2. Fundamental group at infinity

Following [Reference Guilbault10], we start by recalling the algebra of inverse sequences, that will be useful for the definition of the fundamental group at infinity.

Let

\begin{equation*}G_0 \overset {f_1}{\longleftarrow } G_1 \overset {f_2}{\longleftarrow } G_2 \overset {f_3}{\longleftarrow } \ldots \end{equation*}

be an inverse sequence of groups and homomorphisms (such a sequence will be denoted by $\{G_i, f_i\}$ henceforth). Let $i,j$ be two natural numbers such that $i\gt j$ ; after denoting the composition $f_{j}\circ f_{j+1} \ldots \circ f_{i}$ by $f_{ij}$ , we can naturally define a subsequence of $\{G_i, f_i\}$ as follows:

\begin{equation*}G_{i_0} \overset {f_{i_1i_0}}{\longleftarrow } G_{i_1} \overset {f_{i_2i_1}}{\longleftarrow } G_{i_2} \overset {f_{i_3i_2}}{\longleftarrow } \ldots \ .\end{equation*}

Two sequences $\{G_i, f_i\}$ and $\{H_i, g_i\}$ are said to be pro-isomorphic if there exists a commuting diagram of the form

for some subsequences $\{G_{i_k}, f_{i_ki_{k-1}}\}$ and $\{H_{j_k}, g_{j_kj_{k-1}}\}$ of $\{G_i, f_i\}$ and $\{H_j, g_j\}$ , respectively.

The inverse limit of a sequence $\{G_i, f_i\}$ is defined as

\begin{equation*}\underset {\longleftarrow }{\lim } \{G_i, f_i\}=\Big \{(g_0, g_1, \ldots ) \in \prod _{i \in \mathbb {N}}G_i \big | f_i(g_i)=g_{i-1}\Big \},\end{equation*}

with the group operation defined componentwise; this is a subgroup of $\prod _{i \in \mathbb{N}}G_i$ . For each $i \in \mathbb{N}$ , we have a natural projection homomorphism $p_i\,:\, \underset{\longleftarrow }{\lim } \{G_i, f_i\} \rightarrow G_i$ .

It can be easily seen that inverse limits of pro-isomorphic sequences are isomorphic, while we remark that two sequences with isomorphic inverse limit are not necessarily pro-isomorphic; for example the trivial sequence

\begin{equation*}1 \longleftarrow 1 \longleftarrow 1 \longleftarrow \ldots \end{equation*}

and the following one

\begin{equation*} \mathbb{Z} \overset{\cdot 2}{\longleftarrow }\mathbb{Z} \overset{\cdot 2}{\longleftarrow }\mathbb{Z} \overset{\cdot 2}{\longleftarrow } \ldots \end{equation*}

have both inverse limit isomorphic to the trivial group, but it is straightforward that they are not pro-isomorphic.

We distinguish some particular classes of sequences, denoting by $\{G, \textrm{id}_G\}$ a sequence $\{G_i, f_i\}$ such that $G_i \cong G$ and $f_i=\textrm{id}_G$ for all $i\in \mathbb{N}$ , we say that:

• a sequence pro-isomorphic to one of the form $\{1, \textrm{id} \}$ is called pro-trivial;

• a sequence pro-isomorphic to one of the form $\{H, \textrm{id}_H\}$ is called stable;

• a sequence pro-isomorphic to one whose maps are all surjective is called semistable.

In the case of stable sequences, we remark that $H$ is well defined up to isomorphism, since the inverse limit of the sequence is isomorphic to $H$ .

Proof. By definition of inverse limit, if the $f_i$ ’s are surjective we have that the projections $p_i\,:\,G \rightarrow G_i$ are surjective too; by definition of the limit topology on $G$ , we have that the projections are also continuous. Then, each $G_i$ is a continuous image of an amenable group, so it is amenable [Reference Zimmer20, Lemma 4.1.13].

