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

Profinite rigidity and hyperbolic four-punctured sphere bundles over the circle

Tamunonye Cheetham-West Department of Mathematics
Yale University
New Haven, CT, 06511
[email protected]
(Date: Fall 2024)
Abstract.

We show that hyperbolic four-punctured S2S^{2}-bundles over S1S^{1} are distinguished by the finite quotients of their fundamental groups among all 3-manifold groups. To do this, we upgrade a result of Liu to show that the topological type of a fiber is detected by the profinite completion of the fundamental group of a fibered hyperbolic 3-manifold.

1. Introduction

For a compact, connected 3-manifold MM, it is interesting to understand what properties of MM are detected by the profinite completion π1(M)^\widehat{\pi_{1}(M)}; i.e. the inverse limit of finite quotients of π1(M)\pi_{1}(M). If MM is the only compact 3-manifold with profinite completion π1(M)^\widehat{\pi_{1}(M)}, we say that π1(M)\pi_{1}(M) is profinitely rigid among 3-manifold groups.

For example, if two compact 3-manifolds MM and NN have π1(M)^π1(N)^\widehat{\pi_{1}(M)}\cong\widehat{\pi_{1}(N)}, it is a consequence of work of Lott and Lück [Luck1995] [LuckApprox] on the first L2L^{2}-Betti number of compact 3-manifolds that MM is irreducible if and only if NN is. Furthermore, for an irreducible manifold, Wilton and Zalesskii [WZ1] prove that the profinite completion determines whether a manifold is finite-volume hyperbolic. Remarkably, Liu [Y] showed that any set of finite-volume hyperbolic 3-manifolds whose fundamental groups have a fixed common profinite completion is always finite.

Throughout, Σg,p\Sigma_{g,p} will refer to a genus gg surface with pp punctures. Bridson, Reid, and Wilton [BRW] show that whenever the Mapping Class Group Mod(Σg,p)Mod(\Sigma_{g,p}) of a surface Σg,p\Sigma_{g,p} is omnipotent and has the Congruence Subgroup Property (see Section 5 for definitions), hyperbolic Σg,p\Sigma_{g,p} bundles over S1S^{1} with first Betti number 1 with the same finite quotients have a common finite-sheeted cyclic cover. In particular, they show that all hyperbolic Σ1,1\Sigma_{1,1} bundles are profinitely rigid among 3-manifolds. We combine this circle of ideas with the work of Liu [Y] (see Section 3 for a description) to prove:

Theorem 1.1.

Let MM be a hyperbolic Σ0,4\Sigma_{0,4} bundle over S1S^{1}. Then π1(M)\pi_{1}(M) is profinitely rigid among 3-manifold groups.

In particular, the work of Liu [Y] allows us to remove the first Betti number 1 condition.

Theorem 1.2.

Let MM and NN be finite-volume hyperbolic manifolds with π1(M)^π1(N)^\widehat{\pi_{1}(M)}\cong\widehat{\pi_{1}(N)}. Liu’s Thurston-norm and fiber class preserving isomorphism H1(N,)H1(M,)H^{1}(N,{\mathbb{Z}})\to H^{1}(M,{\mathbb{Z}}) induced by an isomorphism Φ:π1(M)^π1(N)^\Phi:\widehat{\pi_{1}(M)}\to\widehat{\pi_{1}(N)} sends fibered classes to fibered classes where the corresponding fiber surfaces have the same topological type.

Theorem 1.1 gives many different examples of hyperbolic 3-manifolds that are distinguished from every other compact 3-manifold. One notable example is the complement of the three-component chain link in S3S^{3} (shown in Figure 1). This manifold is also called the magic manifold [GordonWu] and is denoted as 6136^{3}_{1} in [rolfsen2003knots]. This follows immediately from Theorem 1.1 and Lemma 2.6(1)[EikoKin] which proves the magic manifold is a Σ0,4\Sigma_{0,4}-bundle.

Refer to caption
Figure 1. The magic manifold 6136_{1}^{3}
Corollary 1.3.

The magic manifold has profinitely rigid fundamental group among 3-manifold groups.

Using the ideas in the proof of Theorem 1.1 we can also prove

Theorem 1.4.

Let Σ\Sigma be a finite-type surface for which Mod(Σ)Mod(\Sigma) is omnipotent and has Congruence Subgroup Property. If MϕM_{\phi} and MψM_{\psi} are finite-volume hyperbolic Σ\Sigma-bundles (monodromies ϕ\phi and ψ\psi in Mod(Σ)Mod(\Sigma)) with

π1(Mϕ)^π1(Mψ)^\widehat{\pi_{1}(M_{\phi})}\cong\widehat{\pi_{1}(M_{\psi})}

then MϕM_{\phi} and MψM_{\psi} have a common finite-sheeted cyclic cover and the same volume.

Combining Theorem 1.4 with Theorem 8 [BHMS], one can also prove

Theorem 1.5.

