---
title: 'End-Periodic Homeomorphisms: Theory & Applications'
url: https://www.emergentmind.com/topics/end-periodic-homeomorphisms
type: topic
---

# End-Periodic Homeomorphisms: Theory & Applications

Searching arXiv for recent papers on end-periodic homeomorphisms and related mapping tori, laminations, and graph models.
End-periodic homeomorphisms are homeomorphisms of noncompact surfaces whose dynamics near every end is eventually monotone toward or away from that end. In the modern finitely-ended setting, one requires that there exists \(m>0\) such that for each end \(E\) there is a neighborhood \(U_E\) with either \(f^m(U_E)\subsetneq U_E\) and \(\{f^{nm}(U_E)\}_{n>0}\) a neighborhood basis of \(E\), or \(f^{-m}(U_E)\subsetneq U_E\) and \(\{f^{-nm}(U_E)\}_{n>0}\) a neighborhood basis of \(E\); the two cases define attracting and repelling ends, respectively [2306.03279]. The subject originated in unpublished work of Handel and Miller and was substantially developed by Cantwell, Conlon, and Fenley, where endperiodic automorphisms are analyzed through invariant laminations, reducing curves and lines, escaping sets, and applications to depth one foliations of \(3\)-manifolds [1006.4525]. Recent work places these maps in a quantitative framework involving compactified mapping tori, pants-graph translation length, short-curve theorems, loxodromic actions on infinite-type graphs, train-track analogues on infinite graphs, and arithmetic realizability of Handel–Miller stretch factors [2106.15642], [2306.03279], [2408.07044], [2211.00678], [2408.13401], [2603.20491].

## 1. Definitions at the ends

For surfaces with finitely many ends, all accumulated by genus, an end-periodic homeomorphism \(f\colon S\to S\) is organized by attracting and repelling end neighborhoods. If \(U_+\) and \(U_-\) denote unions of chosen neighborhoods of attracting and repelling ends and satisfy \(f^{\pm1}(U_\pm)\subset U_\pm\), with \(\partial\overline U_\pm\) a union of simple closed curves, then \(U_\pm\) are called tight nesting neighborhoods. A core for \(f\) is a compact subsurface \(Y\subset S\) such that \(S-Y=U_+\sqcup U_-\); the boundary components meeting \(U_+\) and \(U_-\) are the junctures \(\partial_+Y\) and \(\partial_-Y\). The escaping sets are
\[
\mathcal U_+ = \bigcup_{n \geq 0} f^{- n}(U_+),
\qquad
\mathcal U_- = \bigcup_{n \geq 0} f^{n}(U_-),
\]
and the \(\langle f\rangle\)-actions on \(\mathcal U_\pm\) are cocompact, with quotients \(S_\pm=\mathcal U_\pm/\langle f\rangle\) [2306.03279].

The older Handel–Miller formulation uses a noncompact surface \(L\) with finite endset. A periodic end \(e\) of period \(p_e\) is positive if there exists a closed, connected neighborhood \(U_e\) with \(L\setminus U_e\) connected, \(f^{p_e}(U_e)\subset U_e\), \(\bigcap_{n=0}^\infty f^{np_e}(U_e)=\emptyset\), and compact frontier; negative ends are defined by the same conditions for \(f^{-p_e}\). Such a \(U_e\) is an \(f\)-neighborhood, and \(J=\fr U_e\) is an \(f\)-juncture. In that language, \(f\) is endperiodic if all periodic ends are positive or negative, and \(f\) is endperiodic if and only if \(f^p\) is endperiodic for some \(p>0\) [1006.4525].

A later constructive paper uses the equivalent terminology of positive and negative ladders. A positive ladder \(U_+\) is a union of nesting neighborhoods of attracting ends, a negative ladder \(U_-\) is a union of nesting neighborhoods of repelling ends, and if these ladders are tight and disjoint then
\[
Y=\Sigma-(U_+\cup U_-)
\]
is a core for \(f\) [2603.20491].

