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

thanks: The author acknowledges support from NSF grant RTG DMS–1745583.

Automatic Continuity of Pure Mapping Class Groups

Ryan Dickmann 686 Cherry St NW, Atlanta, GA 30332 [email protected]
Abstract.

We completely classify the orientable infinite-type surfaces SS such that PMap(S)\operatorname{PMap}(S), the pure mapping class group, has automatic continuity. This classification includes surfaces with noncompact boundary. In the case of surfaces with finitely many ends and no noncompact boundary components, we prove the mapping class group Map(S)\operatorname{Map}(S) does not have automatic continuity. We also completely classify the surfaces such that PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)}, the subgroup of the pure mapping class group composed of elements with representatives that can be approximated by compactly supported homeomorphisms, has automatic continuity. In some cases when PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} has automatic continuity, we show any homomorphism from PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} to a countable group is trivial.

keywords:
automatic continuity, polish group, topological group, mapping class group, pure mapping class group, boundary, manifold with boundary, noncompact, non-compact
2010 Mathematics Subject Classification:
57K20, 57M07, 57S05, 20F65

1. Introduction

A surface will refer to a second-countable, connected, orientable, 2-manifold, possibly with boundary. Let Homeo(S)\operatorname{Homeo}_{\partial}(S) be the group of (orientation-preserving) homeomorphisms of SS that fix the boundary pointwise. The mapping class group Map(S)\operatorname{Map}(S) is defined to be

Map(S)=Homeo(S)/\displaystyle\operatorname{Map}(S)=\operatorname{Homeo}_{\partial}(S)/\sim

where two homeomorphisms are equivalent if they are isotopic relative to the boundary of SS. A degenerate end will refer to an end with a closed neighborhood homeomorphic to a disk with boundary points removed. Throughout the paper, we assume surfaces do not have degenerate ends, since filling in degenerate ends does not change the underlying mapping class group.

A surface is said to be of infinite type when π1\pi_{1} is infinitely generated, otherwise, it is of finite type. A Polish group is a topological group that is separable and completely metrizable. In the finite-type case, mapping class groups of surfaces are finitely generated and are therefore countable with no interesting Polish group structure. Mapping class groups for infinite-type surfaces, however, are uncountable and are Polish groups when given the quotient topology inherited from the compact-open topology on Homeo(S)\operatorname{Homeo}_{\partial}(S).

Mann [7] proved that certain mapping class groups of infinite-type surfaces without boundary have automatic continuity; i.e., every homomorphism from these groups to a separable group is continuous. Mann also found examples of mapping class groups that admit discontinuous homomorphisms to a finite group and asked which mapping class groups have automatic continuity. Towards this question, we fully classify the pure mapping class groups that have automatic continuity.

Pure mapping class groups. The pure mapping class group of a surface, denoted PMap(S)\operatorname{PMap}(S), is the subgroup of the mapping class group consisting of elements that fix the ends of the surface. A disk with handles will refer to any surface that can be constructed by taking a disk, removing a closed, totally disconnected set from the boundary (whose points become the ends of the surface), and then attaching infinitely many handles accumulating to some subset of the ends. See Figure 1 for an example. The assumption of infinitely many handles is simply to rule out finite-type cases.

Theorem A.

Let SS be an infinite-type surface. Then PMap(S)\operatorname{PMap}(S) has automatic continuity if and only if

  1. (i)

    SS is a connected sum of finitely many disks with handles with any finite-type surface, and

  2. (ii)

    SS has finitely many ends accumulated by genus.

The finite-type surface is necessary in the first condition to capture additional cases with finitely many compact boundary components and finitely many punctures. The final condition is required since, for surfaces with infinitely many ends accumulated by genus, we show there is a discontinuous homomorphism PMap(S)2\operatorname{PMap}(S)\rightarrow\mathbb{Z}_{2} which factors through ω\mathbb{Z}^{\omega}, the infinite countable product. If we equip \mathbb{Z} with the discrete topology, then ω\mathbb{Z}^{\omega} is a Polish group with the product topology. The map to ω\mathbb{Z}^{\omega} is given by the work of Aramayona–Patel–Vlamis in the compact boundary case [1], and this was extended by the author to the noncompact boundary case [5]. More precisely, their works show PMap(S)=PMapc(S)¯\operatorname{PMap}(S)=\overline{\operatorname{PMap}_{c}(S)} when SS has at most one end accumulated by genus, and otherwise PMap(S)\operatorname{PMap}(S) factors into a semidirect product of a special subgroup with n\mathbb{Z}^{n} where nn is finite if and only if there are finitely many ends accumulated by genus. We now discuss PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} further.

Refer to caption
Figure 1. A visualization of a disk with handles with two ends. The handle attaching procedure in this case joins together the bottom two ends of the disk into a single end. By Theorem A, PMap(S)\operatorname{PMap}(S) has automatic continuity for this surface. Note Map(S)=PMap(S)\operatorname{Map}(S)=\operatorname{PMap}(S) for any disk with handles since fixing the boundary forces the ends to be fixed.

Closure of the subgroup of compactly supported maps. We say fMap(S)f\in\operatorname{Map}(S) is compactly supported if ff has a representative that is the identity outside of a compact subset of SS. The subgroup consisting of compactly supported mapping classes is denoted PMapc(S)\operatorname{PMap}_{c}(S) since every compactly supported mapping class is pure. The closure of this subgroup, denoted PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)}, can be described as the subgroup composed of elements with representatives that can be approximated by compactly supported homeomorphisms. We also fully classify the PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} that have automatic continuity.

Theorem B.

Let SS be an infinite-type surface. Then PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} has automatic continuity if and only if SS is a connected sum of finitely many disks with handles with any finite-type surface.

As a consequence of Mann’s result [7] of automatic continuity for the mapping class groups of the sphere minus the Cantor set and the plane minus the Cantor set, Vlamis [12] showed that any homomorphism from these groups to a countable group is trivial. Using a similar but independent proof, we show the following.

Corollary 1.1.

Let SS be a disk with handles. Then every homomorphism from PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} to a countable group is trivial. Therefore, PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} contains no proper normal subgroups of countable index and no proper subgroups of finite index.

This is in stark contrast to the mapping class groups of finite-type surfaces which are residually finite. One natural approach to studying infinite groups is to investigate their finite quotients, but for the PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} of disks with handles, we do not even have countable quotients to work with. Note that PMap(S)\operatorname{PMap}(S) always has a proper normal subgroup of countable index when SS has at least two ends accumulated by genus. In particular, when there are finitely many ends accumulated by genus, PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} is the desired subgroup, and when there are infinitely many ends accumulated by genus, the kernel of the discontinuous homomorphism to 2\mathbb{Z}_{2} discussed above is the desired subgroup.

Mapping class groups. Using the same techniques in the proofs of the above theorems, we are also able to comment on the automatic continuity of the full mapping class groups.

Theorem 1.2.

Suppose SS is an infinite-type surface with finitely many ends and no noncompact boundary components. Then Map(S)\operatorname{Map}(S) does not have automatic continuity.