Let Σ\Sigma be a finite-type surface for which Mod(Σ)Mod(\Sigma) has the Congruence Subgroup Property. Let MϕM_{\phi} and MψM_{\psi} be finite-volume hyperbolic Σ\Sigma-bundles (monodromies ϕ\phi and ψ\psi in Mod(Σ)Mod(\Sigma)) with

π1(Mϕ)^π1(Mψ)^\widehat{\pi_{1}(M_{\phi})}\cong\widehat{\pi_{1}(M_{\psi})}

If all hyperbolic groups are residually finite, then MϕM_{\phi} and MψM_{\psi} have a common finite-sheeted cyclic cover and the same volume.

Acknowledgements.

The author thanks Ian Agol, Autumn Kent, Chris Leininger, Rylee Lyman, Ben McReynolds, Mark Pengitore, and Zhiyi Zhang for helpful conversations about this project. The author thanks Alan Reid for helpful conversations and comments on earlier drafts of this paper. The author is also grateful for corrections to previous versions of this preprint from Martin Bridson, Biao Ma, and Ryan Spitler.

2. Preliminaries

Let Σ\Sigma be a finite-type surface. Given an orientation-preserving homeomorphism f:ΣΣf:\Sigma\to\Sigma, the mapping torus of this homomorphism, MfM_{f} is a 3-manifold (Σ×[0,1])/(x,0)(f(x),1)(\Sigma\times[0,1])/(x,0)\sim(f(x),1). By projecting to the second factor of the product, the 3-manifold MfM_{f} is a fibration with base space S1S^{1} and fiber Σ\Sigma. The mapping torus admits a hyperbolic metric exactly when ff is pseudo-Anosov (up to isotopy) [ThurstonFiber]. At the level of fundamental group, π1(Mf)\pi_{1}(M_{f}) is a semidirect product π1(Σ)f\pi_{1}(\Sigma)\rtimes_{f_{*}}{\mathbb{Z}}, where f:π1(Σ)π1(Σ)f_{*}:\pi_{1}(\Sigma)\to\pi_{1}(\Sigma) is the automorphism induced on fundamental group by ff. It is a theorem of Stallings [StallingsFiber] that whenever π1(M)A\pi_{1}(M)\cong A\rtimes{\mathbb{Z}} for AA a finitely generated normal subgroup, then MM fibers over the circle.

The Thurston norm [Thurston1986ANF] on the homology of an orientable irreducible 3-manifold MM with empty or torus boundary is a seminorm |||\cdot| on H2(M,M;)H_{2}(M,\partial M;{\mathbb{R}}). For βH2(M,M;)\beta\in H_{2}(M,\partial M;{\mathbb{Z}})

|β|=min{χ(S)|[S]=β}|\beta|=\min\{-\chi(S)\,|\,[S]=\beta\}

where SS is a system of surfaces S=S1SnS=S_{1}\cup\dots\cup S_{n}in MM representing the class β\beta, and χ(S)=i=1nmax{0,χ(Si)}-\chi(S)=\sum_{i=1}^{n}\max\{0,-\chi(S_{i})\} where χ(Si)\chi(S_{i}) is the Euler characteristic of SiS_{i}. By Poincaré duality, there is a seminorm on H1(M;)H^{1}(M;{\mathbb{R}}), and any (integral) cohomology class that represents a fibration of MM lies in the cone over a top dimensional face of the unit norm ball called a fibered face of MM. When MM is hyperbolic, the Thurston seminorm is a norm (Theorem 1 [Thurston1986ANF]).

We now consider the profinite completion π1(Mf)^\widehat{\pi_{1}(M_{f})} of the fundamental group of a fibered 3-manifold π1(Mf)^\widehat{\pi_{1}(M_{f})}. One can check that the subspace topology induced on the fiber subgroup π1(Σ)\pi_{1}(\Sigma) coincides with the profinite topology on this subgroup (Lemma 2.2 [BR]), and so there is an exact sequence

1π1(Σ)^π1(Mf)^^11\to\widehat{\pi_{1}(\Sigma)}\to\widehat{\pi_{1}(M_{f})}\to\hat{{\mathbb{Z}}}\to 1

Moreover, Jaikin-Zapirain [JZ] showed that when MM is a compact 3-manifold with π1(M)^π1(N)^\widehat{\pi_{1}(M)}\cong\widehat{\pi_{1}(N)} for NN a fibered 3-manifold, then MM is fibered.

3. Liu’s Theorems

When MM is a finite-volume hyperbolic 3-manifold, Liu [Y] shows that there at most finitely many hyperbolic manifolds N1,,NkN_{1},\dots,N_{k} with π1(Ni)^π1(M)^\widehat{\pi_{1}(N_{i})}\cong\widehat{\pi_{1}(M)}. To show this, Liu proves the following theorems that are crucial for this work

Theorem 3.1 ([Y], Theorem 1.2, Theorem 1.3).

