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

Orientation-preserving homeomorphisms of
Euclidean space are commutators

Megha Bhat Department of Mathematics
CUNY Graduate Center
New York, NY 10016
mbhat@gradcenter.cuny.edu
   Nicholas G. Vlamis Department of Mathematics
CUNY Graduate Center
New York, NY 10016, and
Department of Mathematics
CUNY Queens College
Flushing, NY 11367
nvlamis@gc.cuny.edu
Abstract

We prove that every orientation-preserving homeomorphism of Euclidean space can be expressed as a commutator of two orientation-preserving homeomorphisms. We give an analogous result for annuli. In the annulus case, we also extend the result to the smooth category in the dimensions for which the associated sphere has a unique smooth structure. As a corollary, we establish that every orientation-preserving diffeomorphism of the real line is the commutator of two orientation-preserving diffeomorphisms.

1 Introduction

In 1951, Ore [Ore51] initiated the investigation of groups in which every element can be expressed as a commutator. In particular, he proved that this holds for finite alternating groups and the symmetric group on \mathbb{N}. Much more recently, for each nn\in\mathbb{N}, Tsuboi [Tsu13] showed that the group of orientation-preserving homeomorphisms of the nn-sphere 𝕊n\mathbb{S}^{n} has this property. At roughly the same time, Basmajian–Maskit [BM12] proved that every orientation-preserving isometry of 𝕊n1\mathbb{S}^{n-1}, n\mathbb{R}^{n}, and n\mathbb{H}^{n} can be expressed as a commutator for n3n\geq 3, where n\mathbb{H}^{n} is hyperbolic nn-space.

More than 50 years earlier, Anderson [And58] showed that each element of Homeo+(𝕊2)\operatorname{Homeo}^{+}(\mathbb{S}^{2}) and Homeo+(𝕊3)\operatorname{Homeo}^{+}(\mathbb{S}^{3}) can be expressed as the product of two commutators. Using the generalized Schönflies theorem and the annulus theorem (see the preliminaries below), the techniques used by Anderson in dimension two can be extended to show that, for all nn\in\mathbb{N}, every element of Homeo+(𝕊n)\operatorname{Homeo}^{+}(\mathbb{S}^{n}) can be expressed as a product of two commutators. The same result holds for Homeo+(n)\operatorname{Homeo}^{+}(\mathbb{R}^{n}) and Homeo0(𝕊n1×)\operatorname{Homeo}_{0}(\mathbb{S}^{n-1}\times\mathbb{R}) (a proof for n=2n=2, using ideas of Le Roux–Mann [LRM18], can be found in [Vla24]; the same proof technique can be used to establish the result in the higher dimensional cases as well). Above, Homeo0(M)\operatorname{Homeo}_{0}(M) denotes the connected component of the identity of Homeo(M)\operatorname{Homeo}(M) equipped with the compact-open topology. We note that Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}) coincides with Homeo+(𝕊n)\operatorname{Homeo}^{+}(\mathbb{S}^{n}), Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}) with Homeo+(n)\operatorname{Homeo}^{+}(\mathbb{R}^{n}), and Homeo0(𝕊n×)\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}) with the subgroup of Homeo+(𝕊n×)\operatorname{Homeo}^{+}(\mathbb{S}^{n}\times\mathbb{R}) stabilizing each end of 𝕊n×\mathbb{S}^{n}\times\mathbb{R} (see the preliminaries below).

Given this history and the work of Tsuboi and Basmajian–Maskit, it is natural to ask if every element of Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}) and Homeo0(𝕊n×)\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}) can be expressed as a commutator: our theorem answers this in the affirmative.

Theorem 1.

For each nn\in\mathbb{N}, every element of Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}) and Homeo0(𝕊n×)\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}) can be expressed as a single commutator.

The proof of our theorem uses the same philosophy as Tsuboi’s argument in the spherical case: A group element ff can be expressed as a commutator if and only if there exists a group element gg such that gfgf and gg are conjugate. Tsuboi’s idea is to start with an orientation-preserving homeomorphism ff of a sphere and construct a homeomorphism gg having strong enough hyperbolic dynamics with respect to ff so that gfg\circ f exhibits the same dynamics as gg. He then uses this dynamical picture to guarantee gfg\circ f and gg are conjugate.

The discussion above is a particular instance of a more general phenomenon. A word is an element in a finite-rank free group. Given rr\in\mathbb{N} and a word wx1x2xr{w\in\mathbb{Z}x_{1}*\mathbb{Z}x_{2}*\cdots*\mathbb{Z}x_{r}}, we write w=w(x1,x2,,xr)w=w(x_{1},x_{2},\ldots,x_{r}) and view ww as an expression in the variables x1,,xrx_{1},\ldots,x_{r}. Then, given a group GG, we have a substitution map w:GrGw\colon\thinspace G^{r}\to G given by (g1,,gr)(g_{1},\ldots,g_{r}) maps to w(g1,,gr)w(g_{1},\ldots,g_{r}) in GG. Let w(G)w(G) be the subgroup of GG generated by the set Gw={w(𝐠),w(𝐠)1:𝐠Gr}{G_{w}=\{w(\mathbf{g}),w(\mathbf{g})^{-1}:\mathbf{g}\in G^{r}\}}. Then, w(G)w(G) is a normal subgroup of GG (in fact, it is characteristic). The ww-width of GG is the smallest natural number pp such that every element of w(G)w(G) can be expressed as a product of pp elements in GwG_{w}; if such a number does not exist then the ww-width is said to be infinite (see [Seg09] for more details). For example, if w=xyx1y1xyw=xyx^{-1}y^{-1}\in\mathbb{Z}x*\mathbb{Z}y, then the ww-width of a group is known as its commutator width. In this language, Tsuboi’s theorem implies that Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}) has commutator width one, and our theorem implies Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}) and Homeo0(𝕊n×)\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}) also have commutator width one.

