Anosov Mapping Family in Hyperbolic Dynamics

• Anosov Mapping Family is a class of hyperbolic systems characterized by pseudo-Anosov maps that preserve stable and unstable measured foliations with a scaling dilatation λ > 1.
• It spans both stationary dynamics on finite-type surfaces, where train tracks and Nielsen–Thurston theory govern behavior, and non-stationary sequences of diffeomorphisms with uniform hyperbolicity.
• Explicit constructions in this family link mapping tori, Sol geometric structures, and analytic representation theories, offering precise control over dilatation and geometric transitions.

“Anosov Mapping Family” is used across several distinct but closely related settings in surface and hyperbolic dynamics. In the finite-type surface setting, the central objects are pseudo-Anosov mapping classes, which preserve transverse invariant measured foliations and scale them by a factor $\lambda>1$; in that sense, pseudo-Anosov maps are the surface analogue of “Anosov” behavior (Cumplido et al., 2017, Kozai, 2014). The same phrase also arises in work on explicit families of pseudo-Anosov maps with controlled dilatation, on mapping tori and representation varieties attached to pseudo-Anosov monodromy, on measurable or infinite-type pseudo-Anosov-like systems, and, in a formally different direction, on non-stationary dynamical systems consisting of two-sided sequences of diffeomorphisms with invariant hyperbolic splittings (Acevedo, 2017, Acevedo, 2017).

1. Conceptual scope and basic structures

For a pseudo-Anosov homeomorphism $\phi:S\to S$, the basic invariant data are the stable and unstable measured foliations $(\mathcal F^s,\mu_s)$ and $(\mathcal F^u,\mu_u)$, together with a dilatation $\lambda>1$ such that

$$\phi(\mathcal F^s)=\mathcal F^s,\qquad \phi(\mathcal F^u)=\mathcal F^u,$$

and

$$\phi^*\mu_s=\lambda^{-1}\mu_s,\qquad \phi^*\mu_u=\lambda\,\mu_u.$$

These foliations induce a singular Euclidean metric on the fiber, and for the mapping torus $M_\phi$ they determine a natural singular Sol structure with metric

$ds^2 = e^{2z}dx^2 + e^{-2z}dy^2 + dz^2.$

This is the standard geometric mechanism by which pseudo-Anosov dynamics becomes a three-dimensional geometric structure (Kozai, 2014).

A different formalization appears in the theory of Anosov families. There one considers a two-sided sequence

$f=(f_i)_{i\in\mathbb Z},\qquad f_i:M_i\to M_{i+1},$

on a disjoint union $\phi:S\to S$0 of compact Riemannian manifolds. The family is Anosov if the tangent bundle admits a continuous $\phi:S\to S$1-invariant splitting

$\phi:S\to S$2

and there exist constants $\phi:S\to S$3 and $\phi:S\to S$4 such that for all $\phi:S\to S$5, $\phi:S\to S$6, and $\phi:S\to S$7,

$\phi:S\to S$8

$\phi:S\to S$9

If one can take $(\mathcal F^s,\mu_s)$0, the family is strictly Anosov (Acevedo, 2017, Acevedo, 2017).

A common source of ambiguity is that these two usages are not identical. Pseudo-Anosov mapping families live in mapping class groups and are organized by Nielsen–Thurston dynamics, train tracks, or monodromy, whereas Anosov families in the non-stationary sense are sequences of diffeomorphisms on varying manifolds with uniform hyperbolicity in time (Cumplido et al., 2017, Acevedo, 2017).

2. Abundance in mapping class groups

One major line of work studies pseudo-Anosov maps as a large-scale family inside mapping class groups. For a closed surface $(\mathcal F^s,\mu_s)$1 of genus at least $(\mathcal F^s,\mu_s)$2, with mapping class group $(\mathcal F^s,\mu_s)$3, finite generating set $(\mathcal F^s,\mu_s)$4, Cayley graph $(\mathcal F^s,\mu_s)$5, and ball $(\mathcal F^s,\mu_s)$6, a standard conjecture asserts that the proportion of pseudo-Anosov elements in $(\mathcal F^s,\mu_s)$7 tends to $(\mathcal F^s,\mu_s)$8 as $(\mathcal F^s,\mu_s)$9. A robust weaker result is that this proportion stays bounded away from zero. The key theorem states that there exists a finite subset $(\mathcal F^u,\mu_u)$0 such that for every $(\mathcal F^u,\mu_u)$1, at least one element of $(\mathcal F^u,\mu_u)$2 is pseudo-Anosov. Consequently,