Let M,NM,N be finite volume hyperbolic 3-manifolds and let Φ:π1(M)^π1(N)^\Phi:\widehat{\pi_{1}(M)}\to\widehat{\pi_{1}(N)} be an isomorphism between the profinite completions of their fundamental groups. The isomorphism Φ\Phi induces an isomorphism Φ:H^1(M;)H^1(N;)\Phi_{*}:\widehat{H}_{1}(M;{\mathbb{Z}})\to\widehat{H}_{1}(N;{\mathbb{Z}}) with Φ=h^μ\Phi_{*}=\hat{h}\circ\mu where h:H1(M;)H1(N;)h:H_{1}(M;{\mathbb{Z}})\to H_{1}(N;{\mathbb{Z}}) and μ\mu denotes the scalar multiplication by a unit μ^×\mu\in\hat{{\mathbb{Z}}}^{\times}. The dual homomorphism h:H1(N;)H1(M;)h^{*}:H^{1}(N;{\mathbb{Z}})\to H^{1}(M;{\mathbb{Z}}) is Thurston-norm preserving and sends fibered classes of NN to fibered classes of MM.

Theorem 3.2 ([Y], Corollary 6.2, Corollary 6.3).

Let M,NM,N be finite volume hyperbolic 3-manifolds and let Φ:π1(M)^π1(N)^\Phi:\widehat{\pi_{1}(M)}\to\widehat{\pi_{1}(N)} be an isomorphism between the profinite completions of their fundamental groups. For any connected fiber surface ΣM\Sigma_{M} for MM, there is a connected fiber surface ΣN\Sigma_{N} for NN such that Φ(π1(ΣM)^)π1(ΣN)^\Phi(\widehat{\pi_{1}(\Sigma_{M})})\cong\widehat{\pi_{1}(\Sigma_{N})}. Furthermore, χ(ΣM)=χ(ΣN)\chi(\Sigma_{M})=\chi(\Sigma_{N}).

4. Detecting topological type of fibers using profinite completions

This section aims to prove Theorem 1.2. To begin we analyze surface automorphisms and their induced actions on characteristic quotients of the surface groups. Then let Σ=Σg,n\Sigma=\Sigma_{g,n} be a finite-type genus gg surface with nn punctures and f:π1(Σ)π1(Σ)f:\pi_{1}(\Sigma)\to\pi_{1}(\Sigma) an automorphism. Assume ff is induced by a pseudo-Anosov. We can replace ff with fmf^{m} for some integer mm to obtain a pseudo-Anosov that fixes each puncture of Σ\Sigma (that is, fixes the conjugacy classes in π1(Σ)\pi_{1}(\Sigma) of peripheral loops around each puncture of Σ\Sigma). We define the following numbers:

nf\displaystyle n_{f} =number of fixed primitive conjugacy classes of f in π1(Σ)\displaystyle=\text{number of fixed primitive conjugacy classes of }f\text{ in }\pi_{1}(\Sigma)
n^f\displaystyle\hat{n}_{f} =number of fixed primitive conjugacy classes of f^ in π1(Σ)^\displaystyle=\text{number of fixed primitive conjugacy classes of }\hat{f}\text{ in }\widehat{\pi_{1}(\Sigma)}
Proposition 4.1.

In the setup above, for a pseudo-Anosov ff, n=nf=n^fn=n_{f}=\hat{n}_{f}.

Proof.

Since a pseudo-Anosov is not reducible by definition, n=nfn=n_{f}. Upon completion, each fixed conjugacy class of ff in π1(S)\pi_{1}(S) will give a fixed conjugacy class of f^\hat{f} in π1(Σ)^\widehat{\pi_{1}(\Sigma)}, hence nfn^fn_{f}\leq\hat{n}_{f}. To see that n^fnf\hat{n}_{f}\leq n_{f}, we observe that for any element α\alpha of a fixed conjugacy class of f^\hat{f} in π1(Σ)^\widehat{\pi_{1}(\Sigma)}, we can choose tπ1(Mf)^t\in\widehat{\pi_{1}(M_{f})}, such that tt and α\alpha commute. The element tt can be chosen to be a pre-image of a unit ξ^\xi\in\hat{{\mathbb{Z}}} (which is a topological generator for ^\hat{{\mathbb{Z}}}) under the fixed epimorphism π1(Mf)^^\widehat{\pi_{1}(M_{f})}\to\hat{{\mathbb{Z}}}). Let H=α,t¯H=\overline{\langle\alpha,t\rangle} be the closed abelian group generated by α\alpha and tt.