For example, the mapping class group of the ladder surface, the unique surface with no boundary and exactly two ends each accumulated by genus, does not have automatic continuity. We also extend the reverse direction of Theorem A to the full mapping class group.

Theorem 1.3.

Suppose SS is an infinite-type surface satisfying the conditions of Theorem A. Then Map(S)\operatorname{Map}(S) has automatic continuity.

Outline. In Section 2, we discuss some background on surfaces with noncompact boundary, and in Section 3, we discuss the tools needed to prove the reverse directions of Theorems A and B. In Section 4.1, we prove the reverse directions as well as Theorem 1.3 using adaptations Mann’s techniques [7] and tools from the author for working with surfaces with noncompact boundary [5]. We also use a new extension to a classical lemma of Sierpiński; see Section 3.2. In Section 4.2, we prove the forward directions of Theorems A and B, and in Section 4.3, we prove Corollary 1.1 and Theorem 1.2.

Automatic continuity proofs largely rely on some form of self-similarity in a given group, and in particular, we take advantage of the self-similarity of the mapping class groups of the sliced Loch Ness monsters. A sliced Loch Ness monster is any surface with nonempty boundary, no compact boundary components, infinite genus, and one end. See Figure 3. The key idea is that any sliced Loch Ness monster contains closed proper copies of itself, and therefore, the mapping class group does as well; see Section 2.1. On the other hand, the Loch Ness monster, the unique surface with one end, infinite genus, and empty boundary, does not contain a closed proper copy of itself,111It is unknown whether the mapping class group of the Loch Ness monster contains a proper copy of itself. A group that does not contain a proper copy of itself is known as co-Hopfian. Aramayona–Leininger–McLeay [2] have studied the co-Hopfian property for mapping class groups of infinite-type surfaces, and in particular they found uncountably many examples of pure mapping class groups that are not co-Hopfian. and Domat and the author showed its mapping class group does not have automatic continuity [6].

Once we have found examples of surfaces such that PMap(S)\operatorname{PMap}(S) and PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} have automatic continuity, the main difficulty in proving Theorems A and B is ruling out the zoo of remaining surfaces. Note these theorems consider all surfaces including those with complicated end spaces such as large countable ordinals. Using the tools developed by the author for decomposing surfaces into simpler pieces, we can reduce the complexity of the problem significantly.

Refer to caption
Figure 2. Two subsurfaces KK and KK^{\prime} in the same PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} orbit.

To find discontinuous homomorphisms in the remaining cases, we use the work of Domat [6] who showed PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} admits uncountably many discontinuous homomorphisms to \mathbb{Q} with the discrete topology when SS contains a certain infinite sequence of nondisplaceable subsurfaces. A nondisplaceable subsurface in this case refers to a surface that cannot be mapped off of itself by any map in PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)}. Domat’s proof relies on subsurface projections of Masur–Minksy [8] to construct projection complexes of Bestvina–Bromberg–Fujiwara [3]. For a given finite-type nondisplaceable subsurface KK, a projection complex is built using subsurface projections between subsurfaces in the orbit of KK under the action of PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} on the isotopy classes of subsurfaces in SS. The subsurface projection to KK is a map that takes a subsurface KK^{\prime} (distinct from KK) in the orbit of KK and outputs an element of the power set of the vertex set of the curve graph of KK. The vertex set of the curve graph is defined as the set of isotopy classes of essential simple closed curves. Recall a curve is trivial when it bounds a disk, peripheral when it bounds an annulus, and essential when it is neither trivial nor peripheral. The subsurface projection to KK is defined using the fact that KK\partial K^{\prime}\cap K is a collection of curves and arcs in KK. The arcs are turned into curves by surgering on intervals in K\partial K. The issue for surfaces with noncompact boundary is that there can exist nondisplaceable subsurfaces KK and KK^{\prime} in the same PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} orbit such that KK\partial K^{\prime}\cap K is the union of trivial arcs in KK (a trivial arc is one that bounds a disk). Trivial arcs yield trivial or peripheral curves after surgery using the boundary of KK, so the subsurface projection is not well-defined. See Figure 2 for an example. We will see that the surfaces from Theorem B are exactly those that do not have the special sequences of nondisplaceable subsurfaces needed by Domat.

Acknowledgments

The author would like to thank Kathryn Mann for introducing him to automatic continuity and the many techniques used in Section 3. Thank you to the organizers of the 2019 AIM workshop for surfaces of infinite type. Thank you to Dan Margalit, Roberta Shapiro, and Sanghoon Kwak for comments on an earlier draft. Thank you to an anonymous referee for carefully reading this paper.

2. Surfaces with Noncompact Boundary

Here we discuss some background on surfaces with noncompact boundary needed for the proofs of Theorems A and B. We will assume the reader is familiar with the Richards classification of infinite-type surfaces without boundary [9] as well as the definition of the ends space of a surface, planar ends, and ends accumulated by genus. These definitions apply without adaptation to surfaces with noncompact boundary. The first adaptation needed for noncompact boundary is that we must consider ends accumulated by compact boundary, ends for which every closed neighborhood contains infinitely many compact boundary components, then we must consider the noncompact boundary components.

Boundary chains. Deleting the noncompact boundary components of a surface induces a map π\pi from the ends space of the surface to the ends space of the interior surface (for more details see [5], Section 4.2). For example, in Figure 3 deleting the noncompact boundary components induces a map sending the two ends to the single end of the interior. Suppose ee is an end of the surface that a noncompact boundary component points to, and let eo=π(e)e^{\mathrm{o}}=\pi(e) be the corresponding end of the interior surface. Then we refer to π1(eo)\pi^{-1}(e^{\mathrm{o}}) as a boundary chain, and we refer to any end in π1(eo)\pi^{-1}(e^{\mathrm{o}}) as a boundary end. Note any disk with handles has a single boundary chain, and every end is a boundary end. Other examples of surfaces with a single boundary chain can be constructed by taking a disk, deleting a set from the boundary, and then attaching surfaces without noncompact boundary components where these attached surfaces may be infinite-type and may accumulate to the set of deleted points. An end that is not a boundary end will be called an interior end. Though the boundary chain is formally defined as a set of ends, we can also think of a boundary chain as the corresponding union of noncompact boundary components.

Brown–Messer classification of surfaces. The classification of infinite-type surfaces with boundary is due to Brown and Messer [4]. Roughly speaking, their theorem states that surfaces with boundary are classified up to homeomorphism by the Richards classification data, the ends accumulated by compact boundary, and additional data describing the boundary chains. The major achievement of Brown and Messer was finding a way to represent this boundary chain data, though it is fairly technical. Thus we will not state the actual classification theorem, and instead, we will use tools of the author developed for working with surfaces with boundary [5]. More examples of surfaces with boundary can be found in Section 3 of the previous paper of the author.

Recall a sliced Loch Ness monster is any surface with nonempty boundary, no compact boundary components, infinite genus, and one end. One immediate application of the classification of surfaces is that a sliced Loch Ness monster is determined by the number of boundary components. We will refer to an nn-sliced Loch Ness monster to emphasize the number of boundary components. Note that nn may be infinite.

