphism ff sends weak beat points to weak beat points by Lemma 2.19. It is easy to deduce from this that Aut(X)Aut(X^{\prime}) is the trivial group. On the other hand, the new structure added can be removed without changing the weak homotopy type of the space, see Example 6.1. Therefore, we have that Hn(X)H_{n}(X^{\prime}) is isomorphic to Hn(X)H_{n}(X) and πn(X)\pi_{n}(X^{\prime}) is isomorphic to πn(X)\pi_{n}(X) for every non-negative integer nn.

Finally, we add a new point tt that connects XX^{\prime} to XHGW2X_{H}^{G}\circledast W_{2}, where XHGX_{H}^{G} is the space obtained the proof of Lemma 1.1 for finite groups. In Figure 9 we have the Hasse diagram of the new topological space X¯GH\overline{X}_{G}^{H}. The relations with the point tt are shown in green. The Hasse diagram of XHGW2X_{H}^{G}\circledast W_{2} is painted read and orange. It is easy to check that X¯HG\overline{X}_{H}^{G} satisfies the desired properties.

Refer to caption
Figure 9: Hasse diagram of X¯HG\overline{X}_{H}^{G}.
Proof of Theorem 1.3.

If XX has the same homotopy type of a point, then the result can be deduced from Proposition 6.2. Therefore we can assume that XX does not have the same homotopy type of a point. The idea of the proof is to follow techniques similar to the ones used in the proof of Lemma 1.1. From the simplicial approximation to CW-complexes, [11, Theorem 2C.5.] we get that there exists a finite simplicial complex that is homotopy equivalent to XX. By abuse of notation we continue to write XX for the finite simplicial complex. We apply the McCord functor 𝒳\mathcal{X} to XX in order to obtain a finite T0T_{0} topological space 𝒳(X)\mathcal{X}(X) such that 𝒳(X)\mathcal{X}(X) is weak homotopy equivalent to XX. We can suppose that 𝒳(X)\mathcal{X}(X) does not have beat points or weak beat points; otherwise we can remove them one by one until there are none left. We denote by n=|𝒳(X)|n=|\mathcal{X}(X)| and label the points in 𝒳(X)\mathcal{X}(X), that is, 𝒳(X)={yi}i=1n\mathcal{X}(X)=\{y_{i}\}_{i=1...n}. For each yi𝒳(X)y_{i}\in\mathcal{X}(X) we consider WiW_{i}, where WiW_{i} is the topological space obtained in Example 6.1. We consider Z=𝒳(X)i=1,,nWiZ=\mathcal{X}(X)\cup\bigcup_{i=1,...,n}W_{i}, where we are identifying the point yiy_{i} with x1Wix_{1}\in W_{i} for every i=1,,ni=1,...,n. We define the partial order on ZZ extending the already existing partial orders. To do that we use transitivity, i.e., for x,yZx,y\in Z, xyx\geq y if and only if one of the following situations is satisfied:

  • x,y𝒳(X)x,y\in\mathcal{X}(X) and xx is greater than yy with the partial order defined on 𝒳(X)\mathcal{X}(X).

  • x,yWix,y\in W_{i} for some ii and xx is greater than yy with the partial order defined on WiW_{i}.

  • y𝒳(X)y\in\mathcal{X}(X), xWix\in W_{i} for some ii and xyi(=x1)yx\geq y_{i}(=x_{1})\geq y.

Consider X¯HG=ZXHGW2{t}\overline{X}_{H}^{G}=Z\cup X_{H}^{G}\circledast W_{2}\cup\{t\}, where XHGX_{H}^{G} is the space obtained in the proof of Lemma 1.1 and W2W_{2} the space given in Example 6.1. We extend the partial order defined on ZZ and XHGW2X_{H}^{G}\circledast W_{2} to X¯HG\overline{X}_{H}^{G} by declaring that yj<t>x1y_{j}<t>x_{1} for some hHh\in H, where yjy_{j} is a minimal point in 𝒳(X)\mathcal{X}(X) and x1W2x_{1}\in W_{2}. We prove that fAut(X¯HG)f\in Aut(\overline{X}_{H}^{G}) restricted to ZZ is the identity. In WiW_{i} there are ii weak beat points that we will denote by zjiz_{j}^{i} with j=1,,ij=1,...,i. In fact, we have that the only weak beat points that are not beat points or do not have a bigger beat point are in ZZ. Hence, if zjiWiz_{j}^{i}\in W_{i} is a weak beat point, we have f(zji)=zlkWkf(z_{j}^{i})=z_{l}^{k}\in W_{k} for some lknl\leq k\leq n. By Proposition 2.9 and Lemma 2.19, we have that f(Wi)=Wkf(W_{i})=W_{k}. But WiW_{i} is homeomorphic to WkW_{k} if and only if i=ki=k. By the continuity of ff, f(Wi)=id(Wi)f(W_{i})=id(W_{i}), so f|Z=id(Z)f_{|Z}=id(Z), as we wanted. It is easy to check that tt is also a fixed point for every homeomorphism since yjty_{j}\prec t and f(yj)=yjf(y_{j})=y_{j}. We get Aut(X¯HG)Aut(XHGW2)Aut(XHG)×Aut(W2)Aut(\overline{X}_{H}^{G})\simeq Aut(X_{H}^{G}\circledast W_{2})\simeq Aut(X_{H}^{G})\times Aut(W_{2}). By Proposition 2.27, Example 6.1 and the Proof of Lemma 1.1, Aut(X¯HG)Aut(XHG)Aut(XG)GAut(\overline{X}_{H}^{G})\simeq Aut(X_{H}^{G})\simeq Aut(X_{*}^{G})\simeq G. On the other hand, (X¯HG)(Z{t}XHW2)\mathcal{E}(\overline{X}_{H}^{G})\simeq\mathcal{E}(Z\cup\{t\}\cup X_{H}^{*}\circledast W_{2}), but Z{t}XHW2Z\cup\{t\}\cup X_{H}^{*}\circledast W_{2} does not contain beat points. Therefore, by Corollary 2.13, (Z{t}XHW2)Aut(Z{t}XHW2)\mathcal{E}(Z\cup\{t\}\cup X_{H}^{*}\circledast W_{2})\simeq Aut(Z\cup\{t\}\cup X_{H}^{*}\circledast W_{2}). From here, repeating similar arguments than the ones used before, it can be deduced that Aut(Z{t}XHW2)Aut(XH)HAut(Z\cup\{t\}\cup X_{H}^{*}\circledast W_{2})\simeq Aut(X_{H}^{*})\simeq H.