Following again [Reference Guilbault10], we recall the definition of fundamental group at infinity of a manifold. Let $M$ be a triangulable manifold (henceforth, all manifolds will be triangulable) without boundary, and let $A \subset M$ be any subset; we say that a connected component of $M \setminus A$ is bounded if it has compact closure; otherwise, we say that it is unbounded. A neighborhood of infinity is a subset $N \subset M$ such that $M \setminus N$ is bounded. We say that a neighborhood of infinity is clean if it satisfies the following properties:

• $N$ is closed in $M$ ;

• $N$ is a submanifold of codimension $0$ with bicollared boundary.

It can be easily shown that every neighborhood of infinity contains a clean neighborhood of infinity.

Let $k\geq 0$ be a natural number. A manifold $M$ is said to have $k$ ends if there exists a compact subset $C \subset M$ such that for any other compact subset $D$ such that $C \subset D$ , $M \setminus D$ has exactly $k$ unbounded components; this implies that $M$ contains a clean neighborhood of infinity with $k$ connected components. If such $k$ does not exist, we say that the manifold $M$ has infinitely many ends.

Let us suppose that $M$ is a $1$ -ended manifold without boundary. A clean neighborhood of infinity is called a $0$ -neighborhood of infinity if it is connected and has connected boundary. A sequence of nested neighborhoods of infinity $U_0 \supset U_1 \supset U_2 \supset \ldots$ is neat if $U_{i+1} \subseteq \textrm{int}(U_i)$ for each $i \geq 0$ and $\cap _{i=0}^{+\infty }U_i= \varnothing$ ; by [Reference Guilbault10, Lemma 5] each 1-ended open manifolds admits a neat sequence of clean connected neighborhoods of infinity.

Let $U_0 \supset U_1 \supset U_2 \supset \ldots$ be a neat sequence of nested $0$ -neighborhoods of infinity; a base-ray is a proper map $\rho \,:\,[0,+ \infty ) \rightarrow M$ . For each $i\in \mathbb{N}$ let $p_i\,:\!=\,\rho (i)$ ; up to reparametrization, we can assume that $p_i\in U_i$ and $\rho ([i, +\infty ))\subset U_i$ . The inclusions $U_{i-1} \hookleftarrow U_{i}$ along with the change of basepoints maps induced by $\rho |_{[i-1, i]}$ give rise to the following sequence

\begin{equation*}\pi _1(U_0, p_0) \overset {\lambda _1}{\longleftarrow } \pi _1(U_1, p_1) \overset {\lambda _2}{\longleftarrow } \pi _1(U_2, p_2) \overset {\lambda _3}{\longleftarrow } \ldots .\end{equation*}

We denote the inverse limit of this sequence by $\pi _1^{\infty }(M, \rho )$ . If the sequence is semistable, it can be shown that its pro-isomorphism class does not depend neither on the base-ray $\rho$ nor on the choice of the sequence of neighborhoods of infinity (see e.g. [Reference Geoghegan6, Section 16.1]). In this case, we can define the fundamental group at infinity of $M$ as

\begin{equation*}\pi _1^{\infty }(M)\,:\!=\, \underset {\longleftarrow }{\lim } (\pi _1(U_i, p_i), \lambda _i).\end{equation*}

If $M$ has more than one end, let $U_0 \supset U_1 \supset U_2 \supset \ldots$ be a neat sequence of nested clean neighborhoods of infinity; we say that two base-rays $\rho _1, \rho _2$ determine the same end if their restrictions $\rho _1|_{\mathbb{N}}$ and $\rho _2|_{\mathbb{N}}$ are properly homotopic. In the case of multiple ends, we need to specify the base-ray to define the fundamental group at infinity of each end; however, if the sequence that defines the fundamental group at infinity at each end is semistable we have that if $\rho _1$ and $\rho _2$ determine the same end, then $\pi _1^{\infty }(M, \rho _1)\cong \pi _1^{\infty }(M, \rho _2)$ (see again [Reference Geoghegan6, Section 16.1]).

