4

The title says it all: Is there an aspherical 3-manifold with $H_i(M^3) \cong \mathbb{Z}$ for i = 0 to 3?

  • 1
    Do you know anything about the fundamental group (other than that its abelianization is $\mathbb{Z}$?) What's the motivation for this question? – Jason DeVito - on hiatus Aug 12 '22 at 22:03
  • 1
    Of course, for instance all (orientable) Sol-manifolds satisfy this. – Moishe Kohan Aug 12 '22 at 23:25
  • @Moishe: are you sure? I think I can write down a mapping torus of an Anosov diffeomorphism of $T^2$ (which Wikipedia claims is always a Sol manifold) whose $H_1$ has torsion, by slightly modifying the matrix in my answer below. – Qiaochu Yuan Aug 13 '22 at 19:21

2 Answers2

2

Motivated by Moishe's claim in the comments, let's consider the mapping torus $X$ of the map $f : T^2 \to T^2$ given by the matrix $M = \left[ \begin{array}{cc} 2 & 1 \\ 1 & 1 \end{array} \right]$, which is listed on the Wikipedia article on geometrization as an example of a Sol manifold. I believe but am not sure how to prove that any mapping torus of an endomorphism of $T^2$ has universal cover $\mathbb{R}^3$, so this is a closed aspherical $3$-manifold; I also believe but am not sure how to prove that $X$ is orientable because $f$ preserves the orientation of $T^2$ (since $\det M = 1$). Hopefully someone else can fill in these details.

Given that we get $H_0(X) \cong H_3(X) \cong \mathbb{Z}$; it remains to show that $H_1(X) \cong \mathbb{Z}$ and then we'll get $H_2(X) \cong H^1(X) \cong \text{Hom}(H_1(X), \mathbb{Z}) \cong \mathbb{Z}$ by Poincare duality. $H_1(X)$ is the abelianization of $\pi_1(X)$ which is the semidirect product $\mathbb{Z}^2 \rtimes \mathbb{Z}$ with $\mathbb{Z}$ acting on $\mathbb{Z}^2$ by the matrix $M$. The abelianization of this semidirect product is the product of the coinvariants of this action on $\mathbb{Z}^2$ and $\mathbb{Z}$. Taking coinvariants means quotienting by $M - I = \left[ \begin{array}{cc} 1 & 1 \\ 1 & 0 \end{array} \right]$, which is invertible, so the quotient is trivial; hence the abelianization is $\mathbb{Z}$ as desired.

Qiaochu Yuan
  • 468,795
  • 2
    One way to see the universal cover is $\mathbb R^3$ is to geometrically unroll the space along the mapping torus direction to get a space of the form $T^2\times\mathbb R$.(You can think of this as an infinite stack of $T^2\times[0,1]$s with one end of each identified with the next via the homeomorphism $f$.) Now the universal cover of $T^2\times \mathbb R$ is indeed $\mathbb R^3$. – Cheerful Parsnip Aug 13 '22 at 20:31
  • It's orientable because it is an oriented fiber bundle over an oriented base. You can take the local orientation to be the product of the orientation of the fiber and the orientation of the base. – Connor Malin Aug 20 '22 at 21:49
1

Here's a supplement to Qiaochu's answer, proving that the mapping torus is indeed aspherical and orientable.

To see this, recall that if $T_f = M\times [0,1]/(m,0)\sim (f(m),1)$ is the mapping torus of a homeomorphism $f:M\rightarrow M$, we have a fiber bundle $M\rightarrow T_f\rightarrow S^1$, where the projectoin $\pi:T_f\rightarrow S^1$ is $\pi(m,t) = t \text{ mod } 1$.

This gives a long exact sequence in homotopy groups $$...\rightarrow \pi_k(M)\rightarrow \pi_k(T_f)\rightarrow \pi_k(S^1) \rightarrow \pi_{k-1}(M)\rightarrow ...$$

Since $S^1$ is aspherical, for $k\geq 2$, we thus obtain isomorphisms $\pi_k(M)\rightarrow \pi_k(T_f)$. Thus, if $M$ is also aspherical, then so is $T_f$.

As far as orientability is concerned, we will use Mayer-Vietoris where $U$ denotes the primage in $T_f$ of the northern hemisphere of $S^1$, and $V$ is the preimage of the southern hemisphere. (Technically, we use slightly more than the northern and southern hemispheres, so we actually cover all of $T_f$.)

Because both hemispheres are contractible, both $U$ and $V$ are diffeomorphic to $X\times (-1,1)$, and $U\cap V$ has the homotopy type of two disjoint copies of $X$. I'm thinking of $U$ as having preimage in $X\times [0,1/2]$ in $X\times [0,1]$, and I'm thinking of $V$ has having preimage $X\times [1/2,1]$ in $X\times [0,1]$.

Thus, the inclusion map of $U\cap V\cong X\coprod X\rightarrow U$ is homotopic to the identity on each copy of $X$, while the inclusion map $U\cap V\rightarrow V$ is homotopic to the identity on one copy of $X$, and homotopic to $f$ on the other.

Now, the Mayer-Vietoris homology sequence for this is $$...\rightarrow H_k(U\cap V)\rightarrow H_k(U)\oplus H_k(V)\rightarrow H_k(T_f)\rightarrow H_{k-1}(U\cap V)\rightarrow ...$$

The map $H_k(U\cap V)\rightarrow H_k(U)\oplus H_k(V)$ is a map $H_k(X)\oplus H_k(X)\rightarrow H_k(X)\oplus H_k(X)$ incuded by the inclusions. Based on the above description of the inclusions, this takes an element $(a,b)$ to $(a+b,a+f_\ast(b))$.

Now, focus on the case where $k=\dim T_f = \dim M + 1$ and $M$ is orientable. Mayer-Vietoris tells us that $H_k(T_f)$ is the kernel of hte map $H_{k-1}(U\cap V)\rightarrow H_{k-1}(U)\oplus H_{k-1}(V)$. But note that $H_{k-1}(U)\cong H_{k-1}(V)\cong H_{k-1}(M)\cong \mathbb{Z}$. The map $f_\ast$ is then an isomorphism of $\mathbb{Z}$, so it multiplication by $\pm 1$, and is $-1$ iff $f$ reverses orientation.

Thus, assuming $f$ preserves orientation, the map $(a,b)\mapsto (a+b,a+f_\ast(b))$ is simply $(a,b)\mapsto (a+b,a+b)$, which clearly has kernel $\mathbb{Z}$. Thus, $H_k(T_f)\cong \mathbb{Z}$, so $T_f$ is orientable.

  • 1
    I tried to make the answer quite general, so it proves that the mapping torus of a homeomorphism any aspherical space is aspherical (not just 3-manifolds) and it proves that the mapping torus of an orientation preserving homeomoprhism is orientable, even in the setting of topological manifolds. – Jason DeVito - on hiatus Aug 13 '22 at 20:14