Non-Uniformly Expanding Maps

Updated 29 January 2026

Non-uniformly expanding maps are dynamical systems characterized by average expansion punctuated by intermittent critical or indifferent regions.
Research employs inducing schemes and hyperbolic times to rigorously analyze mixing rates, equilibrium states, and limit laws in these complex systems.
Advanced thermodynamic formalism and symbolic coding techniques reveal crucial insights into multifractal spectra, stochastic stability, and chaotic dynamics.

Non-uniformly expanding maps are dynamical systems, typically continuous or piecewise-smooth transformations on manifolds or intervals, for which typical orbits undergo expansion on average, but may experience regions of indifferent or critical behavior where uniform expansion fails. Such maps frequently arise in one-dimensional dynamics, higher-dimensional systems with critical sets, and random or skew-product settings, and they underpin contemporary developments in thermodynamic formalism, multifractal analysis, equilibrium states, and statistical limit laws. The class is broad, encompassing intermittent maps (e.g., Pomeau–Manneville, Rényi maps), Viana skew-products, circle maps with logarithmic singularities, and non-uniformly expanding local diffeomorphisms—often with complex critical/singular structures.

1. Foundational Definitions and Classes

A map $f:M \to M$ (on a compact Riemannian manifold or interval $M$ ) is termed non-uniformly expanding if almost every orbit satisfies a positive lower bound for the time-average of the logarithm of the norm of the derivative:

$\liminf_{n \to \infty} \frac{1}{n} \sum_{j=0}^{n-1} \log \| Df(f^{j}(x)) \| \geq \lambda > 0.$

In random dynamical systems, the analogous condition is formulated for random compositions $f^n_{\omega} = f_{\omega_{n-1}} \circ \cdots \circ f_{\omega_0}$ , requiring

$\liminf_{n \to \infty} \frac{1}{n} \sum_{j=0}^{n-1} \log \| D f_{\omega_j}(f^j_\omega(x)) \| \geq \lambda.$

Maps may have critical or singular sets $C \subset M$ , with slow recurrence: the measure of points not staying $\delta$ -away from $C$ decays (stretched) exponentially, enforcing that non-uniform expansion dominates statistically for typical orbits (Li et al., 2014, Castro et al., 2012, Araujo et al., 2011).

Key subclasses:

Interval maps with indifferent fixed points: Pomeau–Manneville, Rényi, backward continued fraction maps (Bahsoun et al., 2013, Arima, 4 Nov 2025).
Piecewise expanding maps (finite or countable branches): With finitely or countably many intervals and uniformly expanding away from parabolic or indifferent points (Arima, 19 Oct 2025, Arima, 4 Nov 2025).
Multimodal or unimodal interval maps: Logistic, quadratic family with non-flat critical points and recurrence conditions (Shen, 2011, Araujo et al., 2013).
Higher-dimensional local diffeomorphisms: Viana maps on $S^1 \times I$ , non-uniformly expanding toral maps (Santana, 2021, Li et al., 2014).

2. Inducing Schemes, Hyperbolic Times, and Young Towers

Uniform control of statistical properties for non-uniformly expanding maps is achieved via inducing schemes (first-return maps) and hyperbolic times. A point $x$ has a $(\sigma,\delta)$ -hyperbolic time $n$ if a neighborhood $V_n(x)$ is mapped diffeomorphically with expansion and bounded distortion onto a ball of radius $\delta$ :

$d(f^{k}(y), f^{k}(z)) \leq \sigma^{n-k} d(f^{n}(y), f^{n}(z)), \quad \forall 0 \leq k < n.$

A positive frequency of hyperbolic times guarantees sufficiently regular statistical behavior (Santana, 2021, Li et al., 2014).

Inducing schemes partition the phase space into regions where the first return (or inducing) map is uniformly expanding and admits bounded distortion. The induced system, coded as a (possibly countable) Markov shift, enables construction of Gibbs–Markov–Young towers, which have exponential or polynomial tails for return times—crucial for establishing decay rates of correlations and limit laws (Arima, 19 Oct 2025, Hafouta, 2020).