The following definition is a weakening of the notion of tame manifold.

In other words, this means that every neighborhood of infinity $U$ can be shrunk into a compact subset within $U$ ; we remark that being inward tame implies that each clean neighborhood of infinity has finitely presented fundamental group [Reference Guilbault and Tinsley9, Lemma 2.4]. According to our definitions, it is immediate to check that a tame manifold is also inward tame. On the other hand, well-known examples of inward tame manifolds which are not tame are the exotic universal covers of the aspherical manifolds produced by Davis in [Reference Davis3] (see e.g. [Reference Guilbault, Davis, Fowler, Lafont and Leary11, Example 3.2.5]). Being inward tame ensures some nice properties for the topology at infinity of a manifold, as shown by the following proposition.

Hence, for inward tame manifolds the fundamental group at infinity at each end is well defined (i.e. it does not depend on the base-ray which determines the end or on the chosen sequence of neighborhoods of infinity).

In the introduction, we defined manifolds simply connected at infinity. In the finitely-many-ended case, this definition can be read in terms of pro-triviality of the sequence that defines the fundamental group at infinity at each end.

We stress that a manifold simply connected at infinity need not be inward tame, not even in the 1-ended case [Reference Guilbault, Davis, Fowler, Lafont and Leary11, Example 3.5.11].

We state a result from [Reference Guilbault10] that ensures that, in dimension $n\geq 5$ , a sequence $\{G_i, f_i\}$ pro-isomorphic to one that defines the fundamental group at infinity can be realized as a sequence of fundamental groups of neighborhoods of infinity; we remark that the constraint on the dimension is due to a transversality argument, which we need to assume that some 2-discs are pairwise disjoint. The result is originally stated for inward tame manifolds; however, the author remarks that it is enough to suppose that the fundamental groups of the neighborhoods of infinity are finitely presented [Reference Guilbault10, Section 6].

3. Simplicial volume and amenable covers

Let $X$ be a topological space and let $R$ be either the ring of the integers $\mathbb{Z}$ or of the real numbers $\mathbb{R}$ . We denote by $S_n(X)$ the set of singular $n$ -simplices on $X$ for a given natural number $n\in \mathbb{N}$ . A singular $n$ -chain on $X$ with $R$ -coefficients is a sum $\sum _{i\in I}a_i \sigma _i$ , where $I$ is a (possibly infinite) set of indices, $a_i \in R$ and the $\sigma _i$ ’s are singular simplices. A chain is said to be locally finite if for every compact subset $K \subset X$ there are only finitely many simplices of the chain whose image intersects $K$ . We denote by $C_n^{{\text{lf}}}(X;\, R)$ the module of singular locally finite $n$ -chains on $X$ with $R$ -coefficients. We observe that the obvious extension of the usual boundary operator sends locally finite chains to locally finite chains; hence, it defines a boundary operator on $C_{*}^{{\text{lf}}}(X;\, R)$ ; the homology of this complex with respect to this differential is denoted by $H_{*}^{{\text{lf}}}(X;\,R)$ and is called the locally finite homology of $X$ with $R$ -coefficients. We remark that, if $X$ is compact, then the locally finite homology coincides with the usual singular homology.

From now on, unless otherwise stated, when we omit the coefficients we will understand that $R=\mathbb{R}$ .

We can endow the complex of locally finite chains $C_{*}^{{\text{lf}}}(X)$ with the $\ell ^1$ -norm as follows: for a chain $c=\sum _{i\in I}a_i \sigma _i$ in reduced form we set

\begin{equation*}||c||_1\,:\!=\,\sum _{i \in I}|a_i|.\end{equation*}

We stress that a locally finite chain may have an infinite $\ell ^1$ -norm.

This norm induces a seminorm on $H_{*}^{{\text{lf}}}(X)$ , which we will again denote by $||\cdot ||_1$ .