$(\mathcal F^u,\mu_u)$3

and an explicit lower bound is obtained from the finite correction radius $(\mathcal F^u,\mu_u)$4: $(\mathcal F^u,\mu_u)$5 The proof combines Fathi’s theorem on Dehn twists with Bowditch’s acylindricity of the action on the curve complex, and it extends to subgroups satisfying condition $(\mathcal F^u,\mu_u)$6, including finite index subgroups of $(\mathcal F^u,\mu_u)$7, the Torelli group, and finite index subgroups of the Torelli group (Cumplido et al., 2017).

A more local construction starts with a fixed mapping class $(\mathcal F^u,\mu_u)$8, a curve $(\mathcal F^u,\mu_u)$9 satisfying $\lambda>1$0, and a pure mapping class $\lambda>1$1. In this setting, there is a finite exceptional set $\lambda>1$2 such that if $\lambda>1$3, then

$\lambda>1$4

Here $\lambda>1$5 comes from the Bounded Geodesic Image Theorem, and $\lambda>1$6 satisfies

$\lambda>1$7

The resulting pseudo-Anosov stable lengths are controlled by

$\lambda>1$8

This makes the pseudo-Anosov family effective in terms of subsurface projections and loxodromic domains (Watanabe, 2016).

Generation results show that pseudo-Anosov elements can also form very small algebraic families. The mapping class group of a closed oriented surface is generated by two pseudo-Anosov elements, and if the genus is at least $\lambda>1$9, it is generated by two conjugate pseudo-Anosov elements with arbitrarily large dilatations. The same work gives analogous statements for the hyperelliptic mapping class group, and also parallel constructions with reducible but non-periodic generators (Hirose et al., 2023).

3. Explicit pseudo-Anosov families and dilatation asymptotics

Several papers construct infinite families with explicit control of the stretch factor. One such family $$\phi(\mathcal F^s)=\mathcal F^s,\qquad \phi(\mathcal F^u)=\mathcal F^u,$$0, built using fat train track maps and a train track automaton, has dilatations

$$\phi(\mathcal F^s)=\mathcal F^s,\qquad \phi(\mathcal F^u)=\mathcal F^u,$$1

where

$$\phi(\mathcal F^s)=\mathcal F^s,\qquad \phi(\mathcal F^u)=\mathcal F^u,$$2

For this family, $$\phi(\mathcal F^s)=\mathcal F^s,\qquad \phi(\mathcal F^u)=\mathcal F^u,$$3 is closed, orientable, and has genus

$$\phi(\mathcal F^s)=\mathcal F^s,\qquad \phi(\mathcal F^u)=\mathcal F^u,$$4

Its normalized dilatations converge to the golden-ratio limit, with

$$\phi(\mathcal F^s)=\mathcal F^s,\qquad \phi(\mathcal F^u)=\mathcal F^u,$$5

and the family is orientable if and only if $$\phi(\mathcal F^s)=\mathcal F^s,\qquad \phi(\mathcal F^u)=\mathcal F^u,$$6 is even. The folding decomposition corresponds to a length-$$\phi(\mathcal F^s)=\mathcal F^s,\qquad \phi(\mathcal F^u)=\mathcal F^u,$$7 circuit in the fat train track automaton, and the associated digraph type is fixed for all $$\phi(\mathcal F^s)=\mathcal F^s,\qquad \phi(\mathcal F^u)=\mathcal F^u,$$8 (Hironaka, 2014).