## 2. Handel–Miller laminations and reduction theory

The foundational structural result is the existence, after isotopy, of a pair of invariant laminations \((\Lambda_+,\Lambda_-)\). In the geodesic version, one starts from positive and negative juncture sets and their geodesic tightenings, defines sets of nonescaping juncture components \(X_+\) and \(X_-\), and then forms
\[
G_\pm = \overline{|X_{\mp}|},
\qquad
L_\pm = G_\pm\setminus |X_{\mp}|,
\]
so that the path components of \(L_\pm\) are the laminations \(\Lambda_\pm\). The pairs \((\Gamma_+,\Gamma_-)\) and \((\Lambda_+,\Lambda_-)\) are bilaminations, the laminations are strongly closed, every leaf of \(\Lambda_\pm\) is a one-one immersed copy of \(\mathbb R\), and
\[
\fr U_{\mp}=|\Lambda_\pm|.
\]
The invariant set \(K=|\Lambda_+|\cap |\Lambda_-|\) carries the recurrent dynamics, and the core dynamical system
\[
h:K\to K
\]
is topologically conjugate to a two-ended Markov shift of finite type [1006.4525].

These laminations support a reduction theory parallel in spirit to Nielsen–Thurston theory but adapted to ends. If \(\Lambda_+=\Lambda_-=\emptyset\), then \(f\) is isotopic to a translation. Otherwise one constructs a set \(\mathfrak S\) of reducing curves consisting of rims of crown sets, reducing circles arising from compact border components of the escaping set, and reducing lines associated to noncompact border components. After isotopy, there is an endperiodic automorphism \(g\) that agrees with the Handel–Miller representative on \(|\Lambda_\pm|\), permutes \(\mathfrak S\), and on each \(g\)-periodic noncompact reduced piece \(Q\) the return map \(g^m:Q\to Q\) is isotopic either to a translation or to a pseudo-anosov automorphism. Compact reduced pieces in principal regions are governed by Nielsen–Thurston theory, while compact pieces in the escaping set have trivial dynamics under iteration. The same paper also develops an axiomatic pseudo-geodesic bilamination theory, proves that any such bilamination is ambiently isotopic to the geodesic one, proves a smoothing theorem producing a smooth endperiodic automorphism preserving smooth laminations, and proves the transfer theorem for depth one foliations transverse to a common one-dimensional foliation [1006.4525].

A persistent misconception is that end-periodic dynamics merely describe behavior “at infinity.” Handel–Miller theory shows instead that the nonescaping set supports a finite-type symbolic dynamics, while the escaping sets, principal regions, semi-isolated leaves, arms, nuclei, and crown sets control how that recurrent core is attached to the ends [1006.4525].

## 3. Strong irreducibility and compactified mapping tori

In the finitely-ended, genus-accumulated setting, recent work isolates a stronger irreducibility hypothesis tailored to \(3\)-manifold geometry. A curve \(\alpha\) is reducing for \(f\) if there exist \(m<n\) such that \(f^n(\alpha)\) lies in a nesting neighborhood of an attracting end and \(f^m(\alpha)\) lies in a nesting neighborhood of a repelling end. An end-periodic homeomorphism is strongly irreducible if it has no periodic curves, no periodic lines, and no reducing curves [2306.03279].

The ordinary mapping torus
\[
M_f=S\times[0,1]/(x,1)\sim(f(x),0)
\]
is noncompact because the fiber is noncompact. The compactification is built upstairs in the infinite cyclic cover by
\[
\widetilde M_\infty
=
\{(x,t)\in S\times[-\infty,\infty]\mid x\in \mathcal U_\pm \text{ if } t=\pm\infty\},
\]
with deck transformation
\[
F(x,t)=(f(x),t-1),
\]
where \(\pm\infty-1=\pm\infty\). The compactified mapping torus is
\[
\overline M_f=\widetilde M_\infty/\langle F\rangle,
\]
its interior is \(M_f\), and its boundary is naturally homeomorphic to
\[
\partial \overline M_f \cong S_- \cup S_+.
\]
The action of \(\langle F\rangle\) on \(\widetilde M_\infty\) is properly discontinuous and cocompact, so the quotient is compact [2306.03279].

Under strong irreducibility, \(\overline M_f\) is a compact, irreducible, atoroidal, acylindrical \(3\)-manifold with incompressible boundary, and therefore admits a convex hyperbolic metric with totally geodesic boundary, unique up to isometry. In this setting, strong irreducibility is not merely technical. If periodic curves or lines exist, the compactified mapping torus need not be atoroidal or acylindrical; if reducing curves exist, the core dynamics decomposes and the uniform topological control used in the quantitative theory breaks down [2306.03279].

