Pomeau-Manneville Map Dynamics

• Pomeau–Manneville map is an intermittent one-dimensional system with a neutral fixed point that leads to weak chaos and prolonged laminar phases.
• Its construction, combining a tangent-to-the-identity branch near zero with an expanding branch, underpins the observed polynomial decay of correlations and anomalous transport.
• The model serves as a benchmark for investigating ergodic theory, linear response, thermodynamic formalism, and nonstationary behavior in chaotic systems.

Searching arXiv for relevant papers on the Pomeau–Manneville map to support the article. The Pomeau–Manneville map is a one-dimensional intermittent dynamical system with a neutral fixed point at $x=0$, introduced in the modern intermittency literature as a minimal model of weak chaos. In its standard interval form it combines a tangent-to-the-identity branch near $0$ with an expanding branch away from $0$; the resulting dynamics exhibits long laminar episodes near the neutral point interspersed with expanding excursions. Across its various normalizations, the family has become a reference model for polynomial decay of correlations, stochastic stability, linear response, nonstationary and random compositions, thermodynamic phase transitions, anomalous transport, and open-system renewal phenomena (Bahsoun et al., 2017, Coronel et al., 10 Apr 2025, Klages, 2015).

1. Definitions and parameterizations

A standard piecewise form fixes $\alpha\in(0,1)$ and defines

$$T_\alpha(x)= \begin{cases} x\bigl(1+2^\alpha x^\alpha\bigr), & 0\le x\le \tfrac12,\ 2x-1, & \tfrac12<x\le 1. \end{cases}$$

Equivalent formulations appearing in the literature include $T_\alpha(x)=x+C x^{1+\alpha}\pmod 1$, the mod-$1$ family $f_\alpha(x)=x+x^{1+\alpha}\pmod1$, and the notation $P_{a,z}(x)=x+a x^z\bmod1$. In all cases, $x=0$ is an indifferent fixed point with derivative $0$0, while the map is expanding away from $0$1 (Bahsoun et al., 2017, Shen et al., 2012, Brevitt et al., 2022).

Convention	Formula	Regime boundary stated in the source
Piecewise PM / LSV form	$0$2 on $0$3, $0$4 on $0$5	unique acim for $0$6
Mod-$0$7 PM form	$0$8	no non-trivial acim for $0$9
$0$0 form	$0$1	finite SRB for $0$2, infinite invariant measure for $0$3

The neutral branch controls the intermittency strength. In the formulation $0$4, larger $0$5 makes the map flatter near $0$6 and lengthens laminar trapping. In the more general framework of expanding interval maps with an indifferent point, the PM family sits inside a broader class of piecewise-$0$7 or piecewise-$0$8 maps with full increasing branches, bounded distortion away from the neutral point, and explicit monotonicity conditions on inverse branches (Bonanno et al., 2019).

2. Deterministic ergodic structure and intermittency

In the finite-measure regime, the map admits a unique absolutely continuous invariant probability measure. For $0$9 with $\alpha\in(0,1)$0, one has

$\alpha\in(0,1)$1

so the density is singular but integrable. The same phenomenon is described in the $\alpha\in(0,1)$2 notation by $\alpha\in(0,1)$3, which is integrable for $\alpha\in(0,1)$4 (Shen et al., 2012, Klages, 2015).

A central statistical consequence of the neutral fixed point is polynomial, rather than exponential, mixing. For $\alpha\in(0,1)$5, the correlation function decays as

$\alpha\in(0,1)$6

equivalently at rate $\alpha\in(0,1)$7. This slow decay underlies the anomalous statistics of observables and the sensitivity of limit theorems to the parameter range (Bahsoun et al., 2017, Shen et al., 2012, Leppänen et al., 2015).