A broader asymptotic family is provided by Penner sequences. Starting from a stacked surface $$\phi(\mathcal F^s)=\mathcal F^s,\qquad \phi(\mathcal F^u)=\mathcal F^u,$$9, a shift $$\phi^*\mu_s=\lambda^{-1}\mu_s,\qquad \phi^*\mu_u=\lambda\,\mu_u.$$0, multicurves $$\phi^*\mu_s=\lambda^{-1}\mu_s,\qquad \phi^*\mu_u=\lambda\,\mu_u.$$1 and $$\phi^*\mu_s=\lambda^{-1}\mu_s,\qquad \phi^*\mu_u=\lambda\,\mu_u.$$2, and a connecting curve $$\phi^*\mu_s=\lambda^{-1}\mu_s,\qquad \phi^*\mu_u=\lambda\,\mu_u.$$3, one defines

$$\phi^*\mu_s=\lambda^{-1}\mu_s,\qquad \phi^*\mu_u=\lambda\,\mu_u.$$4

These $$\phi^*\mu_s=\lambda^{-1}\mu_s,\qquad \phi^*\mu_u=\lambda\,\mu_u.$$5 are pseudo-Anosov, and along every rational ray $$\phi^*\mu_s=\lambda^{-1}\mu_s,\qquad \phi^*\mu_u=\lambda\,\mu_u.$$6 the minimal dilatations satisfy

$$\phi^*\mu_s=\lambda^{-1}\mu_s,\qquad \phi^*\mu_u=\lambda\,\mu_u.$$7

The proof uses Penner’s lower bound together with train-track/Perron–Frobenius estimates giving

$$\phi^*\mu_s=\lambda^{-1}\mu_s,\qquad \phi^*\mu_u=\lambda\,\mu_u.$$8

while $$\phi^*\mu_s=\lambda^{-1}\mu_s,\qquad \phi^*\mu_u=\lambda\,\mu_u.$$9 grows linearly in $M_\phi$0 (Valdivia, 2010).

A homologically constrained family is obtained by fixing the dimension $M_\phi$1 of the subspace of $M_\phi$2 fixed by a pseudo-Anosov $M_\phi$3. Defining

$M_\phi$4

one has the explicit comparison

$M_\phi$5

so $M_\phi$6. The corresponding mapping tori satisfy

$M_\phi$7

and the number of conjugacy classes with this constraint grows like a polynomial of degree $M_\phi$8 in $M_\phi$9 (Agol et al., 2014).

Another explicit organization is combinatorial rather than asymptotic. For a family of pseudo-Anosov homeomorphisms $ds^2 = e^{2z}dx^2 + e^{-2z}dy^2 + dz^2.$0 arising from postcritically finite zig-zag interval maps, there is a bijection

$ds^2 = e^{2z}dx^2 + e^{-2z}dy^2 + dz^2.$1

and

$ds^2 = e^{2z}dx^2 + e^{-2z}dy^2 + dz^2.$2

The rationals are arranged in a Farey tree: compatible rationals $ds^2 = e^{2z}dx^2 + e^{-2z}dy^2 + dz^2.$3 and $ds^2 = e^{2z}dx^2 + e^{-2z}dy^2 + dz^2.$4 satisfy $ds^2 = e^{2z}dx^2 + e^{-2z}dy^2 + dz^2.$5, and their Farey sum is

$ds^2 = e^{2z}dx^2 + e^{-2z}dy^2 + dz^2.$6

Within this family,

$ds^2 = e^{2z}dx^2 + e^{-2z}dy^2 + dz^2.$7

so the rational parameter orders the stretch factors monotonically (Farber, 2023).

4. Mapping tori, geometric transitions, and analytic representation families

Pseudo-Anosov mapping families also organize geometric structures on $ds^2 = e^{2z}dx^2 + e^{-2z}dy^2 + dz^2.$8-manifolds. For the mapping torus

$ds^2 = e^{2z}dx^2 + e^{-2z}dy^2 + dz^2.$9

the invariant measured foliations of a pseudo-Anosov $f=(f_i)_{i\in\mathbb Z},\qquad f_i:M_i\to M_{i+1},$0 induce a natural singular Sol structure. If the stable and unstable foliations are orientable and the induced action