A group GG is uniformly simple if there exists kk\in\mathbb{N} such that for all g,fGg,f\in G with gg nontrivial, ff can be expressed as the product of at most kk conjugates of gg and g1g^{-1}. Anderson [And58] showed that Homeo0(𝕊2)\operatorname{Homeo}_{0}(\mathbb{S}^{2}) is uniformly simple, and his argument can be extended to other dimensions using the generalized Schönflies theorem and the annulus theorem; in fact, he showed that kk can be chosen to be eight. Therefore, given any nontrivial word ww, w(Homeo0(𝕊n))=Homeo0(𝕊n)w(\operatorname{Homeo}_{0}(\mathbb{S}^{n}))=\operatorname{Homeo}_{0}(\mathbb{S}^{n}) and the ww-width of Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}) is at most eight. In light of Tsuboi’s result, it is natural to ask for which words Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}) has width one. This can also be asked for Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}) and Homeo0(𝕊n×)\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}); but, unlike Homeo(𝕊n)\operatorname{Homeo}(\mathbb{S}^{n}), neither Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}) nor Homeo0(𝕊n×)\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}) are simple, as the subgroup of compactly supported homeomorphisms is a nontrivial proper normal subgroup. However, the germ at the end of n\mathbb{R}^{n} (resp., at an end of 𝕊n×\mathbb{S}^{n}\times\mathbb{R}) is uniformly simple; this first appears in the (unpublished) thesis of Ling [Lin80] (see also [Man16a] and [Sch11]).

Problem.

Characterize the words for which the word width of Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}) (resp., Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}), Homeo0(𝕊n×)\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R})) is equal to one.

As a corollary to our theorem, we show that the ww-width is one for a particular class of words.

Corollary 2.

Let n,rn,r\in\mathbb{N} with r>1r>1, and let GG denote any one of Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}), Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}), or Homeo0(𝕊n×)\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}). If p1,p2,,pr{0}p_{1},p_{2},\ldots,p_{r}\in\mathbb{Z}\smallsetminus\{0\} and gGg\in G, then there exists g1,g2,,grG{1}g_{1},g_{2},\ldots,g_{r}\in G\smallsetminus\{1\} such that g=i=1rgipig=\prod_{i=1}^{r}g_{i}^{\circ p_{i}}.

The proof of 2 will be given at the end of the note.

A remark on diffeomorphisms

Here, we provide an accounting of the extent to which our methods extend to the smooth category. For a smooth manifold MM, let Diff(M)\operatorname{Diff}(M) denote the group of diffeomorphisms MMM\to M, and let Diff0(M)\operatorname{Diff}_{0}(M) denote the connected component of the identity in Diff(M)\operatorname{Diff}(M) equipped with the compact-open CC^{\infty}-topology. Thurston [Thu74] proved that if MM is closed then Diff0(M)\operatorname{Diff}_{0}(M) is perfect (see [Man16b] for a short proof). Burago–Ivanov–Polterovich [BIP08] showed that Diff0(𝕊n)\operatorname{Diff}_{0}(\mathbb{S}^{n}) is uniformly perfect with commutator width bounded above by four; Rybicki [Ryb11] proved the same for a class of open manifolds, including Diff0(n)\operatorname{Diff}_{0}(\mathbb{R}^{n}) and Diff0(𝕊n×)\operatorname{Diff}_{0}(\mathbb{S}^{n}\times\mathbb{R}). Given this history and the fact that Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}), Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}), and Homeo0(𝕊n×)\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}) all have commutator width one, it is natural to ask if the same holds for diffeomorphisms.

In the case of annuli, we can give a positive answer to this question in specific dimensions, namely in the dimensions nn for which the nn-sphere is known to have a unique smooth structure (with the exception of n=5n=5). In these dimensions, as we will now explain, the preliminary results—as pertain to annuli—presented in Section 2 can be extended to the smooth category. Moreover, for annuli in these dimensions, the arguments throughout the note go through verbatim.

Let 𝔹n\mathbb{B}^{n} be the closed unit ball in n\mathbb{R}^{n}, and let Diff(𝔹n)\operatorname{Diff}_{\partial}(\mathbb{B}^{n}) be the group of diffeomorphisms 𝔹n𝔹n\mathbb{B}^{n}\to\mathbb{B}^{n} fixing a neighborhood of 𝔹n\partial\mathbb{B}^{n} pointwise. A point of weakness in extending our results to the smooth setting is 2.3, which does not extend in all dimensions. However, the proof of 2.3 given below is valid in the smooth category for the dimensions nn in which Diff(𝔹n)\operatorname{Diff}_{\partial}(\mathbb{B}^{n}) is connected. When n{1,2,3}n\in\{1,2,3\}, it is known that Diff(𝔹n)\operatorname{Diff}_{\partial}(\mathbb{B}^{n}) is connected, and when n5n\geq 5, the number of components of Diff(𝔹n)\operatorname{Diff}_{\partial}(\mathbb{B}^{n}) is equal to the number of exotic spheres in dimension nn, which is encoded by the cardinality of the group Θn+1\Theta_{n+1} of h-cobordism classes of homotopy (n+1)(n+1)-spheres; we refer the reader to the historical remarks section of [Kup19] for a more detailed discussion and for references.

The other results from Section 2 can be adapted using the h-cobordism theorem [Sma62], which puts an additional dimension restriction, as it is not known whether the h-cobordism theorem holds in dimension three (it fails in general in dimension four, but this turns out not to be relevant to this note). We refer the reader to Milnor’s notes for details on the h-cobordism theorem and its applications, specifically [Mil65, §9 and Concluding Remarks]. These are the only dimensional obstructions that arise in extending our arguments for annuli to the smooth setting. Accounting for these restrictions allows us to record the following theorem.

Theorem 3.

Let n{4,5}n\in\mathbb{N}\smallsetminus\{4,5\}. If Θn\Theta_{n} is trivial, then Diff0(𝕊n1×)\operatorname{Diff}_{0}(\mathbb{S}^{n-1}\times\mathbb{R}) has commutator width one. ∎

Appealing to [WX17, Corollary 1.15] for the known values of n{1,,61}n\in\{1,\ldots,61\} in which |Θn|=1|\Theta_{n}|=1, we have:

Corollary 4.