Beyond the finite-measure regime, the system passes into infinite-ergodic behavior. In the $\alpha\in(0,1)$8 convention, $\alpha\in(0,1)$9 forces concentration at the neutral fixed point, and for Lebesgue-a.e. initial condition the empirical measures converge to $$T_\alpha(x)= \begin{cases} x\bigl(1+2^\alpha x^\alpha\bigr), & 0\le x\le \tfrac12,\ 2x-1, & \tfrac12<x\le 1. \end{cases}$$0. In the $$T_\alpha(x)= \begin{cases} x\bigl(1+2^\alpha x^\alpha\bigr), & 0\le x\le \tfrac12,\ 2x-1, & \tfrac12<x\le 1. \end{cases}$$1 convention, $$T_\alpha(x)= \begin{cases} x\bigl(1+2^\alpha x^\alpha\bigr), & 0\le x\le \tfrac12,\ 2x-1, & \tfrac12<x\le 1. \end{cases}$$2 yields an infinite invariant measure and vanishing ordinary Lyapunov exponent, with subexponential separation of nearby trajectories and nonstandard ergodic limits (Shen et al., 2012, Klages, 2015).

3. Random perturbations, nonstationarity, and response theory

For random compositions, Bahsoun–Ruziboev–Saussol study iid selections $$T_\alpha(x)= \begin{cases} x\bigl(1+2^\alpha x^\alpha\bigr), & 0\le x\le \tfrac12,\ 2x-1, & \tfrac12<x\le 1. \end{cases}$$3 with law $$T_\alpha(x)= \begin{cases} x\bigl(1+2^\alpha x^\alpha\bigr), & 0\le x\le \tfrac12,\ 2x-1, & \tfrac12<x\le 1. \end{cases}$$4 and random transfer operator

$$T_\alpha(x)= \begin{cases} x\bigl(1+2^\alpha x^\alpha\bigr), & 0\le x\le \tfrac12,\ 2x-1, & \tfrac12<x\le 1. \end{cases}$$5

Under piecewise $$T_\alpha(x)= \begin{cases} x\bigl(1+2^\alpha x^\alpha\bigr), & 0\le x\le \tfrac12,\ 2x-1, & \tfrac12<x\le 1. \end{cases}$$6-onto and uniform-spectral-gap assumptions, the absolutely continuous stationary density $$T_\alpha(x)= \begin{cases} x\bigl(1+2^\alpha x^\alpha\bigr), & 0\le x\le \tfrac12,\ 2x-1, & \tfrac12<x\le 1. \end{cases}$$7 is differentiable in the $$T_\alpha(x)= \begin{cases} x\bigl(1+2^\alpha x^\alpha\bigr), & 0\le x\le \tfrac12,\ 2x-1, & \tfrac12<x\le 1. \end{cases}$$8-norm and satisfies the linear-response formula

$$T_\alpha(x)= \begin{cases} x\bigl(1+2^\alpha x^\alpha\bigr), & 0\le x\le \tfrac12,\ 2x-1, & \tfrac12<x\le 1. \end{cases}$$9

For a uniform law on $T_\alpha(x)=x+C x^{1+\alpha}\pmod 1$0, the random response is exactly one half of the deterministic response: $T_\alpha(x)=x+C x^{1+\alpha}\pmod 1$1 This gives a precise relation between stochastic and deterministic parameter perturbations for PM maps (Bahsoun et al., 2017).

Small additive noise does not destroy the invariant density in the finite-measure regime. For $T_\alpha(x)=x+C x^{1+\alpha}\pmod 1$2 with $T_\alpha(x)=x+C x^{1+\alpha}\pmod 1$3 uniformly distributed on $T_\alpha(x)=x+C x^{1+\alpha}\pmod 1$4, the stationary density $T_\alpha(x)=x+C x^{1+\alpha}\pmod 1$5 converges strongly in $T_\alpha(x)=x+C x^{1+\alpha}\pmod 1$6 to the deterministic density $T_\alpha(x)=x+C x^{1+\alpha}\pmod 1$7 as $T_\alpha(x)=x+C x^{1+\alpha}\pmod 1$8 when $T_\alpha(x)=x+C x^{1+\alpha}\pmod 1$9. The source further states a quantitative bound

$1$0

for $1$1 (Shen et al., 2012).