Recall a disk with handles is a surface that can be constructed by taking a disk, removing a closed, totally disconnected set from the boundary, and then attaching infinitely many handles accumulating to some subset of the ends. Sliced Loch Ness monsters are examples of disks with handles since we can construct any sliced Loch Ness monster by attaching handles to a disk with points removed from the boundary in a way that joins every end to a single end.

Remark 2.1.

We could have equivalently defined a sliced Loch Ness monster as a disk with handles with exactly one end. There are other constructions of sliced Loch Ness monsters that do not start with a single disk which will be useful for the proof of Theorem A. We will discuss these in Section 2.1. See Figure 3 for some examples.

Refer to caption
Figure 3. A 1-sliced, a 2-sliced, and an \infty-sliced Loch Ness monster with the noncompact boundary components represented by the bold lines. The blue shading represents a given compact exhaustion {Ki}\{K_{i}\} for each surface.

Cutting up a surface with boundary. The following result of the author [5] shows that any surface with only ends accumulated by genus can be decomposed in some sense into sliced Loch Ness monsters and Loch Ness monsters.

Lemma 2.2.

Every disk with handles with every end accumulated by genus can be cut along a collection of disjoint essential arcs into sliced Loch Ness monsters.

Furthermore, any infinite-type surface with every end accumulated by genus can be cut along disjoint separating curves into components that are either

  1. (i)

    Loch Ness monsters with compact boundary components added, or

  2. (ii)

    disks with handles with compact boundary components added.

Cutting a surface SS along a curve or arc α\alpha yields a possibly disconnected surface with an identification map between subsets of the boundary such that the quotient is SS and the image of the identified subsets under the quotient is α\alpha. When we say compact boundary components are added, we mean open balls with disjoint closures are removed. These components may have any number of compact boundary components added, and if we add infinitely many we assume they accumulate to some end of the original surface.

Recall a planar end is simply one that is not accumulated by genus. Due to the assumption on degenerate ends in the introduction, a disk with handles automatically has no planar ends, so we can apply the first part of Lemma 2.2 to any disk with handles. Note the second part of Lemma 2.2 does not immediately extend to surfaces with planar ends since filling in a planar boundary end may not be possible; for example, if it is accumulated by compact boundary components or accumulated by boundary chains. For other decomposition results concerning general surfaces, see Section 4 of the work of the author [5].

We will need the following result to justify the forward directions of Theorems A and B.

Lemma 2.3.

An infinite-type surface with nonempty boundary, finitely many boundary chains, no compact boundary components, and no interior ends is a connected sum of disks with handles.

Proof.

By the assumption on interior ends, the only planar ends must be boundary ends. Since there are no compact boundary components and finitely many boundary chains, these planar ends must be degenerate. Since we assumed in the introduction that surfaces do not have degenerate ends, every end must be accumulated by genus and we can apply the second part of Lemma 2.2 to cut the surface along curves into disks with handles with compact boundary components added and Loch Ness monsters with compact boundary added. Note none of the components can be the second type since then there would be interior ends, so we are done. ∎

2.1. Standard pieces of sliced Loch Ness monsters

Now we discuss certain models of the sliced Loch Ness monsters and a standard way to break them into self-similar pieces. We represent the 1-sliced Loch Ness monster as a closed upper half-plane with a handle attached in a small ball about each integer point. Let MiM_{i} be the subsurface bounded by the lines x=i12x=i-\frac{1}{2} and x=i+12x=i+\frac{1}{2} for ii\in\mathbb{Z}. We refer to each MiM_{i} as a standard piece of the 1-sliced Loch Ness monster.

Refer to caption
Figure 4. Shown in the shaded regions are disjoint representatives of the standard pieces for the 2-sliced Loch Ness monster.

Let KnK_{n} denote a compact surface with zero genus and nn boundary components when nn is finite. Let KK_{\infty} denote the surface with 0 genus, no noncompact boundary components, and exactly two ends that are accumulated by compact boundary components. To construct an nn-sliced Loch Ness monster with n2n\geq 2 and possibly infinite, first start with a disjoint union of nn copies of the closed upper half-plane. Now for each interior integer point in the closed upper half-plane apply the following construction: first, remove a small open ball about the integer point in each of the nn copies, then take a copy of KnK_{n} and attach Kn\partial K_{n} to the resulting boundary components. When nn is infinite, we arrange the closed upper half-planes according to a \mathbb{Z}-index. We then equip the components of K\partial K_{\infty} with the natural \mathbb{Z}-index and apply the attaching procedure respecting the two indexes.

To construct the standard pieces in these cases, let MiM_{i} be the subsurface bounded by the nn copies of the lines x=i12x=i-\frac{1}{2} and x=i+12x=i+\frac{1}{2} for ii\in\mathbb{Z}. See Figure 4 for an example where a small open neighborhood of the boundary has been removed from each MiM_{i} to emphasize the standard pieces.

3. Tools

We now discuss the various tools needed in Section 4.1 to show that PMap(S)\operatorname{PMap}(S) and PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} have automatic continuity for certain surfaces with noncompact boundary. These groups are related by the work of Aramayona–Patel–Vlamis in the compact boundary case [1], and this was extended by the author to the noncompact boundary case [5]:

Lemma 3.1.

Let SS be an infinite-type surface.

PMap(S)=PMapc(S)¯H\operatorname{PMap}(S)=\overline{\operatorname{PMap}_{c}(S)}\rtimes H

where Hn1\displaystyle H\cong\mathbb{Z}^{n-1} when there is a finite number n>1n>1 of ends of SS accumulated by genus, Hω\displaystyle H\cong\mathbb{Z}^{\omega} when there are infinitely many ends accumulated by genus, and HH trivial otherwise.

Note this implies PMapc(S)¯=PMap(S)\overline{\operatorname{PMap}_{c}(S)}=\operatorname{PMap}(S) for a one-ended surface such as a sliced Loch Ness monster. Since PMap(S)=Map(S)\operatorname{PMap}(S)=\operatorname{Map}(S) in this case as well, we will always use the latter notation for one-ended surfaces or subsurfaces.

3.1. Automatic Continuity

The standard approach to proving automatic continuity is to prove a stronger but more tractable condition due to Rosendal–Solecki [11]. We say a subset of a group is countably syndetic if countably many left translates cover the entire group. A topological group is Steinhaus if there exists an integer kk such that, for every countably syndetic symmetric subset WW of GG, WkW^{k} contains an open neighborhood of the identity.

Proposition 3.2 (Rosendal–Solecki).

A Steinhaus Polish group has automatic continuity.

We will also need a common fact used in automatic continuity proofs. This result, as well as the result of Rosendal and Solecki, follows from the Baire category theorem.

Proposition 3.3.

Let GG be a Polish group and WGW\subset G a countably syndetic symmetric set. Then there exists a neighborhood UU of the identity in GG such that W2W^{2} is dense in UU.

In some cases, we will rule out automatic continuity by ruling out a weaker property. A topological group is said to have the small index property when any countable index subgroup is open. The following is well-known.

Proposition 3.4.

A Polish group that has automatic continuity has the small index property.

Proof.

