Phantom maps and rational homotopy

The concept of a phantom map is a key to understanding maps with infinite dimensional sources, and has been an important topic in homotopy theory since its discovery ([8, 14]). In this paper, we investigate the relationship between phantom maps and rational homotopy, which is a central concern in the study of phantom maps (see [8, 14, 10, 9, 2, 3, 12])

In this section, we make a brief review of the basic notions and results on phantom maps, and present the main results of this paper.

Given two pointed $CW$-complexes $X$ and $Y$, a map $f:X\longrightarrow Y$ is called a phantom map if for any finite complex $K$ and any map $h:K\longrightarrow X$, the composite $fh$ is null homotopic. Let ${\mathrm{Ph}}(X,Y)$ denote the subset of $[X,Y]$ consisting of homotopy classes of phantom maps.

We are also interested in the subset ${\mathrm{SPh}}(X,Y)$ of ${\mathrm{Ph}}(X,Y)$ consisting of homotopy classes of special phantom maps, defined by the exact sequence of pointed sets

where $e_{Y}:Y\longrightarrow\check{Y}=\underset{p}{\prod}\ Y_{(p)}$ is a natural map called the local expansion (cf. [14, p. 150]). The target $Y$ is usually assumed to be nilpotent of finite type.

The following theorem is due to McGibbon-Roitberg and Iriye; the statements on phantom maps and special phantom maps are Theorem 2 in [10] and Theorem 1.4 in [2] respectively. See Remark 1.1 for the basic notions used in Theorem A.

The following theorem is due to Meier ([12, Theorem 5]). Recall that a space is called an $H_{0}$-space if its rationalization is homotopy equivalent to a product of Eilenberg-MacLane spaces. Let $\hat{{\mathbb{Z}}}$ denote the product $\Pi_{p}\hat{{\mathbb{Z}}}_{p}$ of the $p$-completions of ${\mathbb{Z}}$, in which ${\mathbb{Z}}$ is diagonally contained.

We give generalizations of Theorems A and B as two main theorems, and apply them to calculations of ${\mathrm{Ph}}(X,Y)$ and ${\mathrm{SPh}}(X,Y)$. The proofs are based on the approach introduced in a previous paper [4], which largely extends the rationalization-completion approach developed by Meier and Zabrodsky ([8, Section 5]). Our approach is so general and enables us to give a generalization of Theorem A along with its simple proof. We also provide a generalization of Theorem B along with its dual version; the important idea of Theorem B has been largely overlooked in the literature.

Let ${\mathcal{C}}{\mathcal{W}}$ denote the category of pointed connected $CW$-complexes and homotopy classes of maps and let ${\mathcal{N}}$ denote the full subcategory of ${\mathcal{C}}{\mathcal{W}}$ consisting of nilpotent $CW$-complexes of finite type.

The first main theorem is the following.

The following corollary, and hence Theorem 1.2 can be regarded as a generalization of Theorem A.

Corollary 1.3 is used to obtain new vanishing results for ${\mathrm{Ph}}(X,Y)$ (see Proposition 2.1, Corollary 2.2, and Example 2.4).

Next, to state the second main theorem, we recall the basics of ${\mathrm{Ph}}(X,Y)$ and ${\mathrm{SPh}}(X,Y)$ from [4]. The set ${\mathrm{Ph}}(X,Y)$ (resp. ${\mathrm{SPh}}(X,Y)$) can be described as the orbit space of $[X,F_{Y}]$ (resp. $[X,F^{\prime}_{Y}]$) by the natural action of $[X,\Omega\hat{Y}]$ (resp. $[X,\Omega\check{Y}]$), where $F_{Y}$ (resp. $F^{\prime}_{Y}$) is the homotopy fiber of the profinite completion $c_{Y}:Y\longrightarrow\hat{Y}$ (resp. the local expansion $e_{Y}:Y\longrightarrow\check{Y}$). Hence, we have the exact sequence of pointed sets

(see [4, Lemma 3.5 and Corollary 5.3]). Further, there exist (noncanonical) bijections

where $\check{{\mathbb{Z}}}$ denotes the product $\Pi_{p}{\mathbb{Z}}_{(p)}$ of the $p$-localizations of ${\mathbb{Z}}$, in which ${\mathbb{Z}}$ is diagonally contained (see [4, Proposition 5.4 and Remark 5.6]).

Part 2 of the following corollary is a generalization of Theorem B (see Remark 3.3) and Part 1 is its dual version.

Corollary 1.5 can be applied to find various pairs $(X,Y)$ for which ${\mathrm{Ph}}(X,Y)\cong\underset{i>0}{\Pi}H^{i}(X;\pi_{i+1}(Y)\otimes\hat{{\mathbb{Z}}}/{\mathbb{Z}})$ hold (see Corollary 3.4, Proposition 3.5, Corollary 3.6 and Example 3.7).

The results in this section are proved in Sections 2 and 3.

In this section, we prove Theorem 1.2 and Corollary 1.3, and apply Corollary 1.3 to obtain new vanishing results for ${\mathrm{Ph}}(X,Y)$. In this and next sections, the subscript ($0$) denotes the rationalization of a nilpotent space or a nilpotent group.