If n{1,2,3,6,12,56,61}n\in\{1,2,3,6,12,56,61\}, then Diff0(𝕊n1×)\operatorname{Diff}_{0}(\mathbb{S}^{n-1}\times\mathbb{R}) has commutator width one. ∎

In the above corollary, setting n=1n=1, we obtain that Diff0(𝕊0×)Diff0()×Diff0()\operatorname{Diff}_{0}(\mathbb{S}^{0}\times\mathbb{R})\cong\operatorname{Diff}_{0}(\mathbb{R})\times\operatorname{Diff}_{0}(\mathbb{R}) has commutator width one, yielding the following corollary.

Corollary 5.

Diff0()\operatorname{Diff}_{0}(\mathbb{R}) has commutator width one. ∎

Our proofs as written do not immediately extend to the case of n\mathbb{R}^{n}, and tracing through the arguments, one finds that the issue can be reduced to a question of differentiability at a single point in the proofs of 2.8 and 3.2. For annuli, these issues are pushed off to infinity, allowing us to extend to the smooth category.

Acknowledgements

The authors thank the anonymous referee for their comments and for suggesting we add a discussion of diffeomorphisms, which did not exist in the original draft. The second author is supported by NSF DMS-2212922 and PSC-CUNY Awards #65331-00 53 and #66435-00 54

2 Preliminaries

Before we begin, we will need the generalized Schönflies theorem, the annulus theorem, and the characterizations of Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}), Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}), and Homeo0(𝕊n×)\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}) given in the introduction. An (n1)(n-1)-dimensional submanifold NN of an nn-manifold MM is locally flat if each point of NN has an open neighborhood UU in MM such that the pair (U,UN)(U,U\cap N) is homeomorphic to (n,n1)(\mathbb{R}^{n},\mathbb{R}^{n-1}). Additionally, if XMX\subset M is a closed subset with nonempty interior, then we say XX is locally flat if X\partial X is a locally flat (n1)(n-1)-dimensional submanifold of MM. In an nn-manifold MM, we use the terminology locally flat annulus to refer to a locally flat closed subset of MM that is homeomorphic to 𝕊n1×[0,1]\mathbb{S}^{n-1}\times[0,1].

Theorem 2.1 (Generalized Schönflies Theorem [Bro60, Bro62]).

If Σ\Sigma is a locally flat (n1)(n-1)-dimensional sphere in 𝕊n\mathbb{S}^{n}, then the closure of each component of 𝕊nΣ\mathbb{S}^{n}\smallsetminus\Sigma is homeomorphic to the closed nn-ball. ∎

Theorem 2.2 (Annulus Theorem [Kir69, Qui82]).

The closure of the region co-bounded by two disjoint locally flat (n1)(n-1)-dimensional spheres in 𝕊n\mathbb{S}^{n} is a locally flat annulus. ∎

A homeomorphism of n\mathbb{R}^{n} is stable if it can be factored as a composition of homeomorphisms each of which restricts to the identity on an open subset of n\mathbb{R}^{n}. The annulus theorem is equivalent to the stable homeomorphism theorem, which states that every homeomorphism of n\mathbb{R}^{n} is stable [BG64]. From this, together with Alexander’s trick, one readily deduces that every orientation-preserving homeomorphism of n\mathbb{R}^{n} (resp., 𝕊n\mathbb{S}^{n}) is isotopic to the identity. It is also possible to deduce from the stable homeomorphism theorem that every orientation-preserving homeomorphism of 𝕊n×\mathbb{S}^{n}\times\mathbb{R} that stabilizes the topological ends is isotopic to the identity; however, it is easier to see this fact by using the fragmentation lemma (such a proof can be found for n=1n=1 in [Vla24], which can be generalized to higher dimensions). The fragmentation lemma is a stronger version of the stable homeomorphism theorem that gives control of the open sets being fixed by each homeomorphism in the factorization; it is deduced from the work of Edwards–Kirby [EK71]111The authors learned of the fragmentation lemma from [Man16a]. We also note, for the sake of extending arguments to the smooth setting, that the fragmentation lemma for diffeomorphisms is an exercise (see [Man16b, Lemma 2.1]).. We record these facts in the following statement.

Theorem 2.3.

For nn\in\mathbb{N}, every orientation-preserving homeomorphism of 𝕊n\mathbb{S}^{n} (resp., n\mathbb{R}^{n}) is isotopic to the identity, and every orientation-preserving homeomorphism of 𝕊n×\mathbb{S}^{n}\times\mathbb{R} stabilizing the ends is isotopic to the identity.

The following lemma is required to glue together homeomorphisms defined on disjoint pieces of a sphere, especially in the change of coordinates corollary below and the construction of the conjugating map in 2.7. Before stating the lemma, we need a notion of orientation compatibility for disjoint embeddings of spheres.

Let ι1,ι2:𝕊n1𝕊n\iota_{1},\iota_{2}\colon\thinspace\mathbb{S}^{n-1}\to\mathbb{S}^{n} be embeddings with disjoint locally flat images. Then the annulus theorem implies that there exists a homeomorphism φ:𝕊n1×[0,1]𝕊n\varphi\colon\thinspace\mathbb{S}^{n-1}\times[0,1]\to\mathbb{S}^{n} such that φ(x,0)=ι1(x)\varphi(x,0)=\iota_{1}(x). Let ι:𝕊n1𝕊n\iota\colon\thinspace\mathbb{S}^{n-1}\to\mathbb{S}^{n} be given by ι(x)=φ(x,1)\iota(x)=\varphi(x,1). Then

ι1ι2:𝕊n1𝕊n1\iota^{-1}\circ\iota_{2}\colon\thinspace\mathbb{S}^{n-1}\to\mathbb{S}^{n-1}

is a homeomorphism. We say ι1\iota_{1} and ι2\iota_{2} are compatibly oriented if ι1ι2\iota^{-1}\circ\iota_{2} is orientation preserving.