Let SωS_{\omega} denote the symmetric group on a countably infinite set. This is a Polish group with the compact-open topology. Any countable index subgroup HH determines a homomorphism ϕ\phi to SωS_{\omega} by the left multiplication action on left cosets. The subset of SωS_{\omega} corresponding to permutations that fix HH is open. The pullback of this subset via ϕ\phi is HH, which is open by automatic continuity. ∎

3.2. Sierpiński lemmas

To apply the same techniques used by Mann [7], we need the following result of Sierpiński.

Lemma 3.5 (Sierpiński).

For an infinite countable set Λ\Lambda, there is an uncountable collection of infinite subsets {Ωα}αΓ\{\Omega_{\alpha}\}_{\alpha\in\Gamma} of Λ\Lambda such that any ΩαΩβ\Omega_{\alpha}\cap\Omega_{\beta} is finite for αβ\alpha\neq\beta.

Proof.

Identify Λ\Lambda with \mathbb{Q} via a bijection, and let Γ=\Gamma=\mathbb{R}\setminus\mathbb{Q}. For any given αΓ\alpha\in\Gamma, let Ωα\Omega_{\alpha} be any sequence of rational numbers converging to α\alpha. ∎

We also need an extension that allows us to apply the techniques to infinite unions.

Lemma 3.6.

For any product of infinite countable sets Λ×Λ\Lambda\times\Lambda^{\prime}, there is an uncountable collection of infinite subsets {Ωα}αΓ\{\Omega_{\alpha}\}_{\alpha\in\Gamma} of Λ×Λ\Lambda\times\Lambda^{\prime} such that

  1. (i)

    ΩαΩβ\Omega_{\alpha}\cap\Omega_{\beta} is finite for all αβ\alpha\neq\beta.

  2. (ii)

    Ωα({λ}×Λ)\Omega_{\alpha}\cap(\{\lambda\}\times\Lambda^{\prime}) is infinite for all λΛ\lambda\in\Lambda.

Proof.

Let γ1,γ2,\gamma_{1},\gamma_{2},... be an infinite sequence of irrationals independent over \mathbb{Q}. Now {γn}n=1\{\gamma_{n}\mathbb{Q}\}_{n=1}^{\infty} is a collection of pairwise disjoint dense subsets of the reals. Identify Λ\Lambda via a bijection with \mathbb{N}, and then identify each {n}×Λ\{n\}\times\Lambda^{\prime} by a bijection with γn\gamma_{n}\mathbb{Q}. Let Γ=(n=1γn)\Gamma=\mathbb{R}\setminus(\bigcup_{n=1}^{\infty}\gamma_{n}\mathbb{Q}). For any given αΓ\alpha\in\Gamma, choose any sequence of numbers in n=1γn\bigcup_{n=1}^{\infty}\gamma_{n}\mathbb{Q} converging to α\alpha that includes infinitely many entries of γn\gamma_{n}\mathbb{Q} for all nn, and then let Ωα\Omega_{\alpha} be the corresponding set of tuples (n,γnqn)(n,\gamma_{n}q_{n}) where qnq_{n}\in\mathbb{Q}. ∎

3.3. Tools of Mann

Using the previously discussed tools, we will now introduce the main lemmas for our automatic continuity proofs, Lemmas 3.9 and 3.10. The proofs of these lemmas follow an argument from Section 4 of the paper of Mann [7], and rely on a modified version of Lemma 3.2 from the same paper. The original lemma was applied to the homeomorphism group of a manifold, but the proof can be adapted to work in the mapping class group setting.

We say a collection of disjoint subsurfaces {Si}\{S_{i}\} of SS is admissible when any product ifi\prod_{i}f_{i} is a well-defined homeomorphism of SS for fif_{i} supported in SiS_{i}. This condition is required in the statement of the following lemma since Mann’s proof uses infinite products of homeomorphisms, and these may not always be well-defined. For example, consider a sequence of disjoint essential annuli which all essentially intersect some compact subsurface. An infinite product of Dehn twists about these annuli is not a well-defined mapping class of the surface since, even after isotopy, the annuli accumulate at some point of the compact subsurface, and there is no continuous extension of the infinite twist to this point.

Lemma 3.7 (Mann).

Let SS be an infinite-type surface, and WPMapc(S)¯W\subset\overline{\operatorname{PMap}_{c}(S)} a countably syndetic symmetric set. Let 𝒜\mathcal{A} be an infinite admissible collection of disjoint closed subsurfaces of SS satisfying:

  1. (1)

    There exists an infinite admissible collection of disjoint subsurfaces UiU_{i} of SS such that each UiU_{i} contains an infinite family of disjoint subsurfaces belonging to 𝒜\mathcal{A}.

  2. (2)

    There exists pp\in\mathbb{N} such that, for each A𝒜A\in\mathcal{A}, the subgroup of PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} consisting of maps with support in AA, denoted PMap(A)\operatorname{PMap}(A), has commutator length bounded by pp.

Then there exists A𝒜A\in\mathcal{A} such that PMap(A)\operatorname{PMap}(A) is contained in W8pW^{8p}.

We will also use the following result of the author [5].

Lemma 3.8.

Let SS be a disk with handles. Then every element in PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} can be written as the product of two commutators.

Lemma 3.9.

Let SS be a surface containing a subsurface MM homeomorphic to any sliced Loch Ness monster, {Mn}\{M_{n}\} the standard pieces of this sliced Loch Ness monster, and WPMapc(S)¯W\subset\overline{\operatorname{PMap}_{c}(S)} a countably syndetic symmetric set. Then there is a finite set FF such that

nFMap(Mn)W80\displaystyle\prod_{n\notin F}\operatorname{Map}(M_{n})\subset W^{80}
Proof.

Apply Lemma 3.7 with 𝒜\mathcal{A} consisting of subsurfaces of the form AΛ=nΛM2nA_{\Lambda}=\cup_{n\in\Lambda}M_{2n} for some infinite set Λ\Lambda\subset\mathbb{Z}. Any collection of the MiM_{i} is admissible since we can extend a homeomorphism on the collection via the identity map to SS. Note 𝒜\mathcal{A} satisfies the hypotheses of the lemma, since

  1. (i)

    We can write \mathbb{Z} as a countable disjoint union of infinite sets Λi\Lambda_{i}, and define UiU_{i} to be nΛiM2n\bigcup_{n\in\Lambda_{i}}M_{2n}. Each such set contains a countable union of disjoint elements of 𝒜\mathcal{A}.

  2. (ii)

    Lemma 3.8 implies the same statement for the mapping class group of a disjoint union of sliced Loch Ness monsters. Thus, any element supported in AΛA_{\Lambda} may be written as the product of two commutators.

We conclude that for some such subsurface AΛ𝒜A_{\Lambda}\in\mathcal{A}, we have Map(AΛ)W16\operatorname{Map}(A_{\Lambda})\subset W^{16}. Now we apply Lemma 3.5. Let {Ωα}\{\Omega_{\alpha}\} be an uncountable collection of infinite subsets of Λ\Lambda such that ΩαΩβ\Omega_{\alpha}\cap\Omega_{\beta} is finite for all αβ\alpha\neq\beta. Note we may assume Λ\Lambda and each Ωα\Omega_{\alpha} contain infinitely many negative and positive integers.