Finally, |𝒦(X¯HG)||\mathcal{K}(\overline{X}_{H}^{G})| is clearly the wedge sum of |𝒦(Z)||\mathcal{K}(Z)| and |𝒦(XHGW2)||\mathcal{K}(X_{H}^{G}\circledast W_{2})|. From Remark 2.26 we obtain that 𝒦(XHGW2)\mathcal{K}(X_{H}^{G}\circledast W_{2}) is homotopy equivalent to a point since W2W_{2} is collapsible, which implies that 𝒦(W2)\mathcal{K}(W_{2}) is also collapsible, and XHGX_{H}^{G} is homotopy equivalent to XHX_{H}^{*}. We also get that |𝒦(Z)||\mathcal{K}(Z)| is homotopy equivalent to XX because every WiW_{i} can be removed following the steps of Example 6.1 without changing the weak homotopy type of ZZ. Therefore, for every nn\in\mathbb{N}, we have πn(X)πn(Z)\pi_{n}(X)\simeq\pi_{n}(Z) and Hn(X)Hn(Z)H_{n}(X)\simeq H_{n}(Z). ∎

Proof of Corollary 1.4.

It is an immediate consequence of Theorem 1.3. We only need to consider the wedge sum of Moore spaces and then apply Theorem 1.3. ∎

Proof of Corollary 1.5.

We only need to use the beginning of the construction of Eilenberg-Maclane spaces to obtain a compact CW-complex XX with πn(X)H\pi_{n}(X)\simeq H and possibly non-trivial higher homotopy groups. Therefore, the result is an immediate consequence of Theorem 1.3. ∎

Remark 6.4.

The results obtained in this section are stated in terms of finite groups but it may be possible to get the same results for general groups. The idea could be to use the same constructions described in this section and the theory of [13], which is a generalization of the theory of R.E. Stong [17].

For a compact CWCW-complex XX, (X)\mathcal{E}_{*}(X) and #(X)\mathcal{E_{\#}}(X) are nilpotent groups, see for instance [10, Section 4]. #(X)\mathcal{E}_{\#}(X) ((X)\mathcal{E}_{*}(X)) can be seen as the kernel of a homomorphism of groups. We consider the functor π\pi (H)(H_{*}) between HPolHPol and the category of groups given by π(X)=i=1dim(X)πi(X)\pi(X)=\bigoplus_{i=1}^{dim(X)}\pi_{i}(X) (H(X)=i=1Hi(X)H_{*}(X)=\bigoplus_{i=1}^{\infty}H_{i}(X)). It is easy to check that π\pi (H)(H_{*}) induces a homomorphism of groups, π¯:(X)Aut(π(X))\overline{\pi}:\mathcal{E}(X)\rightarrow Aut(\pi(X)) (H¯:(X)Aut(H(X)))(\overline{H}_{*}:\mathcal{E}(X)\rightarrow Aut(H_{*}(X))). This sends each self-homotopy equivalence to its induced morphism in the homotopy groups (homology groups), where Aut()Aut(\cdot) denotes here the group of automorphisms of a group in the category of groups. Then, #(X)\mathcal{E}_{\#}(X) ((X)\mathcal{E}_{*}(X)) can be seen as the kernel of π¯\overline{\pi} (H¯)(\overline{H}_{*}) and so it is a normal subgroup of (X)\mathcal{E}(X). With the following example we prove that for a general topological space we cannot expect the same result.

Example 6.5.

