Thermodynamic Formalism in Dynamical Systems

• Thermodynamic formalism is a mathematical framework that applies statistical mechanics concepts such as entropy, pressure, and equilibrium states to a wide range of dynamical systems.
• It utilizes tools like transfer operators, spectral theory, and variational principles to derive unique equilibrium measures and analyze phase transitions.
• The framework has broad applications in modeling non-uniform hyperbolic dynamics, neuronal spike-train inference, and quantum systems, enabling multifractal and geometric analyses.

Thermodynamic formalism is a mathematical framework that extends concepts and methods from statistical mechanics—such as entropy, pressure, and equilibrium states—to general dynamical systems. It provides rigorous tools to analyze the statistical, ergodic, and geometric properties of deterministic and random systems, particularly via variational principles, transfer operators, and equilibrium/Gibbs measures. The formalism encompasses a broad spectrum of dynamical models, from compact and uniformly hyperbolic maps to high-dimensional, non-uniformly hyperbolic, non-compact, symbolic, stochastic, quantum, and even correspondence-driven systems.

1. Core Structures: Transfer Operators, Spectral Theory, and Variational Principle

At the heart of thermodynamic formalism is the transfer (Ruelle–Perron–Frobenius) operator, which, for a given potential function φ on a compact metric space X and a continuous map T:X→X, acts as

$$(\mathcal{L}_\varphi f)(x) = \sum_{y:T(y)=x} e^{\varphi(y)} f(y).$$

The spectral properties of this operator are central: under regularity (e.g. Hölder continuity of φ), there exists a unique strictly positive continuous eigenfunction and an eigenmeasure corresponding to its spectral radius. These ingredients yield the unique equilibrium (Gibbs) measure, which realizes the supremum in the variational (maximum entropy) principle

$$P(\varphi) = \sup_{\mu \in \mathcal{M}_T}\left\{ h_\mu(T) + \int \varphi\,d\mu \right\}.$$

This principle generalizes free-energy minimization and links the topological pressure to a duality between entropy and potential averages. For iterated function systems with weights, analogous transfer and Markov operators can be defined, and similar spectral, variational, and equilibrium results hold, including for holonomic (skew-product) measures (Cioletti et al., 2017, Cofré et al., 2020).

2. Generalizations: Non-Uniform Hyperbolicity, Inducing, Countable Markov Models

Thermodynamic formalism extends beyond compact and uniformly hyperbolic systems via the construction of inducing schemes and Markov codings:

• Inducing schemes allow the reduction of dynamics to countable Markov shifts, facilitating fine statistical and multifractal analyses in interval maps and multimodal settings. The induced system admits uniformly expanding dynamics and a countable symbolic coding, to which Sarig's generalization of the Ruelle–Perron–Frobenius theorem applies (Iommi et al., 2012).
• Countable Markov shifts admit a robust version of the formalism with the Gurevich pressure, positive recurrence, and phase transitions (e.g. in support or analyticity) analyzed by the existence and structure of conformal measures and the spectral gap of transfer operators, as established by Sarig, and extended to non-compact attractors and higher-dimensional models (Raszeja, 2021, Jakobson, 2016).
• The interplay between classical and "hidden" pressure (e.g., restricting to non-atomic measures or recurrent configurations) reveals nontrivial phase-transition phenomena, as in the Makarov–Smirnov framework for one-dimensional dynamics (Zhang, 2015), and the characterization of phase transitions in generalized Markov shifts where conformal measures may “jump” support across a critical parameter (Raszeja, 2021).

3. Random, Skew-Product, and Non-Equilibrium Extensions

The formalism accommodates significant generalizations:

• Out-of-equilibrium and random systems: The pressure may be defined for randomly parametrized potentials, often via a supremum taken before temporal averaging, yielding the pressure out of equilibrium. The associated variational principle involves skew-product invariant measures and maximal asymptotic instability functions. Applications to random dynamical systems with Gibbsian base measures establish uniform hyperbolicity under open conditions, extending exponential mixing and large-deviation estimates beyond the i.i.d. regime (Ovadia et al., 7 May 2025).
• Skew-periodic and transient extensions: For systems with noncompact or unbounded orbits (such as ℤ-extensions), one introduces fiber-induced pressure, which only counts configurations confined to bounded regions. This allows precise dimension-theoretic analysis of recurrent, transient, and escaping sets, including dimension gaps, with direct analogs in Kleinian group limit set theory (Gröger et al., 2019).