We argue that this subgroup HH is not procyclic. First, a single topological generator ss for this group cannot live in the kernel π1(Σ)^\widehat{\pi_{1}(\Sigma)} of the epimorphism π1(Mf)^^\widehat{\pi_{1}(M_{f})}\twoheadrightarrow\hat{{\mathbb{Z}}} because tt maps non-trivially under this epimorphism. If, on the other hand, ss maps non-trivially to ^\hat{{\mathbb{Z}}} under the fixed epimorphism above, the subgroup topologically generated by ss also maps non-trivially to ^\hat{{\mathbb{Z}}}, and α\alpha cannot map non-trivially to ^\hat{{\mathbb{Z}}} under this epimorphism. Thus, Hπ1(Mf)^H\leq\widehat{\pi_{1}(M_{f})} is not procyclic. Since HH is abelian and not procyclic, HH is not a closed subgroup of a free profinite group and therefore, HH is not projective by Lemma 7.6.3 [RZ]. By Theorem 9.3 [WZ1], HH is conjugate into the closure of a cusp subgroup P<π1(Mf)P<\pi_{1}(M_{f}). The intersection P¯π1(Σ)^\overline{P}\cap\widehat{\pi_{1}(\Sigma)} is generated by a peripheral element βπ1(Σ)\beta\in\pi_{1}(\Sigma). Since α=Hπ1(Σ)^<P¯π1(Σ)^\langle\alpha\rangle=H\cap\widehat{\pi_{1}(\Sigma)}<\overline{P}\cap\widehat{\pi_{1}(\Sigma)}, α\alpha is in the closure of a peripheral element. Thus, n^fnf\hat{n}_{f}\leq n_{f}. ∎

Corollary 4.2.

For SS a finite-type surface, a homomorphism ϕ:π1(S)π1(S)\phi_{*}:\pi_{1}(S)\to\pi_{1}(S) is induced by a pseudo-Anosov ϕ:SS\phi:S\to S if and only if for all nn\in{\mathbb{Z}} the completed homomorphisms ϕn^:π1(S)^π1(S)^\widehat{\phi_{*}^{n}}:\widehat{\pi_{1}(S)}\to\widehat{\pi_{1}(S)} do not fix a non-peripheral conjugacy class.

Proof.

The forward direction is Proposition 4.1. For the reverse direction observe that ϕ\phi is not periodic, and ϕ\phi is not reducible. ∎

Proof of Theorem 1.2.

When MM (and NN) are closed, Theorem 3.1 shows that Theorem 1.2 holds. Thus, we assume that MM is cusped and fix a fibration of MM over S1S^{1} with fiber a punctured surface ΣM\Sigma_{M} and monodromy φ\varphi. By Theorem 3.2, for a fixed isomorphism Φ:π1(M)^π1(N)^\Phi:\widehat{\pi_{1}(M)}\to\widehat{\pi_{1}(N)}, there is a punctured surface ΣN\Sigma_{N} (with the same complexity as ΣM\Sigma_{M}) which is the fiber of a fibration of NN over S1S^{1} (with monodromy ψ\psi) and Φ(π1(ΣM))^π1(ΣN)^\Phi(\widehat{\pi_{1}(\Sigma_{M}))}\cong\widehat{\pi_{1}(\Sigma_{N})}. The goal is to show that ΣM\Sigma_{M} is homeomorphic to ΣN\Sigma_{N}. The surface ΣM\Sigma_{M} is a genus gMg_{M} surface with nMn_{M} punctures and the surface ΣN\Sigma_{N} is a genus gNg_{N} surface with nNn_{N} punctures. Since χ(ΣM)=χ(ΣN)\chi(\Sigma_{M})=\chi(\Sigma_{N}) by Theorem 3.2, it is sufficient to prove that nN=nMn_{N}=n_{M}. To establish this we argue as follows:

By Theorem 3.1 and Theorem 3.2, we get the following diagram

1{1}π1(ΣM){\pi_{1}(\Sigma_{M})}π1(M){\pi_{1}(M)}{{\mathbb{Z}}}1{1}1{1}π1(ΣM)^{\widehat{\pi_{1}(\Sigma_{M})}}π1(M)^{\widehat{\pi_{1}(M)}}^{\widehat{{\mathbb{Z}}}}1{1}1{1}π1(ΣN)^{\widehat{\pi_{1}(\Sigma_{N})}}π1(N)^{\widehat{\pi_{1}(N)}}^{\widehat{{\mathbb{Z}}}}1{1}1{1}π1(ΣN){\pi_{1}(\Sigma_{N})}π1(N){\pi_{1}(N)}{{\mathbb{Z}}}1{1}Φ|π1(ΣM)^\scriptstyle{\Phi|_{\widehat{\pi_{1}(\Sigma_{M})}}}Φ\scriptstyle{\Phi}×μ\scriptstyle{\times\mu}

where the ×μ:^^\times\mu:\hat{{\mathbb{Z}}}\to\hat{{\mathbb{Z}}} map is multiplication by a profinite unit μ\mu in ^\hat{{\mathbb{Z}}}. Replace φ\varphi and ψ\psi (as needed) with an appropriate power (say k!k! for a large enough natural number kk) such that both φk!\varphi^{k!} and ψk!\psi^{k!} are pure mapping classes.

