Strong Positive Recurrence in Dynamics

• Strong Positive Recurrence (SPR) is a property where dominant invariant dynamics prevent invariant mass or complexity from escaping to infinity.
• SPR is characterized by exponential return-time moments, pressure gaps at infinity, and uniformly positive mass on fixed Pesin blocks or compact sets.
• This concept underpins finite Gibbs measures, unique equilibrium states, and robust statistical properties across symbolic, geometric, and smooth dynamical systems.

Strong positive recurrence (SPR) is a recurrence principle that appears in several branches of dynamics, thermodynamic formalism, and Markov theory, with a common underlying theme: the dominant invariant objects cannot lose mass or complexity to “infinity,” whether infinity is literal noncompact escape, symbolic tails in a countable-state model, or degeneration of hyperbolicity constants in Pesin theory. In the countable-state and matrix setting, SPR is a strengthening of positive recurrence characterized by exponential return-time moments or equivalent spectral-gap criteria (Swart, 2017). In noncompact negative-curvature dynamics, a potential is SPR when its pressure at infinity is strictly smaller than its full pressure (Gouëzel et al., 2020). In recent smooth dynamics, SPR has been defined intrinsically for diffeomorphisms and singular flows by requiring that all sufficiently large-entropy or large-pressure ergodic measures assign uniformly positive mass to a fixed Pesin block or a compact subset of a scaled Pesin block (Buzzi et al., 13 Jan 2025, Li et al., 29 Apr 2026). Across these settings, SPR functions as a rigidity condition that upgrades existence statements to finiteness, uniqueness, Bernoulli structure, spectral gaps, and strong statistical properties.

1. Symbolic and matrix origins

The classical symbolic prototype of SPR comes from irreducible countable-state Markov shifts. For an irreducible graph $\mathcal G$, with $Z_n(a)$ the number of loops of length $n$ at a vertex $a$ and $Z_n^*(a)$ the number of first-return loops of length $n$, the shift is SPR when

$$\limsup_{n\to\infty}\frac1n\log Z_n^*(a) < \limsup_{n\to\infty}\frac1n\log Z_n(a).$$

This inequality expresses an entropy gap between primitive returns and the full loop growth, and the paper on diffeomorphisms identifies it as the symbolic model for smooth SPR (Buzzi et al., 13 Jan 2025).

In matrix language, the corresponding notion is strong $R$-positivity. For an irreducible nonnegative matrix $A$ with critical $h$-transform

$Z_n(a)$0

the matrix is strongly $Z_n(a)$1-positive exactly when the recurrent Markov kernel $Z_n(a)$2 is strongly positive recurrent in the Markov-chain sense, namely when for some state $Z_n(a)$3,

$Z_n(a)$4

for some $Z_n(a)$5 (Swart, 2017). Swart proves that this property is equivalent to a perturbative spectral criterion: lowering finitely many entries of $Z_n(a)$6 while keeping the same support strictly lowers the spectral radius $Z_n(a)$7. More precisely, if $Z_n(a)$8, $Z_n(a)$9, and $n$0, then strong $n$1-positivity implies

$n$2

and conversely, if the set of modified entries is finite and $n$3, then $n$4 is strongly $n$5-positive (Swart, 2017). This gives a practical criterion for SPR that avoids direct return-time estimates.

The same paper reformulates the condition via excursion generating functions. If $n$6 is the logarithmic generating function for excursions away from $n$7, and $n$8, then

$n$9

where $a$0 is the right endpoint of finiteness of $a$1 (Swart, 2017). This is exactly the existence of a nontrivial exponential moment for return times under the critical $a$2-transform.

2. Pressure gaps and infinity

In geometric thermodynamic formalism, SPR is formulated as a pressure gap at infinity. For the geodesic flow on a nonelementary complete connected negatively curved manifold or good orbifold with pinched negative curvature and bounded first derivative of curvature, Gouëzel, Schapira, and Tapie define three versions of pressure at infinity: a geometric critical exponent at infinity $a$3, a Gurevič pressure at infinity $a$4, and a variational pressure at infinity $a$5. They prove that these coincide: $a$6 A Hölder potential $a$7 is then called strongly positively recurrent when

$a$8

equivalently $a$9 (Gouëzel et al., 2020).

This definition makes precise the idea that escaping mass cannot realize full pressure. The variational form is especially explicit: $Z_n^*(a)$0 so SPR excludes sequences of invariant probability measures whose mass escapes to infinity while asymptotically carrying maximal pressure (Gouëzel et al., 2020).