Since all homeomorphisms are assumed to fix the boundary pointwise, we first modify each MiM_{i} by deleting a small regular open neighborhood of the xx-axis so that we can move them into one another with an appropriate homeomorphism. For each α\alpha, let fαf_{\alpha} be a homeomorphism supported in MM with the following property. For each nΩαn\in\Omega_{\alpha}, let fα(M2n)f_{\alpha}(M_{2n}) be the smallest connected subsurface containing the union of M2n+1,M2n+2,,M2k1M_{2n+1},M_{2n+2},...,M_{2k-1} where kΩαk\in\Omega_{\alpha} is the smallest element in Ωα\Omega_{\alpha} larger than nn, so that fαf_{\alpha} maps AΩαA_{\Omega_{\alpha}} into the complementary region. Also let fαf_{\alpha} map the union of M2n+1,M2n+2,,M2k1M_{2n+1},M_{2n+2},...,M_{2k-1} into M2kM_{2k}. Note this homeomorphism exists by the change of coordinates principle. Since {Ωα}\{\Omega_{\alpha}\} is uncountable, there are some α\alpha and β\beta such that fαf_{\alpha} and fβf_{\beta} are in the same left translate gWgW for some gPMapc(S)¯g\in\overline{\operatorname{PMap}_{c}(S)}. Therefore, fα1fβf_{\alpha}^{-1}f_{\beta} and fβ1fαf_{\beta}^{-1}f_{\alpha} are both in W2W^{2}.

If nΩαn\notin\Omega_{\alpha}, then fα(M2n)M2mf_{\alpha}(M_{2n})\subseteq M_{2m} for some mΩαm\in\Omega_{\alpha}. If mΩβm\notin\Omega_{\beta}, then fβ1fα(M2n)f_{\beta}^{-1}f_{\alpha}(M_{2n}) is contained in some M2kM_{2k} where kΩβk\in\Omega_{\beta}. Since ΩαΩβ\Omega_{\alpha}\cap\Omega_{\beta} is finite, we conclude that, with the exception of finitely many values of nΩαn\notin\Omega_{\alpha}, the map fβ1fαf_{\beta}^{-1}f_{\alpha} takes M2nM_{2n} into AΩβAΛA_{\Omega_{\beta}}\subset A_{\Lambda}.

Reversing the role of α\alpha and β\beta, the same argument shows that with only finitely many exceptions of nΩβn\notin\Omega_{\beta}, fα1fβf^{-1}_{\alpha}f_{\beta} takes every M2nM_{2n} into AΛA_{\Lambda}. Let FF^{\prime} be the union of these two exceptional sets of integers. Now write nM2n\bigcup_{n\in\mathbb{Z}}M_{2n} as the union of X1=n(ΩαF)M2nX_{1}=\bigcup_{n\notin(\Omega_{\alpha}\cup F^{\prime})}M_{2n}, X2=n(ΩβF)M2nX_{2}=\bigcup_{n\notin(\Omega_{\beta}\cup F^{\prime})}M_{2n}, and X3=nFM2nX_{3}=\bigcup_{n\in F^{\prime}}M_{2n}.

fβ1fαMap(X1)(fβ1fα)1Map(AΛ)W16, and similarly\displaystyle f_{\beta}^{-1}f_{\alpha}\operatorname{Map}(X_{1})(f_{\beta}^{-1}f_{\alpha})^{-1}\subset\operatorname{Map}(A_{\Lambda})\subset W^{16},\text{ and similarly}
fα1fβMap(X2)(fα1fβ)1Map(AΛ)W16.\displaystyle f_{\alpha}^{-1}f_{\beta}\operatorname{Map}(X_{2})(f_{\alpha}^{-1}f_{\beta})^{-1}\subset\operatorname{Map}(A_{\Lambda})\subset W^{16}.

It follows that Map(X1)\operatorname{Map}(X_{1}), Map(X2)W20\operatorname{Map}(X_{2})\subset W^{20}, so Map(X1X2)W40\operatorname{Map}(X_{1}\cup X_{2})\subset W^{40}. Now we can complete the proof by repeating the above argument to the union of the odd MnM_{n}. ∎

We can strengthen Lemma 3.9 using the upgraded Sierpiński lemma.

Lemma 3.10.

Let SS be any surface containing a countable admissible family of disjoint subsurfaces {Sn}\{S_{n}\} each homeomorphic to a sliced Loch Ness monster, and WPMapc(S)¯W\subset\overline{\operatorname{PMap}_{c}(S)} a countably syndetic symmetric set. Let {Mn,m}m\{M_{n,m}\}_{m\in\mathbb{Z}} be the collection of standard pieces for SnS_{n}. Then there is a finite set FF such that

(n,m)FMap(Mn,m)W80\displaystyle\prod_{(n,m)\notin F}\operatorname{Map}(M_{n,m})\subset W^{80}
Proof.

We show the case where {Sn}n\{S_{n}\}_{n\in\mathbb{N}} is infinite since the finite case is similar. Apply Lemma 3.7 with 𝒜\mathcal{A} consisting of subsurfaces of the form AΛ=n,mΛMn,2mA_{\Lambda}=\bigcup_{n\in\mathbb{N},m\in\Lambda}M_{n,2m} to show Map(AΛ)W16\operatorname{Map}(A_{\Lambda})\subset W^{16} for some infinite Λ\Lambda\subset\mathbb{Z}. Apply Lemma 3.6 to ×Λ\mathbb{N}\times\Lambda, so that {Ωα}α\{\Omega_{\alpha}\}_{\alpha\in\mathbb{R}} are infinite subsets of ×Λ\mathbb{N}\times\Lambda with the properties listed in the lemma. Note we may assume ×Λ\mathbb{N}\times\Lambda and each Ωα\Omega_{\alpha} contain infinitely many positive and negative integers in the second coordinate.

Modify all of the standard pieces slightly as before, then for each α\alpha, let fαf_{\alpha} be a map supported in iSi\bigcup_{i\in\mathbb{N}}S_{i} with the following property. For all (n,m)Ωα(n,m)\in\Omega_{\alpha}, let fα(Mn,2m)f_{\alpha}(M_{n,2m}) be the smallest connected subsurface containing the union of Mn,2m+1,Mn,2m+2M_{n,2m+1},M_{n,2m+2},…, Mn,2k1M_{n,2k-1} where kk is the smallest second component among the elements of Ωα({n}×Λ)\Omega_{\alpha}\cap(\{n\}\times\Lambda) larger than mm. Now the proof is completed as before. ∎

3.4. Fragmentation

To work with the subgroup PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)}, we will use results of the author [5] for decomposing an element of PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} into simpler pieces.

Lemma 3.11.

Let SS be any infinite-type surface and fPMapc(S)¯f\in\overline{\operatorname{PMap}_{c}(S)}. There exist two sequences of compact subsurfaces {Ki}\{K_{i}\} and {Ci}\{C_{i}\}, with each sequence consisting of disjoint surfaces, and g,hPMapc(S)¯g,h\in\overline{\operatorname{PMap}_{c}(S)} such that

  1. (i)

    supp(g)iCi\operatorname{supp}(g)\subseteq\bigcup_{i}C_{i} and supp(h)iKi\operatorname{supp}(h)\subseteq\bigcup_{i}K_{i},

  2. (ii)

    f=hgf=hg.