Nonstationary and random concatenations of PM maps also support strong probabilistic limit laws. For sequential maps with parameters $1$2, Su proves an Almost Sure Invariance Principle under a variance-growth condition $1$3 with

$1$4

and obtains a quenched ASIP for random compositions. Nicol–Török–Vaienti establish self-norming central limit theorems for sequential systems when $1$5, together with quenched CLTs for iid random compositions. In a quasistatic regime with slowly varying parameter curve $1$6, Leppänen–Stenlund identify the physical family $1$7, obtaining convergence in probability for $1$8 and almost sure convergence for $1$9. Korepanov–Leppänen later sharpen the polynomial memory-loss rate for nonstationary and random intermittent compositions to the stationary-order exponent $f_\alpha(x)=x+x^{1+\alpha}\pmod1$0 under return-tail assumptions (Su, 2019, Nicol et al., 2015, Leppänen et al., 2015, Korepanov et al., 2024).

With additive noise at fixed positive amplitude, the annealed transfer operator becomes a compact Hilbert–Schmidt integral operator on $f_\alpha(x)=x+x^{1+\alpha}\pmod1$1, and explicit unique optimal perturbations can be written for maximizing either the infinitesimal change of an observable’s expectation or the spectral gap. The PM map is one of the concrete examples in which these optimal-response formulas are computed (Antown et al., 2021).

4. Thermodynamic formalism and multifractal structure

Recent work gives a detailed phase diagram for Hölder potentials over intermittent maps. For the PM map $f_\alpha(x)=x+x^{1+\alpha}\pmod1$2 and potentials $f_\alpha(x)=x+x^{1+\alpha}\pmod1$3, the topological pressure is

$f_\alpha(x)=x+x^{1+\alpha}\pmod1$4

The intermittent phase is characterized by

$f_\alpha(x)=x+x^{1+\alpha}\pmod1$5

while the stationary phase is the interior of $f_\alpha(x)=x+x^{1+\alpha}\pmod1$6. If $f_\alpha(x)=x+x^{1+\alpha}\pmod1$7 and $f_\alpha(x)=x+x^{1+\alpha}\pmod1$8, then $f_\alpha(x)=x+x^{1+\alpha}\pmod1$9 and $P_{a,z}(x)=x+a x^z\bmod1$0 is real-analytic on all of $P_{a,z}(x)=x+a x^z\bmod1$1. If $P_{a,z}(x)=x+a x^z\bmod1$2, both phases occur: $P_{a,z}(x)=x+a x^z\bmod1$3 is star-convex at $P_{a,z}(x)=x+a x^z\bmod1$4, $P_{a,z}(x)=x+a x^z\bmod1$5 is convex and unbounded, and the phase-transition locus

$P_{a,z}(x)=x+a x^z\bmod1$6

is a topological codimension-$P_{a,z}(x)=x+a x^z\bmod1$7 submanifold, homeomorphic to a closed hyperplane, and exactly the set where the pressure fails to be real-analytic (Coronel et al., 10 Apr 2025).

Temperature scaling produces a second, one-parameter phase-transition picture. For each $P_{a,z}(x)=x+a x^z\bmod1$8, either $P_{a,z}(x)=x+a x^z\bmod1$9 for all $x=0$0, or there is a unique critical inverse temperature $x=0$1 such that $x=0$2 lies in the intermittent phase for $x=0$3, on the transition locus at $x=0$4, and in the stationary phase for $x=0$5. For $x=0$6, the geometric potential $x=0$7 has a non-persistent transition at $x=0$8, whereas for $x=0$9 all temperature transitions are persistent and $0$00 varies continuously with $0$01 (Coronel et al., 10 Apr 2025).

Multifractal analysis for the Manneville–Pomeau map departs sharply from the uniformly hyperbolic case. Iommi–Jordan show that the Birkhoff spectrum can be discontinuous, and more generally that the interval of average values can split into subintervals $0$02 on which the Hausdorff-dimension spectrum is analytic on $0$03 and $0$04 but constant on $0$05. The possibility of discontinuity and plateau behavior is a signature of concentration effects near the indifferent fixed point (Iommi et al., 2013).

A complementary thermodynamic description based on infinite ergodic theory gives explicit formulas for topological pressure and Rényi entropy in the parameterization

$0$06

For $0$07,

$0$08

and the pressure undergoes a second-order phase transition at $0$09. The same framework relates these thermodynamic singularities to Mittag–Leffler fluctuations of ergodic sums and algorithmic complexity (Venegeroles, 2012).