A structurally parallel notion has been introduced for convergence groups equipped with a continuous Gromov–Patterson–Sullivan system. There the entropy at infinity is

$Z_n^*(a)$1

where $Z_n^*(a)$2 records orbit segments joining $Z_n^*(a)$3 to $Z_n^*(a)$4 without crossing other translates in between, and $Z_n^*(a)$5 is the critical exponent of the restricted Poincaré series (Wen, 5 Apr 2026). The group is SPR with respect to $Z_n^*(a)$6 and $Z_n^*(a)$7 if

$Z_n^*(a)$8

This is again a critical gap at infinity, now in a GPS flow-space setting rather than a negatively curved manifold (Wen, 5 Apr 2026).

3. Intrinsic smooth definitions

Recent work defines SPR directly for smooth systems rather than through an a priori symbolic coding. For $Z_n^*(a)$9 diffeomorphisms of closed manifolds, the intrinsic definition uses Pesin blocks. A diffeomorphism $n$0 is SPR if there exists $n$1 such that for each $n$2, there are a Borel $n$3-Pesin block $n$4, a threshold $n$5, and $n$6 such that every ergodic measure $n$7 with $n$8 satisfies

$n$9

Equivalently, every ergodic measure of entropy close enough to the topological entropy must spend a uniformly positive proportion of its mass inside one fixed uniformly hyperbolic block (Buzzi et al., 13 Jan 2025).

The same work interprets this as an entropy gap at infinity in the Pesin bornology. If $$\limsup_{n\to\infty}\frac1n\log Z_n^*(a) < \limsup_{n\to\infty}\frac1n\log Z_n(a).$$0 denotes the family of $$\limsup_{n\to\infty}\frac1n\log Z_n^*(a) < \limsup_{n\to\infty}\frac1n\log Z_n(a).$$1-Pesin blocks, the entropy at infinity is defined by

$$\limsup_{n\to\infty}\frac1n\log Z_n^*(a) < \limsup_{n\to\infty}\frac1n\log Z_n(a).$$2

and one has

$$\limsup_{n\to\infty}\frac1n\log Z_n^*(a) < \limsup_{n\to\infty}\frac1n\log Z_n(a).$$3

for invariant Borel sets $$\limsup_{n\to\infty}\frac1n\log Z_n^*(a) < \limsup_{n\to\infty}\frac1n\log Z_n(a).$$4 such as Borel homoclinic classes (Buzzi et al., 13 Jan 2025). This shows that even on compact manifolds, nonuniform hyperbolicity generates an effective “infinity” through degenerating Pesin constants.

For singular flows, compactness of ordinary Pesin blocks can fail because singularities may lie in their closure. The 2026 paper on singular flows therefore defines SPR intrinsically using compact subsets of scaled Pesin blocks. If $$\limsup_{n\to\infty}\frac1n\log Z_n^*(a) < \limsup_{n\to\infty}\frac1n\log Z_n(a).$$5 is a flow on an invariant Borel set $$\limsup_{n\to\infty}\frac1n\log Z_n^*(a) < \limsup_{n\to\infty}\frac1n\log Z_n(a).$$6 and $$\limsup_{n\to\infty}\frac1n\log Z_n^*(a) < \limsup_{n\to\infty}\frac1n\log Z_n(a).$$7 is a Hölder potential, then $$\limsup_{n\to\infty}\frac1n\log Z_n^*(a) < \limsup_{n\to\infty}\frac1n\log Z_n(a).$$8 is SPR on $$\limsup_{n\to\infty}\frac1n\log Z_n^*(a) < \limsup_{n\to\infty}\frac1n\log Z_n(a).$$9 for $R$0 if there exists $R$1 such that for each $R$2, there are a $R$3-Pesin block, a compact set

$R$4

and numbers

$R$5

such that every ergodic measure $R$6 with $R$7 satisfies

$R$8

The authors explicitly interpret this as the singular-flow analogue of a pressure gap away from infinity (Li et al., 29 Apr 2026).

A closely related potential-theoretic version has been established for smooth surface diffeomorphisms. There, for a continuous potential $R$9, $A$0 is $A$1-SPR for $A$2 if for every $A$3 and every sequence of ergodic measures $A$4 with $A$5,

$A$6

An equivalent formulation is that high-pressure ergodic measures must assign uniformly positive mass to a fixed Pesin set $A$7 (Luo et al., 4 Feb 2026).