We first observe that φ^k!\hat{\varphi}^{k!} and ψ^k!\hat{\psi}^{k!} have an equal number of fixed conjugacy classes in π1(ΣM)^π1(ΣN)^\widehat{\pi_{1}(\Sigma_{M})}\cong\widehat{\pi_{1}(\Sigma_{N})} i.e. n^φk!=n^ψk!\hat{n}_{\varphi^{k!}}=\hat{n}_{\psi^{k!}}. To see this, set tπ1(M)^t\in\widehat{\pi_{1}(M)} to be in the preimage of a unit in ^\hat{{\mathbb{Z}}}, the conjugation action of tt on π1(ΣM)^\widehat{\pi_{1}(\Sigma_{M})} is φ^\hat{\varphi}. The conjugation action of Φ(t)\Phi(t) on π1(ΣN)^\widehat{\pi_{1}(\Sigma_{N})} is ψ^\hat{\psi} by the diagram above. The conjugation action on π1(ΣM)^\widehat{\pi_{1}(\Sigma_{M})} by tk!t^{k!} is φ^k!\hat{\varphi}^{k!} and the conjugation action on π1(ΣN)^\widehat{\pi_{1}(\Sigma_{N})} by Φ(tk!)\Phi(t^{k!}) is ψ^k!\hat{\psi}^{k!}. For every conjugacy class α\alpha of elements in π1(ΣM)^\widehat{\pi_{1}(\Sigma_{M})} fixed by tk!t^{k!}-conjugation, there is a conjugacy class Φ(α)\Phi(\alpha) of elements in π1(ΣN)^\widehat{\pi_{1}(\Sigma_{N})} fixed by Φ(tk!)\Phi(t^{k!})-conjugation, and vice versa. Thus, n^φk!=n^ψk!\hat{n}_{\varphi^{k!}}=\hat{n}_{\psi^{k!}}. By Proposition 4.1, nM=nφk!=n^φk!n_{M}=n_{\varphi^{k!}}=\hat{n}_{\varphi^{k!}} and nN=nψk!=n^ψk!n_{N}=n_{\psi^{k!}}=\hat{n}_{\psi^{k!}}. Thus, nM=nNn_{M}=n_{N}, gM=gNg_{M}=g_{N}, and ΣM\Sigma_{M} is homeomorphic to ΣN\Sigma_{N} as claimed. ∎

Remark 4.3 (M. Bridson).

The hypothesis of hyperbolicity is necessary for the proof of Theorem 1.2. Without hyperbolicity, for instance, every mapping class ϕMod(Σ0,3)\phi\in Mod(\Sigma_{0,3}) gives a group F2ϕF_{2}\rtimes_{\phi}{\mathbb{Z}} which is also the fundamental group of a Σ1,1\Sigma_{1,1}-bundle over S1S^{1}.

5. Distinguishing hyperbolic Σ0,4\Sigma_{0,4} bundles over S1S^{1}

Recall that the Mapping Class Group of a finite type surface Σ\Sigma, here denoted as Mod(Σ)Mod(\Sigma), is the group of orientation-preserving homeomorphisms of Σ\Sigma up to isotopy.

Definition 5.1.

The group Mod(Σ)Mod(\Sigma) is omnipotent if for every independent family of (infinite-order) elements (i.e. having pairwise non-conjugate powers) γ1,,γnMod(Σ)\gamma_{1},\dots,\gamma_{n}\in Mod(\Sigma), there is a positive integer κ\kappa such that for any nn-tuple of positive integers (e1,,en)(e_{1},\dots,e_{n}) there is a finite quotient f:Mod(Σ)Qf:Mod(\Sigma)\twoheadrightarrow Q such that the order of f(γi)f(\gamma_{i}) is κei\kappa e_{i} for all 1in1\leq i\leq n.

Definition 5.2.

A principal congruence quotient of Mod(Σ)Mod(\Sigma) is the image of Mod(Σ)Mod(\Sigma) under the canonical map

Mod(Σ)Out(π1(Σ))Out(π1(Σ)/K)Mod(\Sigma)\to Out(\pi_{1}(\Sigma))\to Out(\pi_{1}(\Sigma)/K)

where K<π1(Σ)K<\pi_{1}(\Sigma) is a characteristic subgroup of finite index. A congruence quotient

Mod(Σ)QMod(\Sigma)\to Q

is a finite quotient that factors through a principal congruence quotient of Mod(Σ)Mod(\Sigma).

Definition 5.3.

The group Mod(Σ)Mod(\Sigma) has the Congruence Subgroup Property when every finite quotient of Mod(Σ)Mod(\Sigma) is a congruence quotient.

Theorem 5.4.

The group Mod(Σ0,4)Mod(\Sigma_{0,4}) is omnipotent and has the Congruence Subgroup Property.