5. Infinite-measure dynamics, global–local mixing, and nonstandard statistics

In the infinite-measure regime, standard mixing notions are replaced by infinite-volume observables and conservative exact dynamics. For a class of interval maps satisfying assumptions (A1)–(A5), including classical PM and LSV maps

$0$10

Bonanno–Lenci prove full global–local mixing relative to the reference measure $0$11. When $0$12, the same full global–local mixing holds relative to the invariant measure $0$13; when $0$14, it holds for a smaller class of global observables. This gives a precise decorrelation statement in the absence of a finite invariant probability measure (Bonanno et al., 2019).

The infinite-ergodic viewpoint also yields explicit limit laws. In the $0$15 or corresponding infinite-measure regime, the Aaronson–Darling–Kac theorem replaces the ordinary ergodic theorem: normalized sums of integrable observables converge in distribution to a Mittag–Leffler random variable. On this basis, one defines generalized Lyapunov and Kolmogorov–Sinai quantities, and the cited source states a generalized Pesin formula $0$16. The same scaling exponent controls anomalous transport in lifted versions of the map (Klages, 2015).

For the finite-measure but slowly mixing regime, higher-order correlation control remains delicate. Functional correlation bounds for time-dependent PM maps imply multivariate normal approximation. For a single map with parameter $0$17, one obtains a multivariate CLT for

$0$18

together with explicit convergence rates via Stein’s method and via Rio’s method in Kantorovich distance (Leppänen, 2017).

6. Spatial lifts, coupling, and open systems

Lifted PM maps on $0$19 generate anomalous transport, but the transport class depends strongly on the lifting scheme. In the superdiffusive lift studied by Brevitt and collaborators, the reduced map is $0$20 and the lift is parameterized by $0$21. Laminar waiting times follow a Pareto law with exponent $0$22, leading to three mean-square-displacement regimes: $0$23 with logarithmic corrections at $0$24 and $0$25. The same study shows, however, that superdiffusion occurs only on the singular set $0$26, and that the classical Lévy-walk matching requires $0$27. For generic noninteger $0$28, the lifted system is either normally diffusive or localized, with localization windows $0$29 and an infinite bifurcation hierarchy. Aging substantially changes the measured generalized diffusion coefficient, and in the related analysis of the same family it removes the suppression transition at $0$30 (Brevitt et al., 2024, Brevitt et al., 2022).

Other lift conventions lead to subdiffusive, rather than superdiffusive, transport. In the spatially extended map discussed in the weak-chaos survey, one obtains

$0$31

and the scaling limit is governed by a time-fractional diffusion equation with Caputo derivative. This suggests that anomalous transport in PM-type systems is not a single universal phenomenon but depends on how the deterministic lift encodes intercell motion (Klages, 2015).

Coupled PM systems preserve intermittency while changing its statistical signatures. For two diffusively coupled PM maps on $0$32, numerical investigations show that any nonzero coupling removes the integrable singularity of the one-dimensional invariant density at the marginal fixed point. Escape from a corner region obeys

$0$33

and the cumulative recurrence law satisfies $0$34 for every $0$35, twice the uncoupled one-dimensional exponent. Phase-space filling and finite-time Lyapunov large deviations exhibit stretched-exponential, rather than purely exponential, decay, and the measured dynamical exponents appear independent of the nonzero coupling strength. In an infinite mean-field setting, the coupled intermittent system admits a unique physical stationary state and polynomial convergence of sufficiently regular densities (Sala et al., 2014, Bahsoun et al., 2023).

Open PM maps with a hole admit a survivor-conditioned renewal theory. Inducing on a base $0$36 away from the neutral point leads to a killed induced operator $0$37, and the number $0$38 of completed survivor returns satisfies an asymptotic law conditioned on survival up to time $0$39. For a conditionally invariant density $0$40 with $0$41, the conditioned return count converges to a geometric distribution: $0$42 For bounded observables satisfying $0$43 near the neutral point, survivor-conditioned Birkhoff sums remain bounded; in particular $0$44 yields bounded survivor-conditioned Lyapunov stretching, and under stronger regularity assumptions the conditioned sum converges to a finite limit (Duvall, 25 Jun 2026).