3. Thermodynamic Formalism and Equilibrium States

Thermodynamic formalism in the non-uniformly expanding setting revolves around the study of topological pressure:

$P_{f}(\varphi) = \sup_{\mu \in M(f)} \left( h_\mu(f) + \int \varphi\, d\mu \right),$

and the existence/uniqueness of equilibrium states for Hölder or continuous potentials $\varphi$ . For random dynamical systems (skew-product or i.i.d. compositions), the pressure is defined via integrals of transfer operator norms over the base $\Omega$ :

$P_{\text{top}}(\varphi) = \lim_{n \to \infty} \frac{1}{n} \int_\Omega \log \| \mathcal{L}_{\varphi_\omega}^n 1 \|_\infty\, dP(\omega).$

Under small variation or hyperbolicity conditions for $\varphi$ , unique equilibrium states arise—even without Markov structure, uniform degree, or expansion everywhere. These equilibrium states possess weak (or full) Gibbs properties and exponential decay of correlations (Castro et al., 2012, Santana, 2021, Stadlbauer et al., 2020, Bilbao et al., 2020).

Ruelle–Perron–Frobenius operators acting on spaces of Hölder (or generalized Hölder, BV, quasi-Hölder) functions admit spectral gaps, leading to exponential or polynomial mixing rates and central limit theorems (Castro et al., 2012, 1711.08052, Hu et al., 2013). Random perturbations preserve the spectral gap and stability of equilibrium states in many cases (Bilbao et al., 2020, Shen, 2011).

4. Multifractal Analysis and Level-Set Spectra

A central focus is the multifractal spectrum of Birkhoff averages:

$\Lambda_\alpha = \left\{ x : \lim_{n \to \infty} \frac{1}{n} \sum_{k=0}^{n-1} \varphi(f^{k} x) = \alpha \right\}, \quad b(\alpha) = \dim_H \Lambda_\alpha.$

Conditional variational principles describe $b(\alpha)$ as the supremum of entropy divided by average expansion over invariant measures with fixed average $\alpha$ :

$b(\alpha) = \sup \left\{ \frac{h_\mu(f)}{\lambda(\mu)} : \int \varphi\, d\mu = \alpha, \lambda(\mu) = \int \log |f'|\, d\mu > 0 \right\},$

with Legendre-type formulas relating $b(\alpha)$ to pressure functions (Arima, 19 Oct 2025, Arima, 4 Nov 2025). For non-uniformly expanding Rényi maps and their relatives, the multifractal spectrum outside central plateau intervals is real-analytic, strictly monotone, and identifies unique equilibrium states achieving prescribed averages (Arima, 4 Nov 2025). The central plateau (from parabolic fixed points) supports measures of full Hausdorff dimension.

Applications include the analysis of Khinchin exponents, arithmetic means in backward continued fractions, and continued fraction expansions (Arima, 4 Nov 2025).

5. Statistical Properties: Mixing Rates and Limit Laws

Statistical properties of non-uniformly expanding maps critically rely on the tail estimates for return times in the inducing scheme:

Exponential tail: $m\{R > n\} \leq Ce^{-cn}$
Polynomial tail: $m\{R > n\} \leq C n^{-\beta}$

Young tower machinery leads to stretched-exponential or exponential decay of quenched correlations for random or deterministic compositions (Li et al., 2014, Castro et al., 2012, Hafouta, 2020). Polynomial tails yield optimal polynomial lower bounds for decay (Hu et al., 2013), which can be strictly quantified for observables supported away from neutral points. Central limit theorems, Berry–Esseen bounds, local limit theorems, and large/moderate deviations principles are verified for suitably regular random or deterministic systems, particularly when exponential tail conditions are met (Hafouta, 2020).