From standard algebraic topology (see e.g. [Reference Löh12, Theorem 5.4]), we have that if $M$ is an oriented connected $n$ -manifold, then $H_n^{{\text{lf}}}(M, \mathbb{Z})\cong \mathbb{Z}$ , and this allows us to define the fundamental class $[M]_{\mathbb{Z}}\in H_n^{{\text{lf}}}(M, \mathbb{Z})$ as a generator of this group. The image of $[M]_{\mathbb{Z}}$ under the change of coefficients map $H_n^{{\text{lf}}}(M, \mathbb{Z}) \rightarrow H_n^{{\text{lf}}}(M)$ is called the real fundamental class and is denoted by $[M]\in H_n^{{\text{lf}}}(M)$ .

If $M$ is an oriented connected $n$ -manifold, the simplicial volume of $M$ is defined as

\begin{equation*}||M||=||[M]||_1.\end{equation*}

We state two results from [Reference Frigerio and Moraschini4] that give finiteness or vanishing conditions on the simplicial volume in terms of the existence of amenable covers with small multiplicity. First, we give some definitions.

We say that a subset $U$ of a topological space $X$ is amenable if $i_*(\pi _1(U,p))$ is an amenable subgroup of $\pi _1(X,p)$ for any choice of basepoint $p \in U$ , where $i\,:\, U \hookrightarrow X$ denotes the inclusion. Following [Reference Frigerio and Moraschini4], we give the following definition.

4. Proof of Theorem 1, Corollary 2, and Theorem 3

Proof of Theorem 1. First, let us deal with the 1-ended case. Since $M$ is inward tame, by Proposition 2.5 we have that $M$ has semistable fundamental group at infinity. Let $\mathcal{G}=\{G_0 \leftarrow G_1 \leftarrow \ldots \}$ be a sequence whose maps are all surjective which is pro-isomorphic to a sequence that realizes the fundamental group at infinity of $M$ . Let $\{U_j\}_{j \in \mathbb{N}}$ be a neat sequence of $0$ -neighborhoods of infinity given by Proposition 2.9, i.e. such that $\{n_j\}_{j\in \mathbb{N}}$ is a strictly increasing sequence, $\pi _1(U_j) \cong G_{n_j}$ and that the inclusions $U_{j+1} \hookrightarrow U_{j}$ induce surjective maps between the fundamental groups for each $j \in \mathbb{N}$ . By Lemma 2.1, we have that $\pi _1(U_j)$ is amenable for each $j \in \mathbb{N}$ . Let $V_j\,:\!=\, U_j \setminus \textrm{int}(U_{j+1})$ ; for $j \geq 0$ , since $U_j \subseteq \textrm{int} (U_{j+1})$ , $V_j$ is a connected codimension- $0$ submanifold of $M$ with two boundary components, which are $\partial U_j$ and $\partial U_{j+1}$ .

For every $j \in \mathbb{N}$ , let us fix a regular neighborhood $N_j$ in $M$ of the boundary component of $U_j$ and a homeomorphism $N_j \cong \partial U_j \times [{-}1,1]$ such that $\partial U_j \times [{-}1,0)$ is contained in $U_{j-1} \setminus U_j$ and $\partial U_{j+1} \times (0,1]$ is contained in $U_{j+1} \setminus U_{j+2}$ , and such that $N_j \cap N_{j+1}= \emptyset$ (this is possible since the sequence is neat). Now, for each $j \in \mathbb{N}$ let us set $\widetilde{V_j}\,:\!=\, V_j \bigcup \partial U_j \times (-1/2,0] \bigcup \partial U_{j+1} \times [0, 1/2)$ . Moreover, let us set $\widetilde{V}_{-1}\,:\!=\, \textrm{int} (M \setminus U_0) \bigcup \partial U_0 \times [0, 1/2)$ . We have that $\widetilde{V_j}$ is open and relatively compact, and that the open cover $\mathcal{V}=\{\widetilde{V_j}\}_{j={-1}}^{+\infty }$ has multiplicity $2$ . Moreover, for each $j \in \mathbb{N}$ we have that $\widetilde{V_j}$ is contained in $\widetilde{U_j}\,:\!=\,U_j \bigcup \partial U_j \times (-1,0]$ ; the $\widetilde{U_j}$ ’s are locally finite and homotopy equivalent to the $U_j$ ’s, so in particular they have an amenable fundamental group. Hence, the image of $\pi _1(\widetilde{V_j})$ in $\pi _1(\widetilde{U_j})$ under the map induced by the inclusion $\widetilde{V_j}\hookrightarrow \widetilde{U_j}$ is also amenable. It follows that $\mathcal{V}$ is amenable at infinity. We can conclude that $||M||\lt + \infty$ thanks to Theorem 3.2.