Lemma 2.4.

Let ι1,ι2:𝕊n1𝕊n\iota_{1},\iota_{2}\colon\thinspace\mathbb{S}^{n-1}\to\mathbb{S}^{n} be embeddings such that Σ1:=ι1(𝕊n1)\Sigma_{1}:=\iota_{1}(\mathbb{S}^{n-1}) and Σ2:=ι2(𝕊n1)\Sigma_{2}:=\iota_{2}(\mathbb{S}^{n-1}) are locally flat and disjoint. If ι1\iota_{1} and ι2\iota_{2} are compatibly oriented, then there exists an isotopy φ:𝕊n1×[0,1]𝕊n\varphi\colon\thinspace\mathbb{S}^{n-1}\times[0,1]\to\mathbb{S}^{n} between ι1\iota_{1} and ι2\iota_{2} whose image is the locally flat annulus co-bounded by Σ1\Sigma_{1} and Σ2\Sigma_{2}.

Proof.

Let AA be the locally flat annulus co-bounded by Σ1\Sigma_{1} and Σ2\Sigma_{2}. Choose an embedding ιA:𝕊n1×[0,1]𝕊n\iota_{A}\colon\thinspace\mathbb{S}^{n-1}\times[0,1]\to\mathbb{S}^{n} whose image is AA and such that ιA|𝕊n1×{0}=ι1\iota_{A}|_{\mathbb{S}^{n-1}\times\{0\}}=\iota_{1}. Let ι=ιA|𝕊n1×{1}\iota=\iota_{A}|_{\mathbb{S}^{n-1}\times\{1\}}. Then τ=ι1ι2\tau=\iota^{-1}\circ\iota_{2} is an orientation-preserving homeomorphism of 𝕊n1\mathbb{S}^{n-1}; hence, τ\tau is isotopic to the identity. Choose an isotopy H:𝕊n1×[0,1]𝕊n1H\colon\thinspace\mathbb{S}^{n-1}\times[0,1]\to\mathbb{S}^{n-1} such that H(x,0)=xH(x,0)=x and H(x,1)=τ(x)H(x,1)=\tau(x). Define h:𝕊n1×[0,1]𝕊n1×[0,1]h\colon\thinspace\mathbb{S}^{n-1}\times[0,1]\to\mathbb{S}^{n-1}\times[0,1] by h(x,t)=(H(x,t),t)h(x,t)=(H(x,t),t). It is readily checked that hh is a homeomorphism and that φ:=ιAh\varphi:=\iota_{A}\circ h is the desired map. ∎

We record as a corollary the main ways in which we will use the results above, each of which being a type of change of coordinates.

Corollary 2.5 (Three change of coordinates principles).

Let nn\in\mathbb{N}.

  1. (1)

    Given any two locally flat (n1)(n-1)-dimensional spheres Σ\Sigma and Σ\Sigma^{\prime} in n\mathbb{R}^{n}, there exists a compactly supported ambient homeomorphism mapping Σ\Sigma onto Σ\Sigma^{\prime}.

  2. (2)

    Given two locally flat nn-dimensional spheres Σ\Sigma and Σ\Sigma^{\prime} in 𝕊n×[0,1]\mathbb{S}^{n}\times[0,1] each separating 𝕊n×{0}\mathbb{S}^{n}\times\{0\} from 𝕊n×{1}\mathbb{S}^{n}\times\{1\}, there exists an ambient homeomorphism mapping Σ\Sigma onto Σ\Sigma^{\prime} and fixing 𝕊n×{0,1}\mathbb{S}^{n}\times\{0,1\} pointwise.

  3. (3)

    Let x,y𝕊nx,y\in\mathbb{S}^{n}. Given a sequence of locally flat annuli {Ak}k\{A_{k}\}_{k\in\mathbb{Z}} with pairwise-disjoint interiors such that AkA_{k} and Ak+1A_{k+1} share a boundary sphere and such that 𝕊n{x,y}=kAk\mathbb{S}^{n}\smallsetminus\{x,y\}=\bigcup_{k\in\mathbb{Z}}A_{k}, there exists τHomeo+(𝕊n)\tau\in\operatorname{Homeo}^{+}(\mathbb{S}^{n}) such that τ(x)=x\tau(x)=x, τ(y)=y\tau(y)=y, and τ(Ak)=Ak+1\tau(A_{k})=A_{k+1}. ∎

We can now introduce the notion of a topologically loxodromic homeomorphism. Tsuboi uses the terminology “topologically hyperbolic homeomorphism”, but the definition given here is seemingly more stringent than the one he gives, and so we introduce a slight modification in our nomenclature to recognize this potential difference.

Definition 2.6.

An orientation-preserving homeomorphism τ:𝕊n𝕊n\tau\colon\thinspace\mathbb{S}^{n}\to\mathbb{S}^{n} is topologically loxodromic if there exists τ+,τ𝕊n\tau^{+},\tau^{-}\in\mathbb{S}^{n} fixed by τ\tau and a sequence of locally flat annuli {An}n\{A_{n}\}_{n\in\mathbb{Z}} with pairwise-disjoint interiors such that

  1. (i)

    𝕊n{τ±}=kAk\mathbb{S}^{n}\smallsetminus\{\tau_{\pm}\}=\bigcup_{k\in\mathbb{Z}}A_{k},

  2. (ii)

    AkA_{k} shares a boundary component with Ak+1A_{k+1},

  3. (iii)

    τ(Ak)=Ak+1\tau(A_{k})=A_{k+1}, and

  4. (iv)

    limk±τk(x)=τ±\lim_{k\to\pm\infty}\tau^{k}(x)=\tau^{\pm} for all x𝕊n{τ+,τ}x\in\mathbb{S}^{n}\smallsetminus\{\tau^{+},\tau^{-}\}.