4. Applications and Computability

• Statistical inference and neuroscience: The formalism rigorously underlies spike-train modeling via maximum-entropy (MaxEnt) principles, providing a conceptual and practical bridge to statistical inference, especially in constructing Gibbs-type models for neuronal population data and extracting biophysical parameters from empirical averages (Cofré et al., 2020).
• Quantum dynamics: The framework has been adapted to dissipative quantum walks, enabling the analysis of dynamical order parameters, large deviation rate functions, and quantum generalizations of centrality measures such as PageRank. In this context, classical large deviation theory merges with quantum jump processes governed by Lindblad dynamics (Garnerone, 2012).
• Computable formalism: Algorithmic approaches to the calculation of equilibrium measures have been developed, delineating conditions under which these measures are computable, notably for non-uniformly expanding or rational dynamical systems with effective pressure/entropy evaluation and “local” Jacobian criteria (Binder et al., 1 Dec 2025).

5. Advanced Frameworks: Correspondences, Flows, Phase Transitions

• Correspondences and multi-valued dynamics: The formalism extends to set-valued maps (correspondences), incorporating transition probability kernels, generalized entropy, and pressure, with equivalence to classical pressure via a shift on orbit spaces. Uniqueness of equilibrium states still holds under specification and Bowen regularity or distance-expanding properties. This provides a foundation for thermodynamics in anti-holomorphic and holomorphic correspondences, including conformal and rational dynamical matings (Li et al., 2023).
• Suspension flows and time-changes: For flows over countable Markov shifts, the classical invariance of thermodynamic quantities under time-changes fails: entropy, pressure, and even the existence of equilibrium states can be destroyed by regular reparametrizations. The space of flows can be topologically stratified according to entropy and phase-transition behavior, with analytic construction tools enabling precise prescription of thermodynamic features (Cipriano et al., 2018).
• Transient potentials and Martin boundary: When the transfer operator is transient (sum over return probabilities converges), the formalism is governed by the Martin compactification, leading to families of conformal measures classified by escape direction, with a direct connection to non-uniqueness of thermodynamic (DLR) states and first-order phase transitions (Shwartz, 2017).

6. Spectral, Geometric, and Dimension Theory

• Conformal and Gibbs measures, spectral gap: Existence, uniqueness, and ergodicity of conformal and equilibrium (Gibbs) measures have been established in broad regimes, including entire and meromorphic transcendental maps with hyperbolic Julia sets and Baker domains (Esparza-Amador et al., 2023, Naderiyan, 7 Jun 2025). The Bowen formula relates the unique zero of pressure to the Hausdorff dimension of dynamically defined sets, allowing for multifractal and geometric analyses.
• Holonomic and correspondence-based approaches: Holonomic measures, extending the setting to skew-product and fibered systems, allow defining variational entropies and pressures in non-classical configurations, accommodating measure-theoretic constraints and non-additive potentials. These generalizations retain uniqueness, variational, and spectral results if the pressure functional is Gateaux-differentiable at the relevant point (Cioletti et al., 2017).

7. Geometric, Differential, and Information-Theoretic Aspects

• Manifold of normalized potentials: The space of potentials yielding the same Gibbs state is an analytic submanifold, and the Gibbs map is analytic. Gradient flows with respect to the weak Riemannian variance metric (derived from asymptotic variance) capture pressure and entropy optimization, supporting general constraint-maximization and Legendre duality (Giulietti et al., 2015).
• Multifractal theory and phase transitions: The formalism provides the basis for multifractal spectra, Lyapunov spectrum, and the detection of critical points and phase transitions (e.g., freezing transitions, first-order support jumps, non-analyticity of pressure).

In summary, thermodynamic formalism functions as a paradigm uniting statistical physics, dynamical systems, ergodic theory, and information theory. Its scope covers transfer-operator spectral analysis and variational principles in broad deterministic, random, geometric, symbolic, and quantum dynamical contexts, with intricate behavior arising from non-uniformity, non-compactness, time-changes, and multi-valued or stochastic dynamical rules (Cioletti et al., 2017, Pinheiro et al., 2022, Ovadia et al., 7 May 2025, Cofré et al., 2020, Zhang, 2015, Binder et al., 1 Dec 2025, Esparza-Amador et al., 2023, Ivrii et al., 2023, Raszeja, 2021, Shwartz, 2017, Peng et al., 2017, Jakobson, 2016, Naderiyan, 7 Jun 2025, Iommi et al., 2012, Cipriano et al., 2018, Giulietti et al., 2015, Li et al., 2023, Gröger et al., 2019, Garnerone, 2012).