In the rest of this section, we derive vanishing results for ${\mathrm{Ph}}(X,Y)$ from Corollary 1.3. Recall the following basic vanishing results:

• •

  ${\mathrm{Ph}}(K,Y)=0$ for a finite complex $K$.

• •

  ${\mathrm{Ph}}(\Omega K,Y)=0$ for a $1$-connected finite complex $K$ and $Y\in{\mathcal{N}}$.

The first is obvious from the definition of a phantom map. The second is a result of Iriye [3, Theorem 1.A].

The $n$-connected cover $X\langle n\rangle$ of $X$ is called $good$ if the canonical map $X\langle n\rangle\longrightarrow X$ induces a monomorphism on the rational homology.

Recall the definition of an $H_{0}$-space from Section 1.

We end this section with an application of Corollary 1.3(2). For a $G$-space $X$, the homotopy orbit space $X_{hG}$ of $X$ (or the Borel construction on $X$) is defined by

where $EG$ is the total space of the universal principal $G$-bundle (see [6]).

In this section, we prove Theorem 1.4 and Corollary 1.5, and apply Corollary 1.5 to find various pairs $(X,Y)$ for which ${\mathrm{Ph}}(X,Y)\cong\underset{i>0}{\Pi}H^{i}(X;\pi_{i+1}(Y)\otimes\hat{{\mathbb{Z}}}/{\mathbb{Z}})$ hold.

For the proof of Theorem 1.4, we need the following lemma.

Let us recall the generalizations of Miller’s theorem [13] and Anderson-Hodgkin’s theorem [1] to obtain many pairs $(X,Y^{\prime})$ with $[X,F_{Y^{\prime}}]\xrightarrow[\cong]{\pi}{\mathrm{Ph}}(X,Y^{\prime})$. A space whose $i^{\rm th}$ homotopy group is zero for $i\leq n$ and locally finite for $i=n+1$ is said to be $n\frac{1}{2}$-connected. Define the classes $\mathcal{A}$, $\mathcal{B}$, $\mathcal{A}^{\prime}$ and $\mathcal{B}^{\prime}$ by

1. $\mathcal{A}$ =

   the class of $\frac{1}{2}$-connected Postnikov spaces, the classifying spaces of compact Lie groups, $\frac{1}{2}$-connected infinite loop spaces and their iterated suspensions.

2. $\mathcal{B}$ =

   the class of nilpotent finite complexes, the classifying spaces of compact Lie groups and their iterated loop spaces.

3. $\mathcal{A}^{\prime}$ =

   the class of ${1}\frac{1}{2}$-connected Postnikov spaces of finite type and their iterated suspensions.

4. $\mathcal{B}^{\prime}$ =

   the class of $BU$, $BO$, $BSp$, $BSO$, $U/Sp$, $Sp/U$, $SO/U$, $U/SO$, and their iterated loop spaces.

If $(X,Y^{\prime})$ is in ${\mathcal{A}}\times{\mathcal{B}}$ or ${\mathcal{A}}^{{}^{\prime}}\times{\mathcal{B}}^{{}^{\prime}}$, then ${[X,\Omega\hat{Y^{\prime}}]=0}$ ([4, Corollary 6.4]), and hence $[X,F_{Y^{\prime}}]\xrightarrow[\cong]{\pi}{\mathrm{Ph}}(X,Y^{\prime})$ and $[X,F^{\prime}_{Y^{\prime}}]\xrightarrow[\cong]{\pi^{\prime}}{\mathrm{SPh}}(X,Y^{\prime})$ ([4, Propositions 6.1 and 2.5]).

Part 1 of the following corollary is a direct generalization of [12, Corollary 2 in Section 2]. For a connected $CW$-complex $K$, $Q(K)$ denotes the infinite loop space defined by $Q(K)=\underset{n}{\rm colim}\,\Omega^{n}\Sigma^{n}K$.

Corollary 3.6, and hence Proposition 3.5 can be regarded as a generalization of [5, Corollary 1.2] (see Remark 3.2).

We end this section with an application of Corollary 1.5(1). The Grassmannians $G_{n}(\mathbb{F})$ and $G_{\infty}(\mathbb{F})$ are defined by $G_{n}(\mathbb{F})=\underset{\longrightarrow}{\lim}_{m}\ G_{n,m}(\mathbb{F})$ and $G_{\infty}(\mathbb{F})=\underset{\longrightarrow}{\lim}_{n}\ G_{n}(\mathbb{F})$ for $\mathbb{F}=\mathbb{R},\mathbb{C},\mathbb{H}$, where the finite Grassmannian $G_{n,m}(\mathbb{F})$ is the space of $n$-dimensional subspaces in $\mathbb{F}^{n+m}$.

If $(G_{n}(\mathbb{F})/G_{n^{\prime},m^{\prime}}(\mathbb{F}),Y)$ is in ${\mathcal{Q}}$, then we can obtain a more precise result (see [4, Example 6.6]).