Furthermore, if SS is a disk with handles we can assume the following:

  1. (i)

    Each Ki\partial K_{i} and Ci\partial C_{i} is a single essential simple closed curve.

  2. (ii)

    SiKiS\searrow\cup_{i}K_{i} and SiCiS\searrow\cup_{i}C_{i} are homeomorphic to SS with compact boundary components added accumulating to some subset of the ends.

We use SKS\searrow K or SKS_{K} to denote the surface obtained from cutting SS along KK. Similar to the definition of cutting along a curve or arc, SKS_{K} is a surface with boundary with an identification map from some subset of SK\partial S_{K} to some subset of K\partial K such that the quotient on SKKS_{K}\sqcup K is homeomorphic to SS. Note we can realize SKS_{K} as a subsurface of SS, in particular the closure of the complement of KK.

4. Proof of Main Results

Now we are ready to prove the results from the introduction. First, we prove the reverse directions of the main theorems as well as Theorem 1.3 using the tools from the previous sections. Then we prove the forward directions using the work Domat [6] and Lemma 2.3. We prove Corollary 1.1 and Theorem 1.2 afterward.

4.1. Reverse Directions of Theorems A and B

First, we will prove the reverse direction of Theorem B, and then we will prove the reverse direction of Theorem A and Theorem 1.3 with a similar method. Recall the Steinhaus property implies automatic continuity by Proposition 3.2. Let Stab(K)\operatorname{Stab}(K) denote the subgroup consisting of maps that pointwise fix a subsurface KK. When KK is finite-type, Stab(K)\operatorname{Stab}(K) is an open neighborhood of the identity, and the collection of all such stabilizers is a neighborhood basis of the identity.

Proposition 4.1.

Let SS be a connected sum of a finite-type surface with finitely many disks with handles. Then PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} is Steinhaus with constant 328.

Proof.

First, we explain the details for the sliced Loch Ness monster and then discuss how to extend the argument to the other cases.

Case 1: sliced Loch Ness monsters. Suppose SS is any sliced Loch Ness monster. Let WW be any countably syndetic symmetric subset of Map(S)\operatorname{Map}(S). By Proposition 3.3, let UU be an open neighborhood of the identity such that W2W^{2} is dense in UU, and find some compact subsurface KK such that Stab(K)U\operatorname{Stab}(K)\subseteq U. Note that any sliced Loch Ness monster SS has a compact exhaustion {Ki}\{K_{i}\} where each SKiS\searrow K_{i} is homeomorphic to SS (see Figure 3). Thus we can assume that SK=SKS_{K}=S\searrow K is homeomorphic to SS by replacing KK with a large enough KiK_{i} if needed. Now we want to show that Map(SK)=Stab(K)Wk\operatorname{Map}(S_{K})=\operatorname{Stab}(K)\subseteq W^{k} for some kk.

Let fMap(SK)f\in\operatorname{Map}(S_{K}) be any element. Let gg be one of the maps produced by applying Lemma 3.11 to ff, and assume the conditions of the second part of Lemma 3.11 hold for gg. Note here we are using Lemma 3.1 which implies Map(SK)=PMapc(SK)¯\operatorname{Map}(S_{K})=\overline{\operatorname{PMap}_{c}(S_{K})}. Let {Ki}\{K_{i}\} be the sequence of compact subsurfaces containing the support of gg. Let {Mi}\{M_{i}\} be the collection of standard pieces for SKS_{K}. We can assume by applying change of coordinates that each MiM_{i} contains exactly one of the KiK_{i}, and each KiK_{i} appears in some MiM_{i}. By Lemma 3.9, we have some cofinite union T=iFMiT=\bigcup_{i\in\mathbb{Z}\setminus F}M_{i} with Map(T)W80\operatorname{Map}(T)\subset W^{80}. Therefore, we can find some gMap(T)g^{\prime}\in\operatorname{Map}(T) such that ggPMapc(SK)g^{\prime}g\in\operatorname{PMap}_{c}(S_{K}). Now let KSKK^{\prime}\subset S_{K} be a compact subsurface bounded by a single curve that contains the support of ggg^{\prime}g. Now using the density of W2W^{2} in Map(SK)\operatorname{Map}(S_{K}), find some element ϕW2\phi\in W^{2} such that ϕ(K)T\phi(K^{\prime})\subset T. It follows that ϕggϕ1W80\phi g^{\prime}g\phi^{-1}\in W^{80}, and therefore ggW84g^{\prime}g\in W^{84}. Finally, this gives gW164g\in W^{164}, and since the above argument also applies to the other element from fragmentation, fW328f\in W^{328}.

Case 2: disks with handles. Now assume SS is a disk with handles, and let WW be a countably syndetic symmetric subset of PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)}. Let UU be an open neighborhood of the identity such that W2W^{2} is dense in UU, and find some compact subsurface KK such that Stab(K)U\operatorname{Stab}(K)\subseteq U. We now claim we can enlarge KK so that each component of SK=SKS_{K}=S\searrow K has exactly one boundary chain and no compact boundary components. First note that there exists a compact exhaustion {Ki}\{K_{i}\} of SS such that each Ki\partial K_{i} is a single component that intersects S\partial S in a union of closed intervals, and the components of SKiS\searrow K_{i} are infinite-type without compact boundary components. To build such an exhaustion, start with the disk with boundary points removed used to construct SS, call it DD, and find a compact exhaustion {Ci}\{C_{i}\} of DD such that each CiC_{i} is a disk and each DCiD\searrow C_{i} is a union of disks with boundary points removed. Then we get the desired exhaustion on SS by attaching handles to DD and modifying the CiC_{i} accordingly. During this last step, we must require the attaching regions for any handle to be disjoint from each Ci\partial C_{i} and that whenever one attaching region of some handle is contained in the interior of CiC_{i}, then then other attaching region is also contained in the interior. Now note each SKiS\searrow K_{i} has a single boundary chain since SS has one boundary chain and all of the boundary components of SKiS\searrow K_{i} point to boundary ends of SS. The claim follows by replacing KK with a large enough KiK_{i} if needed. Now since SKS_{K} is also infinite-type and has no interior ends, Lemma 2.3 implies SKS_{K} is a disjoint union of disks with handles. The complement of any compact subsurface necessarily has a finite number of components, so SKS_{K} is a finite disjoint union.

Let fMap(SK)f\in\operatorname{Map}(S_{K}) be any element. Let gg be one of the maps produced by applying fragmentation to ff, and assume the conditions of the second part of Lemma 3.11 hold for gg when restricted to each component of SKS_{K}. In this case, we need to apply Lemma 3.11 to each component separately and then combine. Let {Ki}\{K_{i}\} be the collection of compact subsurfaces containing the support of gg. Using Lemma 2.2, we can cut SKS_{K} along a collection of disjoint arcs into sliced Loch Ness monsters {Sn}\{S_{n}\}. We can also assume these arcs are chosen to miss the KiK_{i}.

