Curado–Tsallis Constraints in Non-Extensive Statistics

• Curado–Tsallis constraints are fundamental in non-extensive statistics, defining normalization and q-expectation to ensure consistent power-law equilibrium distributions.
• They employ a variational approach that maximizes Tsallis entropy under constant escort averages, effectively eliminating self-reference in the resulting distributions.
• These constraints underlie applications ranging from thermostatistics and quantum systems to cosmological models and causal inference, with experimental bounds closely constraining deviations from extensivity.

The Curado–Tsallis constraints underpin a broad generalization of statistical mechanics and information theory based on non-extensive entropy maximization. These constraints specify both how expectation values should be formulated in Tsallis statistics and the precise variational method leading to non-exponential (power-law) equilibrium distributions. They govern the statistical structure of ensembles, the construction of thermodynamic quantities, and the statistical description of systems where the canonical Boltzmann–Gibbs (BG) formalism breaks down—such as systems with long-range interactions, strong correlations, or anomalous transport. The constraints also provide a foundation for modified cosmological models and the data-driven inference of non-extensivity parameters in astrophysical and cosmological contexts.

1. Definition and Formulation of Curado–Tsallis Constraints

The Tsallis entropy for a probability distribution $\{p_i\}$ is defined as

$S_q = \frac{1 - \sum_i p_i^q}{q-1},$

where $q\in\mathbb{R}$ is the Tsallis index, with $q\to 1$ recovering Shannon (or Boltzmann–Gibbs) entropy (Shen et al., 2017).

In the Curado–Tsallis (CT) formalism, the constraints imposed in variational problems are:

• Normalization: $\sum_i p_i = 1$
• $q$-expectation (escort mean): $\sum_i p_i^q A_i = C$, or equivalently, for observable $A$,

$$\langle A \rangle_q = \frac{\sum_i p_i^q A_i}{\sum_j p_j^q}$$

The essential Curado–Tsallis constraint is the imposition that $\sum_j p_j^q$ is constant and independent of each $p_i$ during the maximization (Shen et al., 2017, Wong et al., 2021).

This ensures that the escort distribution

$P^{(q)}_i \equiv \frac{p_i^q}{\sum_j p_j^q}$

is properly normalized, and that all derived expectation values are self-consistent.

2. Variational Principle, Distribution Families, and Elimination of Self-Reference

Maximizing $S_q$ under the Curado–Tsallis constraints leads to the $q$-exponential equilibrium distributions,

$p_i \propto \exp_q(-\beta(\varepsilon_i - U_q)),$

where $\exp_q(x) \equiv [1 + (1-q)x]_+^{1/(1-q)}$, and $U_q$ is a fixed $q$-mean energy (Shen et al., 2017, Conroy et al., 2014, Wong et al., 2021).

The closed-form, non-self-referential property of the CT prescription arises from treating $\sum_i p_i^q$ as a constant, breaking the feedback loop present in alternative formalisms (e.g., the TMP/OLM method, where $\sum_i p_i^q$ appears inside $p_i$) (Shen et al., 2017).

This variational method generalizes directly to the grand-canonical ensemble, producing $q$-deformed Bose–Einstein and Fermi–Dirac distributions.

3. Implications in Thermostatistics, Information Geometry, and Quantum Extensions

Thermostatistics

Curado–Tsallis constraints define the statistical structure of non-extensive systems, admitting power-law equilibria rather than standard exponentials. The constraints are operationally equivalent to replacing ordinary arithmetic means with escort means, heavily weighting higher-probability events. This is vital in systems with non-local interactions, fractal phase space, or correlation-induced entropy anomalies (Conroy et al., 2014, Shen et al., 2017).

Information Geometry

The CT maximum entropy distribution forms a $\lambda$-exponential family, with $\lambda=1-q$, and admits a duality structure analogous to classical exponential families. $\lambda$-logarithmic divergences correspond to Rényi divergences, and both primal and dual coordinate systems are rooted in CT constraints, embedding non-extensive statistics within a coherent information-geometric framework (Wong et al., 2021).

Quantum Generalization

Quantum analogues involve replacing probability sums with traces, e.g., $S_q(\rho) = (1 - \operatorname{Tr}[\rho^q]) /(q-1)$, and imposing Curado–Tsallis constraints for both energy and particle number (Gonzalez, 2 Feb 2026, Gonzalez et al., 20 Nov 2025). This yields quantum $q$-deformed equilibrium distributions and establishes bounds for observables in quantum thermodynamics and early-universe cosmology.

4. Role in Maximum Entropy Modelling and Statistical Inference

Generalized MaxEnt under Tsallis entropy incorporates CT constraints and, in practical inference problems (density estimation, bias correction), leads to convex programs with Tsallis-based quadratic constraints (Hou et al., 2010). The Tsallis entropy bias (TEB)—the expected reduction in entropy due to sampling—is compensated through an explicit constraint on the candidate distribution's entropy, analytically ensuring unbiased inference.