We the say the sequence of annuli {An}n\{A_{n}\}_{n\in\mathbb{N}} is suited to τ\tau, and we call τ+\tau^{+} the sink of τ\tau and τ\tau^{-} the source. Note that every homeomorphism of n\mathbb{R}^{n} and 𝕊n1×\mathbb{S}^{n-1}\times\mathbb{R} can be viewed as a homeomorphism of 𝕊n\mathbb{S}^{n} that fixes one or two points, respectively, corresponding to the ends, and so we say a homeomorphism of either of these spaces is topologically loxodromic if it is a topologically loxodromic homeomorphism as a homeomorphism of 𝕊n\mathbb{S}^{n}.

The main motivating examples of topologically loxodromic homeomorphisms are loxodromic Möbius transformations of 𝕊n\mathbb{S}^{n}, dilations of n\mathbb{R}^{n}, and translations of 𝕊n×\mathbb{S}^{n}\times\mathbb{R} of the form (x,t)(x,t+t0)(x,t)\mapsto(x,t+t_{0}) for some t0t_{0}\in\mathbb{R}. Of course the latter two examples are just loxodromic Möbius transformations in different coordinates.

In the next proposition, we prove that any two topologically loxodromic homeomorphisms of 𝕊n\mathbb{S}^{n} are conjugate. This is the key fact in Tsuboi’s argument, and the following results appear as part of Tsuboi’s proof in reference to particular maps. We pull the ideas out here and formalize them in the general setting. As a corollary, we provide the analogous statement for n\mathbb{R}^{n} and 𝕊n×\mathbb{S}^{n}\times\mathbb{R}.

Proposition 2.7.

Any two topologically loxodromic homeomorphisms in Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}) are conjugate in Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}).

Proof.

Let τ,σHomeo0(𝕊n)\tau,\sigma\in\operatorname{Homeo}_{0}(\mathbb{S}^{n}) be topologically loxodromic. Let {An}n\{A_{n}\}_{n\in\mathbb{Z}} and {Bn}n\{B_{n}\}_{n\in\mathbb{Z}} be sequences of locally flat annuli suited to τ\tau and σ\sigma, respectively. Let ΣA\Sigma_{A} be the component of A0A_{0} such that A0=ΣAτ(ΣA)\partial A_{0}=\Sigma_{A}\cup\tau(\Sigma_{A}), and similarly, define ΣB\Sigma_{B} so that B0=ΣBσ(ΣB)\partial B_{0}=\Sigma_{B}\cup\sigma(\Sigma_{B}).

Fix embeddings ιA,ιB:𝕊n1𝕊n\iota_{A},\iota_{B}\colon\thinspace\mathbb{S}^{n-1}\to\mathbb{S}^{n} such that the images of ιA\iota_{A} and ιB\iota_{B} are ΣA\Sigma_{A} and ΣB\Sigma_{B}, respectively. Applying 2.4, we obtain embeddings φA,φB:𝕊n1×[0,1]𝕊n\varphi_{A},\varphi_{B}\colon\thinspace\mathbb{S}^{n-1}\times[0,1]\to\mathbb{S}^{n} whose images are A0A_{0} and B0B_{0}, respectively, and such that φA\varphi_{A} (resp., φB\varphi_{B}) is an isotopy between ιA\iota_{A} and τιA\tau\circ\iota_{A} (resp., ιB\iota_{B} and σιB\sigma\circ\iota_{B}). Set φ=φAφB1\varphi=\varphi_{A}\circ\varphi_{B}^{-1}.

Define h:𝕊n𝕊nh\colon\thinspace\mathbb{S}^{n}\to\mathbb{S}^{n} by h(σ±)=τ±h(\sigma_{\pm})=\tau_{\pm} and h(x)=(τnφσn)(x)h(x)=(\tau^{n}\circ\varphi\circ\sigma^{-n})(x) for xBnx\in B_{n}. It is now readily checked that hh is well-defined for the elements of BnBn+1B_{n}\cap B_{n+1}, establishing that hh is a well-defined orientation-preserving homeomorphism of 𝕊n\mathbb{S}^{n}. Now, for xAnx\in A_{n},

(hσh1)(x)\displaystyle(h\circ\sigma\circ h^{-1})(x) =[(τn+1φσ(n+1))σ(σnφ1τn)](x)\displaystyle=\left[(\tau^{n+1}\circ\varphi\circ\sigma^{-(n+1)})\circ\sigma\circ(\sigma^{n}\circ\varphi^{-1}\circ\tau^{-n})\right](x)
=τ(x).\displaystyle=\tau(x).

Hence, τ=hσh1\tau=h\circ\sigma\circ h^{-1}. ∎

Corollary 2.8.

Let GG denote either Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}) or Homeo0(𝕊n×)\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}). Any two topologically loxodromic homeomorphisms in GG that share a sink or a source are conjugate in GG.

Proof.

Viewing two given topologically loxodromic homeomorphisms in GG as homeomorphisms of 𝕊n\mathbb{S}^{n} with a shared sink, a shared source, or both, the conjugating map constructed in 2.7 will preserve any shared source or sink, and hence the conjugation occurs in the appropriate group. ∎

3 Proofs

We prove the theorem first for annuli then for Euclidean spaces.

Theorem 3.1.

For nn\in\mathbb{N}, every element of Homeo0(𝕊n×)\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}) is a commutator.

Proof.

Let fHomeo0(𝕊n×)f\in\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}). Let Σ0=𝕊n×{0}\Sigma_{0}=\mathbb{S}^{n}\times\{0\}, and set t0=0t_{0}=0. Let n0n_{0}\in\mathbb{N} such that A0:=𝕊n×[n0,n0]A_{0}:=\mathbb{S}^{n}\times[-n_{0},n_{0}] contains Σ0f(Σ0)\Sigma_{0}\cup f(\Sigma_{0}) in its interior. Now, choose t1,n1t_{1},n_{1}\in\mathbb{N} such that Σ1=𝕊n×{t1}\Sigma_{1}=\mathbb{S}^{n}\times\{t_{1}\} satisfies Σ1f(Σ1)\Sigma_{1}\cup f(\Sigma_{1}) is contained in the interior of A1:=𝕊n×[n0,n1]A_{1}:=\mathbb{S}^{n}\times[n_{0},n_{1}]. Continuing in this fashion in both directions, we construct sequences {nk}k\{n_{k}\}_{k\in\mathbb{Z}} and {tk}k\{t_{k}\}_{k\in\mathbb{Z}} of integers such that, by setting Σk=𝕊n×{tk}\Sigma_{k}=\mathbb{S}^{n}\times\{t_{k}\} and Ak=𝕊n×[nk,nk+1]A_{k}=\mathbb{S}^{n}\times[n_{k},n_{k+1}], we have Σkf(Σk)\Sigma_{k}\cup f(\Sigma_{k}) is contained in the interior of AkA_{k}.