4. Symbolic codings and exact bridges

A central feature of modern SPR theory is that intrinsic smooth SPR can be transferred to countable-state symbolic systems. For singular flows, the symbolic model is a locally compact topological Markov flow $A$8, obtained as a suspension over a countable-state topological Markov shift $A$9 with roof function $h$0, together with a coding map $h$1 (Li et al., 29 Apr 2026). The symbolic SPR definition for the suspension asks for a finite union of cylinders $h$2, $h$3, and $h$4 such that every ergodic $h$5-invariant measure $h$6 with $h$7 satisfies

$h$8

The exact bridge is

$h$9

and intrinsic flow SPR implies that the coding topological Markov flow is SPR for the lifted potential $Z_n(a)$00 (Li et al., 29 Apr 2026).

For diffeomorphisms, the symbolic coding theorem is similar in spirit. If $Z_n(a)$01 is a $Z_n(a)$02 diffeomorphism and $Z_n(a)$03 is either the whole manifold or a Borel homoclinic class, then for every $Z_n(a)$04 there exists a locally compact countable-state Markov shift $Z_n(a)$05 and a Hölder coding $Z_n(a)$06 such that $Z_n(a)$07, entropy is preserved under projection and lifting for the relevant hyperbolic measures, and if $Z_n(a)$08 is SPR then $Z_n(a)$09 is SPR and

$Z_n(a)$10

(Buzzi et al., 13 Jan 2025). The potential-theoretic surface result gives the analogous statement for equilibrium states: if a Borel homoclinic class is $Z_n(a)$11-SPR for $Z_n(a)$12, then there is a locally compact irreducible Markov shift $Z_n(a)$13 such that $Z_n(a)$14 is SPR and

$Z_n(a)$15

(Luo et al., 4 Feb 2026).

These coding theorems are not merely representational. They identify SPR as the exact hypothesis that permits the import of countable-state thermodynamic formalism, including the spectral theory of transfer operators. In the diffeomorphism paper, a theorem of Cyr–Sarig is invoked to obtain a Banach space $Z_n(a)$16 and a normalized transfer operator

$Z_n(a)$17

with decomposition

$Z_n(a)$18

where $Z_n(a)$19 has rank one and $Z_n(a)$20 has spectral radius $Z_n(a)$21 (Buzzi et al., 13 Jan 2025). A plausible implication is that, in the modern smooth literature, SPR has become the precise smooth-dynamical condition guaranteeing that the symbolic model lies in the spectral-gap regime.

5. Consequences

The main consequences of SPR are strongest when formulated through thermodynamic formalism.

Setting	SPR condition	Representative consequence
Countable matrices	finite perturbations lower $Z_n(a)$22	exponential return moments (Swart, 2017)
Negative curvature	$Z_n(a)$23	finite Gibbs measure (Gouëzel et al., 2020)
Singular flows	compact subset of scaled Pesin block captures all high-pressure measures	unique/finitely many equilibrium states (Li et al., 29 Apr 2026)
Diffeomorphisms	high-entropy measures charge a fixed Pesin block	finitely many MMEs; exponential mixing (Buzzi et al., 13 Jan 2025)
Surface potentials	near-equilibrium measures concentrate on Pesin sets	exponential mixing of equilibrium states (Luo et al., 4 Feb 2026)
GPS systems	$Z_n(a)$24	finite BMS measure (Wen, 5 Apr 2026)

For countable-state matrices, strong $Z_n(a)$25-positivity is equivalent to strong positive recurrence of the critical $Z_n(a)$26-transformed kernel, and in the aperiodic positive recurrent case it is equivalent to geometric ergodicity, for example

$Z_n(a)$27

(Swart, 2017).

For geodesic flows in negative curvature, SPR implies existence of a finite Gibbs measure, hence an equilibrium state, and is equivalent to exponential recurrence with respect to that Gibbs measure. One also has the escape-of-mass inequality

$Z_n(a)$28

for vaguely convergent measures $Z_n(a)$29, which shows directly that near-maximizing sequences cannot lose all mass when $Z_n(a)$30 (Gouëzel et al., 2020).

For singular flows, if $Z_n(a)$31 is SPR for $Z_n(a)$32 on a Borel homoclinic class $Z_n(a)$33, then there is a unique equilibrium state of $Z_n(a)$34 on $Z_n(a)$35, and it is Bernoulli up to a period. If $Z_n(a)$36 is SPR on all of $Z_n(a)$37, then there are finitely many ergodic equilibrium states, each Bernoulli up to a period (Li et al., 29 Apr 2026). Rapid mixing is not automatic: it requires, in addition, the “good asymptotics” hypothesis for the homoclinic class (Li et al., 29 Apr 2026).