Let {Mn,m}\{M_{n,m}\} be the collection of standard pieces for SnS_{n}. By change of coordinates, we can assume each Mn,mM_{n,m} contains exactly one KiK_{i}, and each KiK_{i} appears in some Mn,mM_{n,m}. By Lemma 3.10, there is a finite set FF such that T=(n,m)FMn,mT=\bigcup_{(n,m)\notin F}M_{n,m} and Map(T)W80\operatorname{Map}(T)\subset W^{80}. Proceed as before.

Case 3: connected sums. We now need to consider the general case when SS is a connected sum of a finite-type surface with finitely many disks with handles. Let WW be a countably syndetic symmetric subset of PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)}. Let UU be an open neighborhood of the identity such that W2W^{2} is dense in UU, and find some finite-type subsurface KK such that Stab(K)U\operatorname{Stab}(K)\subseteq U. By choosing KK large enough to contain all the punctures and compact boundary components, we can ensure each component of SKS\searrow K has one boundary chain, no interior ends, and no compact boundary components. By applying Lemma 2.3, we have that SKS\searrow K is homeomorphic to a disjoint union of finitely many disks with handles. Proceed as in the disk with handles case. ∎

Now we show the reverse direction of Theorem A using similar techniques. This will also prove Theorem 1.3, the extension to the full mapping class group.

Proposition 4.2.

Let SS be a surface satisfying the conditions of Proposition 4.1 and with finitely many ends accumulated by genus. Then PMap(S)\operatorname{PMap}(S) and Map(S)\operatorname{Map}(S) are Steinhaus with constant 328.

Proof.

Let WW be a countably syndetic symmetric subset of PMap(S)\operatorname{PMap}(S). Let UU be an open neighborhood of the identity such that W2W^{2} is dense in UU, and find some compact subsurface KK such that Stab(K)U\operatorname{Stab}(K)\subseteq U. As before we can enlarge KK if needed so that SKS\searrow K is homeomorphic to a disjoint union of finitely many disks with handles. Since there are finitely many ends accumulated by genus, we can further assume that each component of SKS\searrow K is a sliced Loch Ness monster. This ensures that Stab(K)PMapc(S)¯\operatorname{Stab}(K)\subset\overline{\operatorname{PMap}_{c}(S)} so we can use fragmentation as in the proof of Proposition 4.1 to show Stab(K)W328\operatorname{Stab}(K)\subset W^{328}. The proof for Map(S)\operatorname{Map}(S) is identical. Note that we must also use a minor adaptation of Lemma 3.7 where WW is a countably syndetic symmetric subset of PMap(S)\operatorname{PMap}(S) or Map(S)\operatorname{Map}(S) instead of PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)}. ∎

4.2. Forward Directions of Theorems A and B

Now we finish the proofs of Theorems A and B using the work of Domat [6]. A nondisplaceable surface in SS will refer to a subsurface KK disjoint from the noncompact boundary components of SS such that f(K)Kf(K)\cap K\neq\varnothing for all representatives of fPMapc(S)¯f\in\overline{\operatorname{PMap}_{c}(S)}. Note a subsurface KK is nondisplaceable if it separates ends; i.e., if SKS\searrow K is disconnected and induces a partition of E(S)E(S) into two or more sets. A subsurface is also nondisplaceable if it separates the ends of the interior surface, so a subsurface that separates boundary components or separates a boundary component from an end is also nondisplaceable. The following result is implicit from Sections 6,7,8, and 10 of Domat’s paper.

Lemma 4.3 (Domat).

Let SS be an infinite-type surface such that either

  1. (i)

    SS has genus at least 3 and there exists an infinite sequence of disjoint nondisplaceable essential annuli that eventually leaves every compact subsurface.

  2. (ii)

    SS has any genus and there exists an infinite sequence of disjoint nondisplaceable essential spheres with n punctures and b boundary components for n+b8n+b\geq 8, and the sequence eventually leaves every compact subsurface.

Then there exists a discontinuous homomorphism ϕ:PMapc(S)¯\phi:\overline{\operatorname{PMap}_{c}(S)}\rightarrow\mathbb{Q}.

Although Domat’s work focused on surfaces with compact boundary, the conditions in Lemma 4.3 hold for some surfaces with noncompact boundary, and the proof goes through without adaptation. For surfaces with only compact boundary components, Domat showed the first condition holds when the interior of SS has at least two ends and at least one end accumulated by genus and the second condition holds when the interior has infinitely many ends. In the case of the Loch Ness monster, there are no finite-type nondisplaceable subsurfaces, so this case had to be handled separately. Domat and the author showed in the appendix of Domat’s paper that Map(S)\operatorname{Map}(S) does not have automatic continuity when SS is a Loch Ness monster.

Proof of Theorem B.

Recall the reverse direction was shown in Proposition 4.1. When SS either has infinitely many interior planar ends, infinitely many compact boundary components, infinitely many boundary chains, or at least one interior end accumulated by genus we will show one of the conditions of Lemma 4.3 is satisfied so that PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} does not have automatic continuity. After possibly filling in the finite number of punctures and capping the finite number of compact boundary components, we apply Lemma 2.3 to conclude SS is a connected sum of finitely many disks with handles. The original SS can then be obtained by connect summing with a finite-type surface with punctures and boundary components.

Case 1: infinitely many interior planar ends. When SS has infinitely many planar interior ends, there is a closed neighborhood UU of one of these ends such that each component of U\partial U is compact, and UU has infinitely many planar ends. Now this case reduces to the cases originally considered by Domat, and the second condition of Lemma 4.3 holds.

Case 2: infinitely many compact boundary components. When SS has infinitely many compact boundary components, there is some end accumulated by compact boundary, and every closed neighborhood of this end contains infinitely many compact boundary components. Let {Ui}\{U_{i}\} be a system of closed neighborhoods of this end such that Ui+1UiU_{i+1}\subset U_{i} for all ii, and i=1Ui=\bigcap_{i=1}^{\infty}U_{i}=\varnothing. Now the second condition of Lemma 4.3 holds by inductively choosing an essential punctured sphere in some sufficiently small UiU_{i} that misses the previously chosen punctured spheres.

Case 3: infinitely many boundary chains. When SS has infinitely many boundary chains, there is some end such that every closed neighborhood has infinitely many boundary chains. Let {Ui}\{U_{i}\} be a system of closed neighborhoods of this end satisfying the two properties from the previous case. Since the interior of each UiU_{i} has infinitely many ends, we can use an inductive procedure as before to show the second condition of Lemma 4.3 holds.

Case 4: at least one interior end accumulated by genus. When SS has an interior end accumulated by genus, there is a closed neighborhood UU of this end such that each component of U\partial U is compact, and UU has infinite genus. We can assume SS is not the Loch Ness monster since this case was ruled out by Domat and the author, so the interior of SS has at least two ends. Now this case reduces to the cases considered by Domat, and the first condition of Lemma 4.3 holds. The interior of SS having at least two ends ensures that we can find annuli that separate ends of the interior. ∎

Proof of Theorem A.