Now, let gHomeo0(𝕊n×)g^{\prime}\in\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}) such that g(Ak)=Ak+1g^{\prime}(A_{k})=A_{k+1}. Note, by construction, kAk\bigcup_{k\in\mathbb{Z}}A_{k} is all of 𝕊n×\mathbb{S}^{n}\times\mathbb{R}, and so gg^{\prime} is topologically loxodromic. As both (gf)(Σk)(g^{\prime}\circ f)(\Sigma_{k}) and Σk+1\Sigma_{k+1} are locally flat annuli contained in the interior of Ak+1A_{k+1}, we can choose hkHomeo0(𝕊n×)h_{k}\in\operatorname{Homeo}_{0}(\mathbb{S}^{n}\times\mathbb{R}) that is supported in the interior of Ak+1A_{k+1} and satisfies hk(g(f(Σk)))=Σk+1h_{k}(g^{\prime}(f(\Sigma_{k})))=\Sigma_{k+1}. The sequence {hk}k\{h_{k}\}_{k\in\mathbb{Z}} consists of homeomorphisms with pairwise-disjoint supports, and hence, we can define h=khkh=\prod_{k\in\mathbb{Z}}h_{k}. Set g=hgg=h\circ g^{\prime}.

As the support of each hkh_{k} is contained in the interior of Ak+1A_{k+1}, we have that

g(Ak)=hk(g(Ak))=hk(Ak+1)=Ak+1;g(A_{k})=h_{k}(g^{\prime}(A_{k}))=h_{k}(A_{k+1})=A_{k+1};

in particular, gg is topologically loxodromic. Let BkB_{k} be the locally flat annulus co-bounded by Σk\Sigma_{k} and Σk+1\Sigma_{k+1}. Then, as (gf)(Σk)=Σk+1(g\circ f)(\Sigma_{k})=\Sigma_{k+1}, we have (gf)(Bk)=Bk+1(g\circ f)(B_{k})=B_{k+1}. Therefore, gfg\circ f is topologically loxodromic. Moreover, gg and gfg\circ f share the same sink, and hence they are conjugate by 2.8. This establishes that ff can be expressed as a commutator. ∎

Theorem 3.2.

For nn\in\mathbb{N}, every element of Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}) is a commutator.

Proof.

Letting 𝕊0\mathbb{S}^{0} be a singleton, the proof in the annulus case with n=0n=0 shows that every element of Homeo0()\operatorname{Homeo}_{0}(\mathbb{R}) can be expressed as a commutator. We may now assume that n>1n>1.

Fix z𝕊nz\in\mathbb{S}^{n}, and identify Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}) with the stabilizer of zz in Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}). Let ff be an element of Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}) such that f(z)=zf(z)=z. We will show that ff can be expressed as a commutator of a pair of elements in Homeo0(𝕊n)\operatorname{Homeo}_{0}(\mathbb{S}^{n}) that each fix zz. The identity homeomorphism is clearly a commutator, so we may assume that ff is not the identity, and therefore there exists y𝕊ny\in\mathbb{S}^{n} such that f(y)yf(y)\neq y. By continuity, there exists a locally flat ball D0D_{0} centered at yy such that f(D0)D0=f(D_{0})\cap D_{0}=\varnothing. Choose a locally flat ball D1D_{1} centered at zz that is disjoint from D0f(D0)D_{0}\cup f(D_{0}). Again by continuity, by shrinking D1D_{1} if necessary, we may assume that f(D1)f(D_{1}) is also disjoint from D0f(D0)D_{0}\cup f(D_{0}).

First, in D1D_{1}, we proceed identically as we did in the annulus case. Choose a locally flat annulus A1A_{1} such that D1f(D1)\partial D_{1}\cup f(\partial D_{1}) is contained in the interior of A1A_{1} and such that A1A_{1} is disjoint from D0f(D0)D_{0}\cup f(D_{0}). Now choose a locally flat ball D2D_{2} centered at zz such that D2f(D2)D_{2}\cup f(D_{2}) is disjoint from A1A_{1}. Choose a locally flat annulus A2A_{2} such that A1A_{1} and A2A_{2} share a boundary component and D2f(D2)\partial D_{2}\cup f(\partial D_{2}) is contained in the interior of A2A_{2}. Continuing in this fashion, we build a sequence of locally flat annuli {Ak}k\{A_{k}\}_{k\in\mathbb{N}} and locally flat balls {Dk}k\{D_{k}\}_{k\in\mathbb{N}} such that AkA_{k} and Ak+1A_{k+1} share a boundary, Dkf(Dk)\partial D_{k}\cup f(\partial D_{k}) is contained in the interior of AkA_{k}, and kDk={z}\bigcap_{k\in\mathbb{N}}D_{k}=\{z\}. Note that the last condition is not a priori guaranteed, but at each stage we are free to choose DkD_{k} to have radius less than 1/k1/k, forcing the intersection of the DkD_{k} to be {z}\{z\}.