Proof.

The group Mod(Σ0,4)PSL(2,)(/2)2Mod(\Sigma_{0,4})\cong PSL(2,{\mathbb{Z}})\ltimes({\mathbb{Z}}/2{\mathbb{Z}})^{2} (Proposition 2.7 [farb2011primer]), and this is a virtually free group. Virtually free groups are omnipotent by a theorem of Bridson-Wilton (Theorem 4.3 [BW]). That Mod(Σ0,4)Mod(\Sigma_{0,4}) has congruence subgroup property is a theorem of Diaz-Donagi-Harbater [CSPDDH]. ∎

The following lemma makes use of the proofs of Theorem 2.4 and Lemma 2.5 [BRW].

Lemma 5.5.

For hyperbolic 3-manifolds MM and NN that fiber over S1S^{1} with Σ0,4\Sigma_{0,4} fiber and monodromies φ\varphi and ψ\psi respectively, let Φ:π1(M)^π1(N)^\Phi:\widehat{\pi_{1}(M)}\to\widehat{\pi_{1}(N)} be an isomorphism that identifies the closures of the fiber subgroups of MM and NN. The monodromies φ\varphi and ψ\psi are conjugate elements of Mod(Σ0,4)Mod(\Sigma_{0,4}), and therefore MM and NN are homeomorphic.

We complete the proof of Theorem 1.1 assuming Lemma 5.5.

Proof of Theorem 1.1.

Let Σ0,4MS1\Sigma_{0,4}\hookrightarrow M\to S^{1} be a fixed fibration. By [WZ1], [JZ], and Theorem 3.2, a 3-manifold NN with π1(N)^π1(M)^\widehat{\pi_{1}(N)}\cong\widehat{\pi_{1}(M)} will be hyperbolic and will fiber over S1S^{1} with fiber a surface SS with Thurston norm 2. Fix an identification Φ:π1(M)^π1(N)^\Phi:\widehat{\pi_{1}(M)}\to\widehat{\pi_{1}(N)}. By Theorem 3.2, there is a connected fiber surface ΣN\Sigma\hookrightarrow N with Φ(π1(Σ0,4))^π1(Σ)^\Phi(\widehat{\pi_{1}(\Sigma_{0,4}))}\cong\widehat{\pi_{1}(\Sigma)}. By Theorem 1.2, Σ\Sigma is also homeomorphic to Σ0,4\Sigma_{0,4}, and therefore by Lemma 5.5 MNM\cong N. ∎

We now prove Lemma 5.5.

Proof of Lemma 5.5.

We denote the monodromies of MM and NN by φ\varphi and ψ\psi respectively, and we will refer to the outer automorphisms of F3F_{3} (the free group on three generators) induced by φ,ψ\varphi,\psi by the same names. Let Σ,Σ\Sigma,\Sigma^{\prime} be the corresponding aligned fiber surfaces (both homeomorphic to Σ0,4\Sigma_{0,4}) for MM and NN respectively. We have the following exact sequences

1{1}π1(Σ){\pi_{1}(\Sigma)}π1(M){\pi_{1}(M)}{{\mathbb{Z}}}1{1}1{1}π1(Σ)^{\widehat{\pi_{1}(\Sigma)}}π1(M)^{\widehat{\pi_{1}(M)}}^{\widehat{{\mathbb{Z}}}}1{1}1{1}π1(Σ)^{\widehat{\pi_{1}(\Sigma^{\prime})}}π1(N)^{\widehat{\pi_{1}(N)}}^{\widehat{{\mathbb{Z}}}}1{1}1{1}π1(Σ){\pi_{1}(\Sigma^{\prime})}π1(N){\pi_{1}(N)}{{\mathbb{Z}}}1{1}Φ|π1(Σ)^\scriptstyle{\Phi|_{\widehat{\pi_{1}(\Sigma)}}}Φ\scriptstyle{\Phi}\scriptstyle{\cong}

where all the unlabelled vertical arrows are canonical inclusions of groups into their profinite completions.