6. Stochastic Stability and Robustness

Non-uniform expansion persists under small random perturbations for a broad class of maps (interval, higher-dimensional, skew-products), provided summability, slow recurrence, and integrability conditions for hyperbolic times are satisfied (Shen, 2011, Araujo et al., 2013, Alves et al., 2010). Strong stochastic stability entails the convergence of stationary measure densities in $L^1$ (not just weak*-convergence) to those of the unperturbed SRB measure (Alves et al., 2010, Shen, 2011). Cone contraction, projective metrics, and uniform spectral gap arguments undergird continuity results for equilibrium states and pressure functions in both deterministic and random settings (Castro et al., 2012, Bilbao et al., 2020).

7. Symbolic Dynamics, Coding, and Markov Structures

Symbolic models—often based on countable Markov shifts—can be constructed for invariant measures of non-uniformly expanding interval maps, even in the presence of critical points, discontinuities, and infinite distortion (Lima, 2018). The coding map is Hölder continuous and typically finite-to-one onto the natural extension, recovering product structure and Markov partitions. Weak Markov or inducing structures on non-hyperbolic sets allow for robust symbolic decompositions, facilitating multifractal analysis, periodic point counting, and fine classification of equilibrium states. Notably, the coding techniques subsume classical Markov and Hofbauer tower frameworks (Lima, 2018).

8. Applications and Illustrative Examples

Pomeau–Manneville maps: Polynomial decay of correlations; infinite distortion near neutral points (Hu et al., 2013, Duvall, 2019).
Rényi backward continued fraction maps: Complete multifractal analysis for Khinchin exponents and backward digits; analytic dimension spectra (Arima, 4 Nov 2025).
Viana maps and quadratic families: Randomly driven skew-products and induced maps with critical/singular sets; strong stochastic stability and exponential mixing (Li et al., 2014, Santana, 2021, Alves et al., 2010).
Circle maps with logarithmic singularities: Chaotic dynamics and existence of SRB measures for parametrized singular maps (Takahasi et al., 2011).
Random random non-uniformly expanding systems: Limit theorems for random towers, storchastic stability, invariant density continuity (Hafouta, 2020, Alves et al., 2010).

9. Advanced Topics and Open Directions

Recent developments extend the scope of non-uniformly expanding theory via:

Relative pressure for noncompact invariant sets: Carathéodory-type constructions and variational principles for pressure on arbitrary random sets (Stadlbauer et al., 2020).
Flatness criteria for potentials: Optimal transportation and Wasserstein contraction approaches to spectral gaps and decay rates without inducing or Markov partitions (1711.08052).
Phase transitions and multifractal spectra: Identification of analytic sectors and plateaux in dimension spectra, especially in systems with parabolic or neutral fixed points (Arima, 19 Oct 2025, Arima, 4 Nov 2025).
Random thermodynamic formalism: Stability, existence, and uniqueness of equilibrium states in highly non-uniform settings without Markov or bounded-degree assumptions (Stadlbauer et al., 2020, Bilbao et al., 2020).

Ongoing research seeks to generalize these mechanisms to higher dimensions, potentials of unbounded variation, partially hyperbolic systems, and systems with infinite or uncountable critical structures.

Key References (arXiv ids for further reading):

Area	Example Map/Class	arXiv Reference
Inducing/Young towers	Interval maps with parabolic points	(Arima, 19 Oct 2025, Arima, 4 Nov 2025)
Random non-uniform expansion	Skew-product, Viana maps	(Li et al., 2014, Stadlbauer et al., 2020, Santana, 2021)
Strong stochastic stability	Interval diffeomorphisms, Viana maps	(Alves et al., 2010, Shen, 2011)
Flatness/waterstein contraction	Circle/intermittent maps	(1711.08052, Hu et al., 2013)
Symbolic coding	Non-uniformly expanding interval maps	(Lima, 2018)
Multifractal spectra	Backward continued fractions	(Arima, 4 Nov 2025)