The constraints also furnish a rigorous method for selecting smoothing parameters in Lidstone-type estimators by requiring the estimator to satisfy a TEB-corrected Tsallis entropy (Hou et al., 2010).

5. Applications in Cosmology, Astrophysics, and Causal Inference

Cosmological Models

Generalized Friedmann and Boltzmann equations with CT constraints describe non-extensive universes. The scaling exponent (often denoted $\delta$ or $\beta$) quantifies non-additivity in horizon entropy, leading to modified Hubble expansion laws. Observational constraints from Big Bang Nucleosynthesis (BBN), cold dark matter relics, CMB, and large-scale structure tightly restrict $|q-1|$ to $\lesssim 10^{-2}$–$10^{-5}$ (Jizba et al., 2023, Ghoshal et al., 2021, Gonzalez, 2 Feb 2026, Gonzalez et al., 20 Nov 2025, Mendoza-Martínez et al., 2024, Astashenok et al., 2024, Sadeghnezhad et al., 10 Dec 2025).

Astrophysical Relics and Dark Matter

During WIMP freeze-out or primordial element synthesis, CT constraints yield modified freeze-out conditions and relic densities, with model predictions confronted against Planck and CMB+BAO data to extract allowed $q$ intervals (Gonzalez et al., 20 Nov 2025, Jizba et al., 2023).

Causal Inference and Entropic Inequalities

In the analysis of classical and quantum causal structures, Tsallis entropies and mutual informations subject to CT constraints provide a generalization of Shannon-type entropic constraints. This establishes operational inequalities (e.g., dimension-dependent bounds on conditional mutual information) that discriminate classical, quantum, and supra-classical correlations in graphical models (Vilasini et al., 2019).

Summary Table: Operational Formulations of Core CT Constraints

Domain	Constraint Structure	Key Outcome
Statistical Physics	$\sum_i p_i = 1$, $\sum_i p_i^q A_i = C$	Non-self-referential $q$-exponential family
Quantum Systems	$\operatorname{Tr}\rho = 1$, $\operatorname{Tr}[\rho^q O] = C$	Quantum $q$-deformed statistics
MaxEnt Inference	$T[\bar P] \geq T[\widehat P] + \Delta T$	Bias-corrected density estimates
Cosmological Friedmann Laws	$S \sim A^\delta$, Friedmann eqs. w/ $q$-entropic factors	Modified expansion, $q$ from BBN/DM data
Causal Structure Analysis	$I_q(\cdot:\cdot|\cdot) \leq f(q,d_1,d_2)$	Classical/quantum constraint separation

6. Experimental and Observational Constraints on the CT Parameter

Multiple lines of investigation place stringent bounds on the allowed departure of $q$ (or related scaling exponents $\delta$, $\beta$, $\gamma$) from the extensive limit:

• BBN and Cold Dark Matter: $|q-1|\lesssim 10^{-2}$, with most likely values extremely close to unity, supporting an almost extensive early-universe plasma (Jizba et al., 2023, Ghoshal et al., 2021, Gonzalez et al., 20 Nov 2025).
• CMB, BAO, and SNe Data: Parameter scans in Tsallis-modified cosmologies restrict the non-extensive exponent to $|q-1|\lesssim 10^{-2}$–$10^{-3}$, with cosmological models fitting observational data only for minute deviations from extensivity (Gonzalez, 2 Feb 2026, Mendoza-Martínez et al., 2024, Sadeghnezhad et al., 10 Dec 2025, Astashenok et al., 2024).
• New Dark Energy Models: Embedding Tsallis entropy in extended holographic dark energy or Rastall gravity frameworks introduces new model-specific bounds but typically still requires $\delta$ very near unity (Sadeghnezhad et al., 10 Dec 2025).

7. Impact and Theoretical Significance

Curado–Tsallis constraints are foundational in structuring the formalism of non-extensive statistics. They not only guarantee internal consistency (elimination of self-reference; normalization of escort averages) but also connect deep mathematical frameworks (information geometry via $\lambda$-duality, non-additive entropy composition, generalized Cramér–Rao inequalities) to physically observable consequences in thermodynamics, cosmology, and inference theory. Their operational flexibility makes them central to ongoing efforts to characterize and test deviations from extensivity in laboratory, astrophysical, and cosmological systems.

References:

(Vilasini et al., 2019, Shen et al., 2017, Conroy et al., 2014, Wong et al., 2021, Jizba et al., 2023, Gonzalez et al., 20 Nov 2025, Gonzalez, 2 Feb 2026, Ghoshal et al., 2021, Mendoza-Martínez et al., 2024, Sadeghnezhad et al., 10 Dec 2025, Astashenok et al., 2024, Hou et al., 2010, Furuichi, 2010)