## 4. Capacity, pants graphs, short curves, and volume

Because the surface is of infinite type, constants cannot depend on the topology of the whole fiber in the finite-type manner of Brock’s theorem. The replacement is the capacity of \(f\), the pair \((\chi(f),\xi(f))\), where
\[
\chi(f)=\max_{Y\subset S}\chi(Y)
\]
with the maximum taken over all cores \(Y\subset S\), and
\[
\xi(f)=\xi(\partial \overline M_f)=\xi(S_+)+\xi(S_-)=2\xi(S_+).
\]
The pants graph \(\mathcal P(S)\) is defined with edge lengths \(1\) on one-holed torus moves and \(2\) on four-holed sphere moves, but for infinite-type \(S\) it is disconnected. The asymptotic translation distance is therefore
\[
\tau(f) = \inf_{P\in\mathcal{P}(S)} \liminf_{n \to \infty} \frac{d(P, f^n(P))}n,
\]
and for an \(f\)-invariant component \(\Omega\subset\mathcal P(S)\),
\[
\tau_\Omega(f) = \inf_{P \in \Omega} \lim_{k \to \infty} \frac{d_{\mathcal P}(P,f^k(P))}{k}.
\]
The central lower bound is
\[
C_2\tau(f)\le (\overline M_f),
\]
where \(C_2\) depends only on the capacity of \(f\) [2306.03279].

Earlier work proved the companion upper bound
\[
(\overline M_f)\le V_{oct}\,\tau(f),
\]
and the stronger componentwise estimate
\[
(\overline M_f-P_\Omega)\le V_{oct}\,\tau(f,\Omega).
\]
Together these give
\[
C_2\tau(f)\le (\overline M_f)\le C_1\tau(f),
\]
with constants depending only on capacity, extending Brock’s finite-type pseudo-Anosov theorem to strongly irreducible end-periodic homeomorphisms. The same program also produces invariant pants-graph components whose induced boundary pants decompositions have uniformly bounded length and whose translation lengths are within a uniform multiplicative constant of the global asymptotic translation length [2106.15642], [2306.03279].

A related quantitative theorem compares Handel–Miller laminations to short curves in hyperbolic structures. If \(\Lambda^+\) and \(\Lambda^-\) are the positive and negative Handel–Miller laminations, then for any \(D,\varepsilon>0\) there exists \(K=K(D,\epsilon)\) such that for any atoroidal end-periodic \(f:S\to S\) with
\[
\capgen(f)=\chi(f)+\xi(f)^2\le D,
\]
there exists \(s\in \AH(M_f)\) with the property that for any connected, compact subsurface \(Y\subset S\subset M_f\),
\[
d_Y(\Lambda^+,\Lambda^-)\ge K \Rightarrow \ell_s(\partial Y)\le \varepsilon.
\]
This is an end-periodic analogue of Minsky’s short-curve theorem from the finite-type pseudo-Anosov setting [2408.07044].

## 5. Loxodromic actions and graph-theoretic analogues

End-periodic homeomorphisms also act on several infinite-type arc and curve graphs. One construction starts with an endperiodic map \(t\colon S\to S\), a finite-type witness subsurface \(Y\), and a mapping class \(g\in Mod(Y)\), and sets
\[
f=tg.
\]
Under explicit hypotheses on the boundary behavior of \(Y\) under \(t\) and the projection-distance condition
\[
d_Y(t(\partial_-Y),t^{-1}(\partial_+Y))+61\le \ell_Y(g),
\]
the map \(f\) is again endperiodic and acts loxodromically on the ambient graph \(A(S)\). The same method produces loxodromic actions on the relative arc graph, the omnipresent arc graph and grand arc graph, and separating curve graphs. In the examples emphasized there, the constructed \(f=tg\) is strongly irreducible in the sense of Field–Kim–Leininger–Loving [2211.00678].