Let g0g_{0}^{\prime} be a topologically loxodromic homeomorphism of 𝕊n\mathbb{S}^{n} that fixes zz, that maps Ak+1A_{k+1} to AkA_{k} for kk\in\mathbb{N}, and that maps f(D1)f(\partial D_{1}) to D0\partial D_{0}. For k0k\leq 0, let Ak=(g0)(n+1)(A1)A_{k}=(g_{0}^{\prime})^{\circ(n+1)}(A_{1}). Then, (g0f)(Dk+1)(g_{0}^{\prime}\circ f)(\partial D_{k+1}) and Dk\partial D_{k} are locally flat spheres in the annulus AkA_{k}, and so we may choose hkHomeo(𝕊n)h_{k}\in\operatorname{Homeo}(\mathbb{S}^{n}) supported in the interior of AkA_{k} such that hk(g0(f(Dk+1)))=Dkh_{k}(g_{0}^{\prime}(f(\partial D_{k+1})))=\partial D_{k}. As the hkh_{k} have pairwise disjoint support, we can define h=khkh=\prod_{k\in\mathbb{N}}h_{k}. Let g0=hg0g_{0}=h\circ g_{0}^{\prime}. Then, for k{0}k\in\mathbb{N}\cup\{0\}, setting TkT_{k} to be the locally flat annulus bounded by Dk\partial D_{k} and Dk+1\partial D_{k+1}, we have (g0f)(Tk+1)=Tk(g_{0}\circ f)(T_{k+1})=T_{k}.

At this point, g0fg_{0}\circ f behaves like a topologically loxodromic homeomorphism when restricted to D1D_{1}, and so now we have to edit g0g_{0} in the complement of D1D_{1}. This portion of the argument is a version of Tsuboi’s argument for spheres. Let B0=g0(f(D0))B_{0}=g_{0}(f(D_{0})), and note that f(B0)f(D0)f(B_{0})\subset f(D_{0}) and g0(f(B0))B0g_{0}(f(B_{0}))\subset B_{0}. Moreover, observe that B0f(D0)A0B_{0}\cup f(D_{0})\subset A_{0}; hence, modifying g0g_{0} in either f(D0)f(D_{0}) or B0B_{0} will not change the fact that g0g_{0} is topologically loxodromic.

Fix xB0x\in B_{0}. We may assume that g0(f(x))=xg_{0}(f(x))=x; indeed, if not, then we may post compose g0g_{0} with a homeomorphism of 𝕊n\mathbb{S}^{n} supported in B0B_{0} that maps g0(f(x))g_{0}(f(x)) to xx. And, as just noted above, this edited version of g0g_{0} remains topologically loxodromic. We will now recursively edit g0g_{0} in f(D0)f(D_{0}) so that g0fg_{0}\circ f will be topologically loxodromic with zz its sink and xx its source.

Below, given a subset XX, we let XoX^{\mathrm{o}} denote its interior. Let Σ0=B0\Sigma_{0}=\partial B_{0}, let B0=f(B0)B_{0}^{\prime}=f(B_{0}), let B1=g0(B0)B_{1}=g_{0}(B_{0}^{\prime}), and let Σ1=B1\Sigma_{1}=\partial B_{1}. Note Σ0Σ1=\Sigma_{0}\cap\Sigma_{1}=\varnothing, and therefore Σ0\Sigma_{0} and Σ1\Sigma_{1} co-bound a locally flat annulus, which we label T1T_{-1}. Then, by continuity, we can choose a locally flat ball B1B_{1}^{\prime} in the interior of B0B_{0}^{\prime} centered at f(x)f(x) such that g0(B1)B1og_{0}(B_{1}^{\prime})\subset B_{1}^{\mathrm{o}}. We can then choose σ1Homeo(𝕊n)\sigma_{1}\in\operatorname{Homeo}(\mathbb{S}^{n}) such that σ1\sigma_{1} is supported in f(B0)f(B_{0}) and such that σ1(f(Σ1))=B1\sigma_{1}(f(\Sigma_{1}))=\partial B_{1}^{\prime}. Set B2=g0(B1)B_{2}=g_{0}(B_{1}^{\prime}), and set Σ2=B2\Sigma_{2}=\partial B_{2}. Now, set T2T_{-2} to be the locally flat annulus co-bounded by Σ1\Sigma_{1} and Σ2\Sigma_{2}, and set g1=g0σ1g_{1}=g_{0}\circ\sigma_{1}. Then, (g1f)(Tk)=Tk1(g_{1}\circ f)(T_{k})=T_{k-1} for all k{0,1}k\in\mathbb{N}\cup\{0,-1\}. Note that B1B_{1}^{\prime} can be chosen arbitrarily small.

Continuing in this fashion, for each mm\in\mathbb{N}, we obtain a locally flat annulus TmT_{-m} and a homeomorphism σm\sigma_{m} of 𝕊n\mathbb{S}^{n} supported in a ball of radius 1/m1/m centered at f(x)f(x) such that

  1. (1)

    𝕊n{x,z}=kTk\mathbb{S}^{n}\smallsetminus\{x,z\}=\bigcup_{k\in\mathbb{Z}}T_{k},

  2. (2)

    (gmf)(Tk)=Tk1(g_{m}\circ f)(T_{k})=T_{k-1} for every integer kmk\geq-m, and

  3. (3)

    gm=gm1σm=g0(σ1σ2σm)g_{m}=g_{m-1}\circ\sigma_{m}=g_{0}\circ(\sigma_{1}\circ\sigma_{2}\circ\cdots\circ\sigma_{m}).

These properties guarantee that limmgm\lim_{m\to\infty}g_{m} exists, call it gg, and that (gf)(Tk)=Tk1(g\circ f)(T_{k})=T_{k-1} for every kk\in\mathbb{Z}; hence, gfg\circ f is topologically loxodromic. Moreover, gg is topologically loxodromic as gg agrees with g0g_{0} outside of B0f(D0)B_{0}\cup f(D_{0}); in particular, g(Ak)=g0(Ak)=Ak+1g(A_{k})=g_{0}(A_{k})=A_{k+1} for all kk\in\mathbb{Z}. And, as gg and gfg\circ f share the same sink, they are conjugate in Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}) by 2.8. Thus, ff can be expressed as a commutator in Homeo0(n)\operatorname{Homeo}_{0}(\mathbb{R}^{n}). ∎

We finish by providing a proof of 2.

Proof of 2.