The isomorphism Φ\Phi embeds π1(M)\pi_{1}(M) and π1(Σ)\pi_{1}(\Sigma) in π1(N)^\widehat{\pi_{1}(N)} as dense subgroups of π1(N)^\widehat{\pi_{1}(N)} and π1(Σ)^\widehat{\pi_{1}(\Sigma^{\prime})} respectively. We consider Φ(π1(Σ))<Φ(π1(M))\Phi(\pi_{1}(\Sigma))<\Phi(\pi_{1}(M)). Let KiK_{i} be the intersection of all subgroups of Φ(π1(Σ))\Phi(\pi_{1}(\Sigma)) of index i\leq i. The tower {Ki|i}\{K_{i}\,|\,i\in\mathbb{N}\} is a cofinal tower of characteristic subgroups of Φ(π1(Σ))\Phi(\pi_{1}(\Sigma)). Set Li=Ki^π1(Σ)L_{i}=\widehat{K_{i}}\cap\pi_{1}(\Sigma^{\prime}) to obtain a corresponding characteristic tower of π1(Σ)\pi_{1}(\Sigma^{\prime}). Conjugation by elements of Φ(π1(M))\Phi(\pi_{1}(M)) on Φ(π1(Σ))\Phi(\pi_{1}(\Sigma)) induces outer automorphisms in Out(Φ(π1(Σ))/Ki)Out(π1(Σ)^/Li^)Out(\Phi(\pi_{1}(\Sigma))/K_{i})\cong Out(\widehat{\pi_{1}(\Sigma^{\prime})}/\widehat{L_{i}}) for all ii. Similarly, conjugation by elements of π1(N))\pi_{1}(N)) on π1(Σ)\pi_{1}(\Sigma^{\prime}) induces outer automorphisms in Out(π1(Σ)/Li)Out(π1(Σ)^/Li^)Out(\pi_{1}(\Sigma^{\prime})/L_{i})\cong Out(\widehat{\pi_{1}(\Sigma^{\prime})}/\widehat{L_{i}}) for all ii.

The monodromy φ\varphi is induced for the extension

1Φ(π1(Σ))Φ(π1(M))11\to\Phi(\pi_{1}(\Sigma))\to\Phi(\pi_{1}(M))\to{\mathbb{Z}}\to 1

by the conjugation action of an element tΦ(π1(M))t\in\Phi(\pi_{1}(M)) where the rightmost {\mathbb{Z}} is a dense subgroup of ^\hat{{\mathbb{Z}}} (and tt maps to a generator of this ^\hat{{\mathbb{Z}}} under the fixed profinite epimorphism π1(N)^^\widehat{\pi_{1}(N)}\twoheadrightarrow\hat{{\mathbb{Z}}}). Likewise, the monodromy ψ\psi is induced by the conjugation action of an element sπ1(N)<π1(N)^s\in\pi_{1}(N)<\widehat{\pi_{1}(N)} that maps to a generator of {\mathbb{Z}}. Since the images of tt and ss under the fixed π1(N)^^\widehat{\pi_{1}(N)}\twoheadrightarrow\hat{{\mathbb{Z}}} both topologically generate ^\hat{{\mathbb{Z}}}, the actions of tt and ss (and therefore the actions of φ\varphi and ψ\psi) induce outer automorphisms φiOut(Φ(π1(Σ))/Ki)Out(π1(Σ)^/Li^)\varphi_{i}\in Out(\Phi(\pi_{1}(\Sigma))/K_{i})\cong Out(\widehat{\pi_{1}(\Sigma^{\prime})}/\widehat{L_{i}}) and ψiOut(π1(Σ)/Li)Out(π1(Σ)^/Li^)\psi_{i}\in Out(\pi_{1}(\Sigma^{\prime})/L_{i})\cong Out(\widehat{\pi_{1}(\Sigma^{\prime})}/\widehat{L_{i}}) such that φi\varphi_{i} and ψi\psi_{i} generate conjugate cyclic subgroups Ci<Out(π1(Σ)^/Li^)C_{i}<Out(\widehat{\pi_{1}(\Sigma^{\prime})}/\widehat{L_{i}}).

Assume, for sake of contradiction, that φ\varphi and ψ\psi are independent in Mod(Σ0,4)Mod(\Sigma_{0,4}), i.e. that φ\varphi and ψ\psi do not have conjugate powers. Since Mod(Σ0,4)Mod(\Sigma_{0,4}) is omnipotent there is a finite quotient q:Mod(Σ0,4)Qq:Mod(\Sigma_{0,4})\twoheadrightarrow Q for which o(q(φ))o(q(ψ))o(q(\varphi))\neq o(q(\psi)). Because Mod(Σ0,4)Mod(\Sigma_{0,4}) has the Congruence Subgroup Property, qq factors through a principal congruence homomorphism

qi:Mod(Σ0,4)Out(π1(Σ)/Ki)q_{i}:Mod(\Sigma_{0,4})\to Out(\pi_{1}(\Sigma)/K_{i})

for some ii\in\mathbb{N}. By construction, φi=qi(φ)\varphi_{i}=q_{i}(\varphi) and ψi=qi(ψ)\psi_{i}=q_{i}(\psi). By the previous paragraph, φi\varphi_{i} and ψi\psi_{i} generate conjugate cyclic subgroups of Out(Φ(π1(Σ)/Ki)Out(\Phi(\pi_{1}(\Sigma)/K_{i}), and this contradicts the claim that o(q(φ))o(q(ψ))o(q(\varphi))\neq o(q(\psi)). Thus, φ\varphi and ψ\psi are not independent. In particular, there are positive integers m,nm,n such that φm\varphi^{m} and ψ±n\psi^{\pm n} are conjugate.