A distinct line of work builds one-dimensional analogues on infinite graphs with finitely many ends. A generalized endperiodic graph map is a cellular homotopy equivalence \(f\colon G\to G\) for which each end is attracting or repelling; it is endperiodic if the restrictions to \(f\)-neighborhoods of ends are homeomorphisms that send edges to edges. Adapting Bestvina–Handel theory, every generalized endperiodic map is conjugate to a generalized endperiodic relative train track map via a combinatorially bounded homotopy equivalence. If \(F\) is such a representative, its largest Perron–Frobenius eigenvalue \(\lambda(F)\) is canonical, characterized by
\[
\log(\lambda)=\sup_{G'\subseteq G\text{ finite},~ \mathfrak{c} \text{ conjugacy class of }\pi_1(G)}
\limsup_{n\to\infty}
\frac{\log\bigl(l_w(P_{G'}(f_*^n(\mathfrak{c})))\bigr)}{n},
\]
and when \(\lambda(F)\ge 1\),
\[
h_{\rm top}(F)=\log \lambda(F).
\]
Relative train track representatives minimize both \(\lambda\) and topological entropy in the proper homotopy class. The authors are explicit that this graph-theoretic framework is a one-dimensional analogue of the surface theory and do not claim equivalence with Handel–Miller laminational train tracks [2408.13401].

The graph-mapping-torus analogue is equally close. If \(\Gamma\) is an infinite connected graph with finitely many ends and \(g:\Gamma\to\Gamma\) is end-periodic, then the mapping torus \(Z_g\) is homotopy equivalent to a finite \(2\)-complex \(W_g\) via a flow-preserving embedding, with disjoint \(1\)-subcomplexes \({}_+W_g\) and \({}_-W_g\) such that
\[
Z_g \cong W_g - ({}_+W_g\cup {}_-W_g).
\]
With additional hypotheses, the compactified mapping torus embeds in the mapping torus of a homotopy equivalence of a finite graph via a \(\pi_1\)-injective, flow-preserving map [2511.14976].

## 6. Stretch factors, realizability, and broader context

The arithmetic range of end-periodic stretch factors is now known exactly. For an end-periodic homeomorphism \(f\), the intersection \(\Lambda^+\cap \Lambda^-\) determines a Markov decomposition of the complement of the escaping points, and the Handel–Miller stretch factor \(\lambda(f)\) is the spectral radius of the corresponding incidence matrix, equivalently the exponential of the topological entropy on the action of \(f\) on \(\Lambda^+\cap\Lambda^-\). The realizability theorem states:
\[
\lambda \text{ is the stretch factor of an end-periodic homeomorphism } f:\Sigma\to\Sigma
\text{ if and only if } \lambda \text{ is a weak Perron number.}
\]
In particular, given any weak Perron number \(\lambda\), there is a connected infinite-type surface with finitely many ends all accumulated by genus and an end-periodic homeomorphism \(f:\Sigma\to\Sigma\) whose Handel–Miller stretch factor equals \(\lambda\) [2603.20491].

This arithmetic picture differs sharply from the finite-type pseudo-Anosov case. The same paper emphasizes that every pseudo-Anosov stretch factor is bi-Perron, while realizability of all bi-Perron numbers is open, whereas in the end-periodic category realizability is completely settled: exactly weak Perron numbers occur [2603.20491].

The broader context remains foliation theory. In the smooth depth one setting, the monodromy on noncompact leaves is endperiodic, and the transfer theorem shows that if two depth one foliations are transverse to a common one-dimensional foliation and the induced monodromy on one leaf preserves a pseudo-geodesic bilamination satisfying the Handel–Miller axioms, then the induced monodromy on the other leaf does as well. In more recent \(3\)-manifold work, compactified mapping tori of end-periodic homeomorphisms appear as the pieces obtained after collapsing trivial pieces of a co-oriented depth-one foliation of a \(3\)-manifold [1006.4525].

End-periodic homeomorphisms therefore occupy a position analogous, but not identical, to pseudo-Anosov homeomorphisms. Their invariant laminations replace stable and unstable laminations only after one incorporates escaping sets and end neighborhoods; their mapping tori become canonical hyperbolic objects only after compactification; and their quantitative theory depends not on the topology of a compact fiber but on capacity, end complexity, and related finite invariants extracted from the dynamically nontrivial core.

Source: https://www.emergentmind.com/topics/end-periodic-homeomorphisms