Without loss of generality, it is enough to prove the statement with pip_{i}\in\mathbb{N} for each i{1,,r}i\in\{1,\ldots,r\}, as we can always replace an element with its inverse in the decomposition to switch the sign of the exponent.

Fix gGg\in G. We have shown in the above proofs that there exist f,hGf,h\in G such that g=fhg=f\circ h and both ff and hh are topologically loxodromic. Applying the same techniques to ff, we can write f=f1f2f=f_{1}\circ f_{2} with both f1f_{1} and f2f_{2} topologically loxodromic. Therefore, continuing this splitting as many times as necessary, we can write g=i=1rfig=\prod_{i=1}^{r}f_{i} with fif_{i} topologically loxodromic. Now, every power of a topologically loxodromic homeomorphism is itself topologically loxodromic. Therefore, by 2.7 and 2.8, fipif_{i}^{\circ p_{i}} is conjugate to fif_{i}. Choose hih_{i} such that fi=hifipihi1f_{i}=h_{i}\circ f_{i}^{\circ p_{i}}\circ h_{i}^{-1}. Then, setting gi=hifihi1g_{i}=h_{i}\circ f_{i}\circ h_{i}^{-1}, we have fi=gipif_{i}=g_{i}^{\circ p_{i}}, and hence, g=i=1rgipig=\prod_{i=1}^{r}g_{i}^{\circ p_{i}}. ∎

References

  • [And58] R. D. Anderson, The algebraic simplicity of certain groups of homeomorphisms, Amer. J. Math. 80 (1958), 955–963. MR 98145
  • [BG64] Morton Brown and Herman Gluck, Stable structures on manifolds. I, II, III, Ann. of Math. (2) 79 (1964), 1–58. MR 158383
  • [BIP08] Dmitri Burago, Sergei Ivanov, and Leonid Polterovich, Conjugation-invariant norms on groups of geometric origin, Groups of diffeomorphisms, Adv. Stud. Pure Math., vol. 52, Math. Soc. Japan, Tokyo, 2008, pp. 221–250. MR 2509711
  • [BM12] Ara Basmajian and Bernard Maskit, Space form isometries as commutators and products of involutions, Trans. Amer. Math. Soc. 364 (2012), no. 9, 5015–5033. MR 2922617
  • [Bro60] Morton Brown, A proof of the generalized Schoenflies theorem, Bull. Amer. Math. Soc. 66 (1960), 74–76. MR 117695
  • [Bro62]  , Locally flat imbeddings of topological manifolds, Ann. of Math. (2) 75 (1962), 331–341. MR 133812
  • [EK71] Robert D. Edwards and Robion C. Kirby, Deformations of spaces of imbeddings, Ann. of Math. (2) 93 (1971), 63–88. MR 283802
  • [Kir69] Robion C. Kirby, Stable homeomorphisms and the annulus conjecture, Ann. of Math. (2) 89 (1969), 575–582. MR 242165
  • [Kup19] Alexander Kupers, Some finiteness results for groups of automorphisms of manifolds, Geom. Topol. 23 (2019), no. 5, 2277–2333. MR 4019894
  • [Lin80] Wensor Ling, The algebraic structure of geometric automorphism groups, ProQuest LLC, Ann Arbor, MI, 1980, Thesis (Ph.D.)–Princeton University. MR 2630795
  • [LRM18] Frédéric Le Roux and Kathryn Mann, Strong distortion in transformation groups, Bull. Lond. Math. Soc. 50 (2018), no. 1, 46–62. MR 3778543
  • [Man16a] Kathryn Mann, Automatic continuity for homeomorphism groups and applications, Geom. Topol. 20 (2016), no. 5, 3033–3056, With an appendix by Frédéric Le Roux and Mann. MR 3556355
  • [Man16b]  , A short proof that Diffc(M){\rm Diff}_{c}(M) is perfect, New York J. Math. 22 (2016), 49–55. MR 3484676
  • [Mil65] John Milnor, Lectures on the hh-cobordism theorem, Princeton University Press, Princeton, NJ, 1965, Notes by L. Siebenmann and J. Sondow. MR 190942
  • [Ore51] Oystein Ore, Some remarks on commutators, Proc. Amer. Math. Soc. 2 (1951), 307–314. MR 40298
  • [Qui82] Frank Quinn, Ends of maps. III. Dimensions 44 and 55, J. Differential Geometry 17 (1982), no. 3, 503–521. MR 679069
  • [Ryb11] Tomasz Rybicki, Boundedness of certain automorphism groups of an open manifold, Geom. Dedicata 151 (2011), 175–186. MR 2780744
  • [Sch11] Paul A. Schweitzer, Normal subgroups of diffeomorphism and homeomorphism groups of n\mathbb{R}^{n} and other open manifolds, Ergodic Theory Dynam. Systems 31 (2011), no. 6, 1835–1847. MR 2851677
  • [Seg09] Dan Segal, Words: notes on verbal width in groups, London Mathematical Society Lecture Note Series, vol. 361, Cambridge University Press, Cambridge, 2009. MR 2547644
  • [Sma62] S. Smale, On the structure of manifolds, Amer. J. Math. 84 (1962), 387–399. MR 153022
  • [Thu74] William Thurston, Foliations and groups of diffeomorphisms, Bull. Amer. Math. Soc. 80 (1974), 304–307. MR 339267
  • [Tsu13] Takashi Tsuboi, Homeomorphism groups of commutator width one, Proc. Amer. Math. Soc. 141 (2013), no. 5, 1839–1847. MR 3020870
  • [Vla24] Nicholas G. Vlamis, Homeomorphism groups of self-similar 2-manifolds, In the Tradition of Thurston III: Geometry and Dynamics (Ken’ichi Ohshika and Athanase Papadopoulos, eds.), Springer, 2024, pp. 105–167.
  • [WX17] Guozhen Wang and Zhouli Xu, The triviality of the 61-stem in the stable homotopy groups of spheres, Ann. of Math. (2) 186 (2017), no. 2, 501–580. MR 3702672