8
$\begingroup$

Are there smooth closed manifolds $M^n$ in every dimension $n \geq 3$ with trivial mapping class groups and with $H^1(M^n;\mathbb{Z}/2\mathbb{Z})$ arbitrarily large?

I am under the impression that "generically" a manifold will have no mapping class group (but maybe I am totally mistaken). There are presumably lots of constructions of such manifolds with trivial mapping class groups. So I guess I'm hoping to hear of some such construction where $H^1(M^n;\mathbb{Z}/2\mathbb{Z})$ can get very large.

$\endgroup$
2
  • $\begingroup$ Not enough is known about mapping class groups of high-dimensional manifolds to answer your question. But my suspicion is high-dimensional manifolds with trivial mapping class groups are relatively rare. I.e one would have to construct them as very specific manifold types that are highly stabilized in some sense. This brings up another question -- what kind of "generic" manifold construction are you interested in? Depending on how you build your manifold, your question could easily have different answers. $\endgroup$ Jul 12, 2021 at 20:42
  • $\begingroup$ @RyanBudney I didn't have a particular model in mind for random manifolds, although I would be interested in an answer in any model. $\endgroup$
    – user101010
    Jul 27, 2021 at 17:29

2 Answers 2

5
$\begingroup$

I think that the construction in Belolipetsky--Lubotzky, Finite Groups and Hyperbolic Manifolds (https://arxiv.org/abs/math/0406607) provides a more algebraic construction of such manifolds for any $n$.

One of the results in this paper (Theorem 3.1) is the following : given groups $\Gamma \triangleright \Delta \triangleright M$ with $\Gamma/\Delta$ finite and $F = \Delta/M$ a nonabelian free group, there exists $\Gamma \ge D \ge \Delta$ and $B \le D$ of finite index with $N_\Gamma(B) = B$ (in fact they prove that for any finite group $G$ there are infinitely many $B$ with $N_\Gamma(B)/B$ isomorphic to $G$).

They use this to construct closed hyperbolic manifolds with trivial isometry group ; by Mostow rigidity the mapping class group of $X$ is isomorphic to the isometry group of $X$ (every homeomorphism is isotopic to an isometry), so these manifolds will have trivial mapping class group. Without too much details their construction uses a specific lattice $\Gamma$ in the isometry group of hyperbolic $n$-space $\mathbb H^n$, and subgroups $D, \Delta, M, B$ as in the previous paragraph ; the manifold is the quotient of $\mathbb H^n$ by $B$ (their construction ensures that $B$ is torsion-free though $\Gamma$ will likely not be).

Now to look at Betti numbers, for which we need to dive a bit more into the details of this paper. Since $X$ is aspherical we have $$b_1(X, \mathbb Q) = b_1(B, \mathbb Q) \ge b_1(D, \mathbb Q).$$ On the other hand $b_1(D, \mathbb Q)$ is the dimension of the fixed subspace of $D/\Delta$ in $H^1(\Delta, \mathbb Q)$ (conjugation action on the normal subgroup $\Delta$ and its morphisms to $\mathbb Z$). By the definition of $D$ in the paper of Belolipetsky--Lubotzky (p. 4 in the arxiv version) we have that it acts trivially on the part of the cohomology coming from the surjective morphism $\Delta \to F$. So we have that $b_1(X, \mathbb Q)$ is larger than the rank of $F$. It remains to remark that in the data for Theorem 3.1 the quotient $\Delta/M$ can be taken to have arbitrarily large rank (since a non-abelian free group contains free subgroups of finite index with arbitrarily large rank), so we get examples with $b_1(B, \mathbb Q)$ arbitrarily large.

$\endgroup$
4
  • $\begingroup$ Thank you for this insightful answer! I am confused about your very last sentence. In particular, I don't see why the fact about subgroups of non-abelian free groups applies in this instance to show that the data in Theorem 3.1 can be taken with $\Delta/M$ of arbitrary rank. Could you explain that a little more? $\endgroup$
    – user101010
    Jul 27, 2021 at 17:28
  • $\begingroup$ Let $\Delta'$ be the preimage in $\Delta$ of any finite-index subgroup $F$ of $\Delta/M$, and $\Delta''$ its normal core in $\Gamma$. Then the hypotheses will still be satisfied for $\Gamma \triangleright \Delta'' \triangleright M$, and $\Delta''/M$ is a finite-index subgroup of $F$ so it has rank at least that of $F$. Since we can choose $F$ of arbitrarily large rank we can also do so for $\Delta/M$. $\endgroup$ Jul 28, 2021 at 14:27
  • 1
    $\begingroup$ Sorry for this delay, why is $M$ a subgroup of $\Delta''$? $\endgroup$
    – user101010
    Sep 20, 2021 at 13:58
  • $\begingroup$ i was a bit hasty writing the previous comment, i don't think $M \subset \Delta''$ in general. But i think we can replace $M$ with $M'' = \Delta'' \cap M$ and keep the hypotheses right for $\Gamma \triangleright \Delta'' \triangleright M''$. $\endgroup$ Sep 29, 2021 at 7:58
3
$\begingroup$

Here is one such construction in dimension three. Suppose that $G$ is a finite, connected, simple graph without symmetries. We build a manifold $M^3$ by taking manifolds with boundary for the vertices of $G$ and gluing them in pairs “along” the edges. We arrange matters so that the vertex manifolds are hyperbolic and with totally geodesic boundary. We also require that the vertex manifolds have no reflection symmetries. Finally we glue boundary components in pairs according to the edges of the graph, using high powers of pseudo-Anosov maps.

If the power is high enough, then any mapping class of $M$ sends vertex manifolds to each other, and so preserves the graph structure. Thus the manifold $M$ has no nontrivial mapping classes.

We can arrange also for each vertex manifold to contribute to homology. Thus the rank of the first (co)homology is linear in the number of vertex manifolds.

Note that, with a bit more work, such manifolds are hyperbolic. Thus we know that their mapping class groups are finite. The messing about (with high powers of pseudo-Anosov maps) is needed to enforce “rigidity”: that is, to reduce the analysis from symmetries of $M$ to symmetries of $G$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.