$f=(f_i)_{i\in\mathbb Z},\qquad f_i:M_i\to M_{i+1},$1

does not have $f=(f_i)_{i\in\mathbb Z},\qquad f_i:M_i\to M_{i+1},$2 as an eigenvalue, then the singular Sol structure on $f=(f_i)_{i\in\mathbb Z},\qquad f_i:M_i\to M_{i+1},$3 deforms into a family of nearby singular hyperbolic structures with cone singularities along the singular locus $f=(f_i)_{i\in\mathbb Z},\qquad f_i:M_i\to M_{i+1},$4. These hyperbolic structures degenerate to a transversely hyperbolic foliation and, after rescaling, limit to the Sol structure as projective structures. The cone angles may be chosen to decrease from multiples of $f=(f_i)_{i\in\mathbb Z},\qquad f_i:M_i\to M_{i+1},$5 (Kozai, 2014).

The same pseudo-Anosov family viewpoint appears in quantum and analytic representation theory. For a punctured surface $f=(f_i)_{i\in\mathbb Z},\qquad f_i:M_i\to M_{i+1},$6 with negative Euler characteristic, there is an analytic family of representations

$f=(f_i)_{i\in\mathbb Z},\qquad f_i:M_i\to M_{i+1},$7

of the mapping class group acting on $f=(f_i)_{i\in\mathbb Z},\qquad f_i:M_i\to M_{i+1},$8, where $f=(f_i)_{i\in\mathbb Z},\qquad f_i:M_i\to M_{i+1},$9. The family interpolates between the multicurve representation at $\phi:S\to S$00 and the $\phi:S\to S$01-character variety representation at $\phi:S\to S$02. It is bounded for $\phi:S\to S$03, unitary for $\phi:S\to S$04, densely defined on the unit circle away from roots of unity, and finite-dimensional at roots of unity distinct from $\phi:S\to S$05, where it identifies with the “Hom” version of the Reshetikhin–Turaev/BHMV TQFT representation (Costantino et al., 2012).

For pseudo-Anosov $\phi:S\to S$06 with dilatation $\phi:S\to S$07, this representation family gives an explicit level estimate for quantum detection: if $\phi:S\to S$08 is a primitive $\phi:S\to S$09 root of unity and

$\phi:S\to S$10

then

$\phi:S\to S$11

This estimate is derived from train-track data, a Perron–Frobenius incidence matrix, the Ham–Song estimate, and a triangulation-based detection lemma (Costantino et al., 2012).

5. Generalized, measurable, and infinite-type families

Classical pseudo-Anosov families can be weakened while retaining part of the expanding/contracting architecture. A continuously varying family

$\phi:S\to S$12

provides measurable pseudo-Anosov maps on the sphere. Each $\phi:S\to S$13 is semi-conjugate to the inverse limit of the core tent map of slope $\phi:S\to S$14, via

$\phi:S\to S$15

The invariant geometry is a pair of transverse full-measure measured turbulations $\phi:S\to S$16 and $\phi:S\to S$17, scaled by

$\phi:S\to S$18

Every $\phi:S\to S$19 is topologically transitive, ergodic with respect to a background Oxtoby–Ulam measure, has dense periodic points, and satisfies

$\phi:S\to S$20

For a dense $\phi:S\to S$21, full-measure set of parameters, $\phi:S\to S$22 is measurable pseudo-Anosov but not generalized pseudo-Anosov, and its turbulations are nowhere locally regular (Boyland et al., 2023).

Infinite-type analogues arise on the infinite Jacob’s ladder surface $\phi:S\to S$23, with regular infinite-sheeted cover

$\phi:S\to S$24

and deck group generated by the handle shift $\phi:S\to S$25. Penner-type pseudo-Anosov maps on $\phi:S\to S$26 lift to “pseudo-Anosov-like” maps $\phi:S\to S$27 on $\phi:S\to S$28. The lifted foliations remain invariant and satisfy

$\phi:S\to S$29

After passing to the commuting lift, the dynamics admits a countable Markov partition and a finite-to-one semiconjugacy to a countable Markov shift whose transition matrix has block-tridiagonal form

$\phi:S\to S$30

These lifts are topologically transitive, topologically mixing, ergodic, and null recurrent with respect to a natural $\phi:S\to S$31-finite invariant measure (Agarwal et al., 24 Aug 2025).