For diffeomorphisms, SPR yields exactly one local measure of maximal entropy on an SPR Borel homoclinic class, and finitely many ergodic MMEs for an SPR system with positive topological entropy (Buzzi et al., 13 Jan 2025). It also implies effective intrinsic ergodicity: for Hölder or quasi-Hölder observables $Z_n(a)$38,

$Z_n(a)$39

for invariant measures $Z_n(a)$40 on the class (Buzzi et al., 13 Jan 2025). Exponential mixing holds on ergodic components of the appropriate iterate: $Z_n(a)$41 The same paper lists large deviations, almost sure invariance principle, central limit theorem, law of iterated logarithm, arcsine law, and record statistics among the consequences (Buzzi et al., 13 Jan 2025).

For Hölder potentials on $Z_n(a)$42 surface diffeomorphisms, the 2026 potential-theoretic extension proves that if

$Z_n(a)$43

then $Z_n(a)$44 is $Z_n(a)$45-SPR for $Z_n(a)$46 for some $Z_n(a)$47. Consequently there are at most finitely many ergodic equilibrium states, weak-* limits of ergodic pressure-maximizing sequences are ergodic equilibrium states, Lyapunov exponents converge along such sequences, and exponentially mixing equilibrium states are obtained under mixing hypotheses (Luo et al., 4 Feb 2026).

For GPS systems, SPR implies divergence of the full Poincaré series at the critical exponent and, under the comparison hypothesis

$Z_n(a)$48

for every $Z_n(a)$49, it implies finiteness of the BMS measure on the associated flow space (Wen, 5 Apr 2026). This extends classical negative-curvature finiteness criteria and covers relatively Anosov groups as well as new higher-rank examples (Wen, 5 Apr 2026).

6. Scope, examples, and terminological caveats

SPR is broad but not universal. In the smooth literature, one prominent scope theorem states that every $Z_n(a)$50 surface diffeomorphism with positive entropy is SPR (Buzzi et al., 13 Jan 2025). The potential-theoretic extension shows that every Hölder potential with oscillation smaller than topological entropy is SPR on a $Z_n(a)$51 surface diffeomorphism with positive entropy, and gives a $Z_n(a)$52 variant involving the correction term $Z_n(a)$53 (Luo et al., 4 Feb 2026). For singular flows, every $Z_n(a)$54 three-dimensional flow with positive topological entropy is SPR in the zero-potential case, and Hölder potentials with

$Z_n(a)$55

are SPR as well (Li et al., 29 Apr 2026). In negative curvature, compact or convex-cocompact manifolds satisfy $Z_n(a)$56 for every Hölder $Z_n(a)$57, so they are automatically SPR whenever $Z_n(a)$58 is finite (Gouëzel et al., 2020). For GPS systems, geometrically finite groups with parabolic gap and relatively Anosov groups are SPR, and free-product constructions yield additional higher-rank examples beyond the relatively Anosov category (Wen, 5 Apr 2026).

The terminology is not uniform across mathematics. In the matrix and symbolic literature, strong positive recurrence is tightly linked to positive recurrence, spectral radii, and exponential moments (Swart, 2017). In geometry and thermodynamic formalism, it is a pressure-gap property at infinity (Gouëzel et al., 2020, Wen, 5 Apr 2026). In smooth ergodic theory, it is an intrinsic Pesin-block condition for high-entropy or high-pressure measures (Buzzi et al., 13 Jan 2025, Luo et al., 4 Feb 2026, Li et al., 29 Apr 2026). By contrast, the number-theoretic “strong recurrence” of the Riemann zeta function concerns self-approximation under vertical shifts and is equivalent to the Riemann hypothesis only in the $Z_n(a)$59 case; it is not a notion of strong positive recurrence (Nakamura, 2010). Likewise, the ergodic-theoretic “set of strong recurrence” for subsets of $Z_n(a)$60 concerns positive return correlations

$Z_n(a)$61

and the paper studying it explicitly notes that it does not introduce a separate notion called SPR (Mountakis, 2022). In stochastic-process papers on reflecting Brownian motion or piecewise Ornstein–Uhlenbeck diffusions, the proved notion is often positive recurrence, sometimes with exponential ergodicity in special regimes, but not an explicit SPR definition (Bramson, 2010, Dieker et al., 2011, Bramson et al., 2010).

Taken together, these developments show that “strong positive recurrence” is best understood not as a single formula shared across all disciplines, but as a family of structurally analogous gap conditions. In each setting, SPR identifies the regime where recurrent core dynamics dominate escape mechanisms strongly enough to force finite invariant measures and robust statistical structure.