Following [BRW], apply omnipotence and CSP to the cyclic subgroup of Mod(Σ0,4)Mod(\Sigma_{0,4}) generated by φ\varphi to show that there is a congruence quotient q:Mod(Σ0,4)Out(π1(Σ)/Ki)q:Mod(\Sigma_{0,4})\to Out(\pi_{1}(\Sigma)/{K_{i}}) for which mnmn divides o(q(φ))=o(q(ψ))o(q(\varphi))=o(q(\psi)). Since

o(q(φ))m=o(q(φm))=o(q(ψn))=o(q(ψ))n\frac{o(q(\varphi))}{m}=o(q(\varphi^{m}))=o(q(\psi^{n}))=\frac{o(q(\psi))}{n}

it follows that m=nm=n. Since φ,ψ\varphi,\psi are pseudo-Anosovs in Mod(Σ0,4)Mod(\Sigma_{0,4}) with a common conjugate power, there is a mapping class ξ\xi in Mod(Σ0,4)Mod(\Sigma_{0,4}) with (ξφξ1)m=ξφmξ1=ψm(\xi\varphi\xi^{-1})^{m}=\xi\varphi^{m}\xi^{-1}=\psi^{m}. Since Mod(Σ0,4)Mod(\Sigma_{0,4}) is a (non-free) virtually free group, it splits as a finite graph of groups with finite vertex and edge groups (Theorem 1 [karrass_pietrowski_solitar_1973]) and so Mod(Σ0,4)Mod(\Sigma_{0,4}) acts on a simplicial tree TT. The elements ξφξ1\xi\varphi\xi^{-1} and ψ\psi fix a common bi-infinite geodesic α\alpha in TT. The stabilizer of α\alpha is cyclic (since all torsion in Mod(Σ0,4)Mod(\Sigma_{0,4}) fixes a vertex in TT) and so ξφξ1=ψ±\xi\varphi\xi^{-1}=\psi^{\pm}.

6. Theorem 1.4 and further observations

This program for proving Theorem 1.1 works to show that when Mod(Σ)Mod(\Sigma) is omnipotent and has the Congruence Subgroup Property, hyperbolic Σ\Sigma-bundles over S1S^{1} whose fundamental groups have isomorphic profinite completions are cyclically commensurable. For the proof of Lemma 5.5, it was crucial that two pseudo-Anosovs in Mod(Σ0,4)Mod(\Sigma_{0,4}) have a common power (up to conjugacy) if and only if the two pseudo-Anosovs are conjugate. In general, however, this is not true (see [rootsModS] for example).

Proof of Theorem 1.4.

Let Σ\Sigma be a finite-type surface with negative Euler characteristic for which Mod(Σ)Mod(\Sigma) is omnipotent and has the Congruence Subgroup Property. By Theorem 1.2 and [WZ1], any 3-manifold NN with the same profinite completion as a hyperbolic Σ\Sigma-bundle MM over S1S^{1} is a hyperbolic Σ\Sigma-bundle over S1S^{1}. An analog of Lemma 5.5 shows that the corresponding monodromies have a common conjugate power. This common conjugate power is the monodromy of a common finite cyclic cover of MM and NN. Thus MM and NN are cyclically commensurable. ∎

In their work [BHMS] on hierarchically hyperbolic quotients of mapping class groups, Behrstock-Hagen-Sisto-Martin prove

Theorem 6.1 (Theorem 8, [BHMS]).

Let SS be a connected orientable surface of finite type of complexity at least 2. If all hyperbolic groups are residually finite, then the following holds. Let g,hMod(S)g,h\in Mod(S) be pseudo-Anosovs with no common proper power, and let qq\in{\mathbb{Q}}. Then there exists a finite group GG and a homomorphism ϕ:Mod(S)G\phi:Mod(S)\to G such that ord(ϕ(g))/ord(ϕ(h))=qord(\phi(g))/ord(\phi(h))=q.

This theorem allows us to prove Theorem 1.5, reducing the question of commensurability of profinitely equivalent Σ\Sigma-bundles over S1S^{1} when Mod(Σ)Mod(\Sigma) has CSP to the residual finiteness of all hyperbolic groups.

Proof of Theorem 1.5.

Let Σ\Sigma be a finite-type surface with negative Euler characteristic for which has the Congruence Subgroup Property. By Theorem 1.2 and [WZ1], any 3-manifold NN with the same profinite completion as a hyperbolic Σ\Sigma-bundle MM over S1S^{1} is a hyperbolic Σ\Sigma-bundle over S1S^{1}. If all hyperbolic groups are residually finite, then Theorem 6.1 above can be combined with Lemma 5.5 to show that the corresponding monodromies will have a common conjugate power. This common conjugate power is the monodromy of a common degree finite cyclic cover of MM and NN. Thus MM and NN are cyclically commensurable manifolds with the same volume. ∎

References