Recall the reverse direction was shown in Proposition 4.2. Now we consider two cases, and then we are done by using Lemma 2.3 as in Theorem B.

Case 1: infinitely many ends accumulated by genus. First, suppose SS has infinitely many ends accumulated by genus. Let ϕ1:PMap(S)ω\phi_{1}:\operatorname{PMap}(S)\rightarrow\mathbb{Z}^{\omega} be the projection mapping given by Lemma 3.1, and let ϕ2:ω(2)ω\phi_{2}:\mathbb{Z}^{\omega}\rightarrow(\mathbb{Z}_{2})^{\omega} be the mod 2 homomorphism. Now we use the discontinuous homomorphism ψ:(2)ω2\psi:(\mathbb{Z}_{2})^{\omega}\rightarrow\mathbb{Z}_{2} from Example 1.4 of Rosendal [10]. Composing all of the above homomorphisms yields a discontinuous homomorphism ψϕ2ϕ1:PMap(S)2\psi\circ\phi_{2}\circ\phi_{1}:\operatorname{PMap}(S)\rightarrow\mathbb{Z}_{2}. To show that ψϕ2ϕ1\psi\circ\phi_{2}\circ\phi_{1} is discontinuous we use the discontinuity of ψ\psi and the fact that the ϕi\phi_{i} are surjective and thus open by the open mapping theorem for Polish groups.

Case 2: finitely many ends accumulated by genus. Now suppose SS has finitely many ends accumulated by genus and satisfies one of the conditions for Lemma 4.3. Note PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} is a countable index subgroup of PMap(S)\operatorname{PMap}(S) by Lemma 3.1. By Lemma 4.3, there is a map ϕ:PMapc(S)¯\phi:\overline{\operatorname{PMap}_{c}(S)}\rightarrow\mathbb{Q} such that ker(ϕ)\ker(\phi) is not open. Note we also have that ker(ϕ)\ker(\phi) is not open in PMap(S)\operatorname{PMap}(S). Since ker(ϕ)\ker(\phi) is countable index in PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)}, it must also be countable index in PMap(S)\operatorname{PMap}(S). Now PMap(S)\operatorname{PMap}(S) cannot have automatic continuity by Proposition 3.4. ∎

4.3. Consequences

Now we finish with the proofs of some additional results.

Proof of Theorem 1.2.

Recall the Loch Ness monster case was shown by Domat and the author [6], so we can assume the interior of SS has at least two ends. By Lemma 4.3, PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} has a nonopen countable index subgroup. Since PMap(S)\operatorname{PMap}(S) is finite index in Map(S)\operatorname{Map}(S) by the assumption of finitely many ends, we can apply the same proof as the second case of Theorem A to show Map(S)\operatorname{Map}(S) does not have automatic continuity. ∎

Proof of Corollary 1.1.

Suppose SS is a disk with handles. Proceeding by contradiction, let HH be the kernel of a nontrivial map from PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} to a countable group with the discrete topology. By automatic continuity, HH is open and closed. Since HH is closed, it suffices to show that it contains every compactly supported mapping class since then HH is dense in PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)}, and in fact H=PMapc(S)¯H=\overline{\operatorname{PMap}_{c}(S)}. Since HH is open, it contains Stab(K)\operatorname{Stab}(K) for some compact subsurface KK. Now let ϕ\phi be any compactly supported mapping class. Since ϕ\phi fixes the boundary pointwise, we can isotope it so that it is supported in a subsurface KK^{\prime} that does not intersect the boundary. Since SS is a disk with handles, its interior is a Loch Ness monster, and there exists some homeomorphism supported in the interior that takes KK^{\prime} into the complement of KK. Therefore, a conjugate of ϕ\phi lies in Stab(K)H\operatorname{Stab}(K)\subset H, and we are done since HH is normal. There are no proper finite index subgroups in PMapc(S)¯\overline{\operatorname{PMap}_{c}(S)} since any index nn subgroup determines a nontrivial homomorphism to the symmetric group on nn elements via the left multiplication action on the left cosets. ∎

References

  • [1] Aramayona, J; Patel, P; Vlamis, N. The first integral cohomology of pure mapping class groups. Int. Math. Res. Not. IMRN. (2020), no. 22, 8973-8996. \mrev4216709, \zbl1462.57020, 10.1093/imrn/rnaa229, \arx1711.03132.
  • [2] Aramayona, J; Leininger, C; McLeay, A. Big mapping class groups and the co-Hopfian property. Michigan Math. J. 74 (2024), no. 2, 253-281. \mrev4739839, \zbl7855086, 10.1307/mmj/20216075, \arx2101.07188.
  • [3] Bestvina, M; Bromberg, K; Fujiwara, K. Constructing group actions on quasi-trees and applications to mapping class groups. Publ. Math. IHÉS. 122 (2015), 1-64. \mrev3415065, \zbl1372.20029, 10.1007/s10240-014-0067-4, \arx1006.1939.
  • [4] Brown, E; Messer, R. The classification of two-dimensional manifolds. Trans. Am. Math. Soc. 255 (1979), 377-402. \mrev0542887, \zbl0391.57010, 10.2307/1998182.
  • [5] Dickmann, R. Mapping class groups of surfaces with noncompact boundary components. Algebraic & Geometric Topology. 23 (2023), no. 6, 2777-2821. \mrev4640140, \zbl07786947, 10.2140/agt.2023.23.2777, \arx2101.00564.
  • [6] Domat, G. Big pure mapping class groups are never perfect. Appendix with Ryan Dickmann. Math. Res. Lett. 29 (2022), no. 3, 691-726. \mrev4516036, \zbl07626171, 10.4310/MRL.2022.v29.n3.a4, \arx2007.14929.
  • [7] Mann, K. Automatic continuity for homeomorphism groups and big mapping class groups. Michigan Math. J. 74 (2024), no. 1, 215-224. \mrev4718498, \zbl7828366, 10.1307/mmj/20216095, \arx2003.01173.
  • [8] Masur, H; Minsky, Y. Geometry of the complex of curves II: hierarchical structure. Geom. Funct. Anal. 10 (2000), no. 4, 902-974. \mrev1791145, \zbl0972.32011, 10.1007/PL00001643, \arxmath/9807150.
  • [9] Richards, I. On the classification of noncompact surfaces. Trans. Amer. Math. Soc. 106 (1963), 259-269. \mrev0143186, \zbl0156.22203, 10.2307/1993768.
  • [10] Rosendal, C. Automatic continuity of group homomorphisms. Bull. Symbolic Logic. 15 (2009), no. 2, 184-214. \mrev2535429, \zbl1173.03037, 10.2178/bsl/1243948486.
  • [11] Rosendal, C; Solecki, S. Automatic continuity of homomorphisms and fixed points on metric compacta. Israel J. Math. 162 (2007), 349-371. \mrev2365867, \zbl1146.22003, 10.1007/s11856-007-0102-y.
  • [12] Vlamis, N. Three perfect mapping class groups. New York J. Math. 27 (2021), 468-474. \mrev4226155, \zbl1459.57022, \arx2012.05993.