A different deformation scheme starts from any pseudo-Anosov map $\phi:S\to S$32 on a surface of genus $\phi:S\to S$33 and produces a family “derived from pseudo-Anosov” by perturbing $\phi:S\to S$34 near fixed points. The resulting map $\phi:S\to S$35 has a nontrivial invariant compact set $\phi:S\to S$36, hyperbolic on $\phi:S\to S$37, and an attracting basin outside $\phi:S\to S$38. There is a unique invariant measure $\phi:S\to S$39, supported on $\phi:S\to S$40, with respect to which $\phi:S\to S$41 is mixing. The stable vector field integrates to a renormalized flow $\phi:S\to S$42 satisfying

$\phi:S\to S$43

and the Poincaré return map on a transversal is a uniquely ergodic generalized interval exchange transformation with a wandering interval, semi-conjugated to a self-similar interval exchange transformation. When $\phi:S\to S$44 is $\phi:S\to S$45, both the flow and the GIET are $\phi:S\to S$46 (Carrand, 2021).

The asymptotic geometry of such families can also be encoded in continuum theory. For a pseudo-Anosov map on a compact orientable surface, the second-order Hausdorff metric $\phi:S\to S$47 on the hyperspace of subcontinua allows one to identify the stable foliation with a hypercontinuum $\phi:S\to S$48. In the non-negative curvature case,

$\phi:S\to S$49

and for every nontrivial stable arc $\phi:S\to S$50,

$\phi:S\to S$51

For arbitrary curvature, the correct object is the closure of the accessible arcs $\phi:S\to S$52. Under an analogous condition for cw-expansive homeomorphisms, topological mixing and density of stable continua follow (Artigue, 2015).

6. Non-stationary Anosov families and their dynamical infrastructure

In the non-autonomous setting, the theory of Anosov families develops a direct analogue of classical hyperbolic dynamics. The set of Anosov families is open in the space $\phi:S\to S$53 of two-sided sequences of $\phi:S\to S$54-diffeomorphisms equipped with the strong (Whitney) topology. This openness uses invariant cone conditions, continuity of the splitting, and adapted metrics making the family strictly Anosov. The notion depends on the chosen Riemannian metrics on the components $\phi:S\to S$55, but it is invariant under uniformly equivalent metrics on the total space (Acevedo, 2017).

Local invariant manifolds also persist. For $\phi:S\to S$56 and a sequence of radii $\phi:S\to S$57, the local stable and unstable sets are defined by

$\phi:S\to S$58

$\phi:S\to S$59

A graph-transform construction yields local unstable manifolds $\phi:S\to S$60 and local stable manifolds $\phi:S\to S$61, tangent respectively to $\phi:S\to S$62 and $\phi:S\to S$63, with the expected invariance and contraction properties. Under additional compatibility conditions, these admissible manifolds coincide with the dynamical sets $\phi:S\to S$64 and $\phi:S\to S$65, and depend continuously on the base point (Acevedo, 2017).

For the class

$\phi:S\to S$66

the standard hyperbolic toolkit survives in full. Such families admit canonical coordinates, are expansive, satisfy the shadowing property, and exhibit Markov partitions. Canonical coordinates mean that for sufficiently small $\phi:S\to S$67 there exists $\phi:S\to S$68 such that $\phi:S\to S$69 implies

$\phi:S\to S$70

Expansivity gives a uniform $\phi:S\to S$71 separating distinct bi-infinite orbits; shadowing asserts that every sufficiently small pseudo-orbit is uniquely shadowed by a true orbit; and the Markov partition theorem provides a sequence of proper rectangles with the usual stable/unstable boundary relations (Acevedo et al., 2020).

Taken together, these results show that “Anosov Mapping Family” names not a single theorem or object but a coherent research domain organized by hyperbolicity. On finite-type surfaces the dominant theme is the construction, prevalence, and deformation of pseudo-Anosov families; in non-autonomous dynamics it is the persistence of invariant splitting, local product structure, shadowing, and symbolic coding for sequences of maps (Cumplido et al., 2017, Acevedo, 2017, Acevedo et al., 2020).