If $M$ has $k$ ends and $k \in \mathbb{N}_{\geq 2}$ , thanks to Remark 2.10 we can simply apply the same procedure to each end separately and get the same conclusion.

Proof of Corollary 2. In the 1-ended case, for each $j\in \mathbb{N} \cup \{-1\}$ let us take $\widetilde{V_j}$ as in the proof of Theorem 1: since $\pi _1(M)$ is amenable, we have that the $\widetilde{V_j}$ ’s form an amenable open cover with the same properties as above, so by Theorem 3.3 we have that $||M||=0$ .

If $M$ has more than one end, the conclusion follows again from Remark 2.10.

We extend Theorem 1 and Corollary 2 to the case of finitely-many-ended manifolds that are simply connected at infinity (which, as remarked in Section 2, need not be inward tame).

Proof of Theorem 3. As usual, we write the proof in the 1-ended case, noting that it can be generalized to the finitely-many-ended one thanks to Remark 2.10. We are going to prove that if $M$ is simply connected at infinity then the fundamental groups of neighborhoods of infinity of $M$ are finitely presented, so that we can apply Proposition 2.9. Let $U_0 \supset U_1 \supset U_2 \supset \ldots$ be a neat sequence of nested $0$ -neighborhoods of infinity; by Lemma 2.7, the induced sequence on the fundamental groups is pro-trivial, i.e. (up to taking a subsequence) it fits in a diagram of the following form:

where $\lambda _j\,:\, \pi _1(U_{j+1})\rightarrow \pi _1(U_j)$ is the map induced by the inclusion $U_{j+1} \hookrightarrow U_j$ which, by commutativity of the diagram, is the trivial map. Let us fix $j\in \mathbb{N}$ , let $K$ be the compact subset $U_j \setminus \textrm{int} U_{j+1}$ . By Van Kampen Theorem, we have that

\begin{equation*}\pi _1(U_j) \cong \pi _1(K) *_{\pi _1(\partial U_{j+1})} \pi _1(U_{j+1}).\end{equation*}

Let $\mu$ denote the map induced by the inclusion $\partial U_{j+1} \hookrightarrow K$ . Since the elements of $\pi _1(U_{j+1})$ are trivial in the amalgamated product, we can simplify the standard presentation of the amalgamated product by setting the generators of $\pi _1 (U_{j+1})$ equal to $1$ . In the end, the presentation of $\pi _1(U_j)$ is given by the generators of $\pi _1(K)$ and the relations given by the ones of $\pi _1(K)$ and the image via $\mu$ of the generators of $\pi _1(\partial U_{j+1})$ . From the compactness of $K$ and $\partial U_{j+1}$ , it follows that $\pi _1(U_j)$ is finitely presentable. Therefore we can apply Proposition 2.9 to obtain a sequence of simply connected neighborhoods of infinity, and then we can conclude that $||M||\lt +\infty$ as in the proof of Theorem 1. If in addition $\pi _1(M)$ is amenable, the vanishing of $||M||$ is deduced as in the proof of Corollary 2.