Applying the construction obtained in the proof of Theorem 1.3 we can get a topological space XX such that Aut(X)Aut(X) is trivial, (X)=S3\mathcal{E}(X)=S_{3} and XX is weak homotopy equivalent to a circle, where S3S_{3} denotes the symmetric group on a set of 33 elements. In Figure 10 we present the Hasse diagram of XX. By construction, every self-homotopy equivalence of XX fixes the blue, black, green and orange parts of the Hasse diagram. On the other hand, the only part that contributes to the homotopy groups or homology groups is the black one. Therefore we can deduce that #(X)=(X)=(X)=S3\mathcal{E}_{\#}(X)=\mathcal{E}_{*}(X)=\mathcal{E}(X)=S_{3}, which implies that #(X)\mathcal{E}_{\#}(X) and (X)\mathcal{E}_{*}(X) are not nilpotent groups.

Refer to caption
Figure 10: Hasse diagram of XX.
Remark 6.6.

It is not difficult to show that the topological spaces obtained in the proof of Theorem 1.3 satisfy that (X)=#(X)=(X)\mathcal{E}_{*}(X)=\mathcal{E}_{\#}(X)=\mathcal{E}(X). Then it is easy to find more examples such as Example 6.5.

Acknowledgment. We thank Jesús Antonio Álvarez López for posing the question that led to one of the main results of this paper. We wish to express our gratitude to Jonathan A. Barmak for his useful comments and suggestions that substantially improved this manuscript. The authors gratefully acknowledges the many helpful corrections of Andrei Martínez Finkelshtein, Elena Castilla and Jaime J. Sánchez-Gabites during the revision of the paper.

References

  • [1] P. S. Alexandroff. Diskrete Räume. Mathematiceskii Sbornik (N.S.), 2(3):501–519, 1937.
  • [2] L. Babai. Finite digraphs with given regular automorphism groups. Period. Math. Hung., 1:257,270, 1980.
  • [3] J. A. Barmak. Algebraic topology of finite topological spaces and applications, volume 2032. Springer, 2011.
  • [4] J. A. Barmak. Automorphism groups of finite posets II. Preprint. arXiv:2008.04997, 2020.
  • [5] J. A. Barmak and E. G. Minian. Automorphism groups of finite posets. Discrete Math., 309(10):3424–3426, 2009.
  • [6] G. Birkhoff. On groups of automorphisms. Rev. Un. Mat. Argentina, 11:155–157, 1946.
  • [7] P. J. Chocano, M. A. Morón, and F. R. Ruiz del Portal. Topological realizations of groups in alexandroff spaces. Rev. R. Acad. Cien. Serie A. Mat., DOI: 10.1007/s13398-020-00964-7, 2021.
  • [8] N. Cianci and M. Ottina. Smallest weakly contractible non-contractible topological spaces. Proc. Edinburgh Math. Soc., 63(1):263–274, 2020.
  • [9] C. Costoya and A. Viruel. Every finite group is the group of self-homotopy equivalences of an elliptic space. Acta Math., 213(1):49–62, 2014.
  • [10] C. Costoya and A. Viruel. A primer on the group of self-homotopy equivalences: a Rational Homotopy Theory approach. Graduate J. Math., 5(1):76..87, 2020.
  • [11] A. Hatcher. Algebraic topology. Cambridge Univ. Press, Cambridge, 2000.
  • [12] D. W. Kahn. Realization problems for the group of homotopy classes of self-equivalences. Math. Ann., 220(1):37–46, 1976.
  • [13] M. J. Kukieła. On homotopy types of Alexandroff spaces. Order, 27(1):9–21, 2010.
  • [14] J. P. May. Finite spaces and larger contexts. Unpublished book, 2016.
  • [15] M. C. McCord. Singular homology groups and homotopy groups of finite topological spaces. Duke Math. J., 33(3):465–474, 1966.
  • [16] I. Rival. A fixed point theorem for finite partially ordered sets. J. Combin. Theory A 21, pages 309–318, 1976.
  • [17] R. E. Stong. Finite topological spaces. Trans. Amer. Math. Soc., 123(2):325–340, 1966.
  • [18] M. C. Thornton. Spaces with given homeomorphism groups. Proc. Amer. Math. Soc., 33(1):127–131, 1972.
  • [19] J. V. Whittaker. On Isomorphic Groups and Homeomorphic Spaces. Ann. of Math., 78(1):74–91, 1963.

P.J. Chocano, Departamento de Álgebra, Geometría y Topología, Universidad Complutense de Madrid, Plaza de Ciencias 3, 28040 Madrid, Spain

E-mail address:pedrocho@ucm.es

M. A. Morón, Departamento de Álgebra, Geometría y Topología, Universidad Complutense de Madrid and Instituto de Matematica Interdisciplinar, Plaza de Ciencias 3, 28040 Madrid, Spain

E-mail address: ma_moron@mat.ucm.es

F. R. Ruiz del Portal, Departamento de Álgebra, Geometría y Topología, Universidad Complutense de Madrid and Instituto de Matematica Interdisciplinar , Plaza de Ciencias 3, 28040 Madrid, Spain

E-mail address: R_Portal@mat.ucm.es