Bernstein Functions at Work: Coalescents, Copulas, and Subordination

Abstract: Several positivity questions in stochastic processes, dependence modeling, fractional analysis, and renewal theory reduce to a common recognition task: after normalization, identify the object as a Laplace transform, a potential density, an inverse-flow coefficient, or a finite kernel average, and then read the sign pattern from that representation. We develop this recognition calculus for completely monotone functions, Bernstein functions, special Bernstein functions, and probabilistic realizations through subordinators and mixing measures. The main affirmative results settle three narrowly stated source questions in the conventions used by their source papers. Möhle's Problem 6.3 on the block-counting process of exchangeable coalescents with residual singleton mass (dust) is proved by a finite-simplex ordered-pair kernel certificate. For the Pearse--Bondell power-divergence copula generators, we prove complete monotonicity of the inverse throughout the remaining strict negative range $λ\le-1$ identified in their Section 3.8. Together with the special cases already verified in the source paper, this yields Archimedean copulas in every dimension for $λ\le-1$. The Bendikov--Cygan monotonicity question for discrete renewal sequences attached to special Bernstein functions is answered by representing the potential kernel as a Gamma average of a nonincreasing density. Supporting representation and boundary results cover Sibisi's Prabhakar--Pollard $Q$-measure, the Mecke--Nagel--Weiss atom at zero, and the cubic branch criterion $a<sup>2\ge3b$.

• The paper establishes a unifying recognition calculus using Laplace transforms and Bernstein function properties to resolve open problems in stochastic analysis.
• It applies finite-simplex kernel certificates and sign-inductive arguments to prove complete monotonicity in coalescent processes and power-divergence copulas.
• The study provides explicit frameworks for subordination and renewal theory, delineating both regime characterizations and structural obstructions.

Bernstein Functions at Work: Analytical Recognition Principles and Probabilistic Applications

Overview

The paper "Bernstein Functions at Work: Coalescents, Copulas, and Subordination" (2607.04467) rigorously addresses positivity and monotonicity questions in stochastic and analytic settings via a unifying recognition calculus rooted in the theory of completely monotone functions, Bernstein functions, and their probabilistic representations. By leveraging canonical analytic-measure correspondences—Laplace transforms, subordinate processes, and mixing measures—the paper resolves several open problems in coalescent theory, dependence modeling via copulas, and renewal sequences attached to special Bernstein functions. The author systematically employs measure-theoretic, covariance-based, and sign-inductive arguments to link function-theoretic structure with probabilistic realizability, delineating both regime characterizations and structural obstructions.

Analytical Recognition Principle and Representations

The central principle asserts that, after suitable normalization, objects of interest (e.g., survival functions, moment sequences, subordination transforms) can be identified as Laplace transforms or mixtures thereof, enabling their sign patterns to be inferred directly from properties of the representing measures. The canonical representations are based on the Bernstein-Widder, Lévy–Khintchine, and Bochner subordination theorems, providing an explicit calculus for classifying complete monotonicity, the Bernstein property, and their special subclasses.

A paradigmatic example is given by the Prabhakar–Pollard Q-measure, whose Laplace transform is identified as an a-stable subordination of a normalized Pollard measure, yielding a direct proof of complete monotonicity in applicable parameter regimes (Theorem 3.2). This machinery is extended to formulate constructive solutions to Laplace-survival and Poissonization problems for random variables and count processes (Theorem 3.4), establishing explicit realizations within the framework of exponential race or interarrival constructions.

Positive Resolution of Source Problems

Möhle's Block-Counting Process

Theorem 4.2 provides a definitive answer to Möhle's Problem 6.3 on block-counting processes in exchangeable coalescents with residual singleton mass (dust). The main technical achievement is the finite-simplex ordered-pair kernel certificate for second-moment domination, showing that for any $n \geq 1$ and every point in the ranked simplex, the variance is uniformly below the mean, i.e., $E[Y(n,u)^2] \leq E[Y(n,u)]$. The proof leverages integral identities and ordered pair domination arguments rooted in the structure of multinomial occupancy and residual mass, relying fundamentally on the nonnegativity properties induced by simplex constraints.

Power-Divergence Copulas

For the Pearse-Bondell power-divergence copula generators [27], the paper closes the remaining gap in their complete monotonicity regime. Theorem 4.5 proves that the strict inverse of the power-divergence generator is strictly completely monotone for all parameters $\lambda \leq -1$. This extends the validity of the generator as an Archimedean copula in all dimensions via the Kimberling criterion, using explicit positive generalized power series flows and inductive sign arguments based on the inverse ODE structure. The authors detail explicit positivity of all intermediate exponents, ensuring the full range of parameters for valid multidimensional copulas.

Bernstein Structure and Renewal Monotonicity

Utilizing inverse-ODE sign induction, Theorem 5.1 characterizes the Bernstein branch of a cubic inverse-polynomial (arising in generalized dual relativistic-diffusion frameworks [22]) by the discriminant criterion $a^2 \geq 3b$, precisely delineating the analytic regime of positivity for such inverse-symbols.

Theorem 5.3 resolves the Bendikov–Cygan monotonicity question for discrete renewal sequences attached to special Bernstein functions, showing that the renewal sequence is always nonincreasing under natural normalization. The argument is built on expressing the renewal sequence as a Gamma average of a nonincreasing potential density, with covariance-based sign control rooted in the monotonicity of the underlying measure.

Boundary Results and Obstruction Certificates

The analysis includes explicit boundary cases and obstructions, demarcating the limits of the recognition principle and the class of objects admiting positive representations. Notably:

• Counterexample 6.1 (to Rastegar–Roitershtein) demonstrates the necessity of specific structural constraints in characterizations of exponential distributions by constructing a law with a valid source equation but lying outside the conclusion class.
• Counterexample 6.3 (to Jonckheere–Shneer) provides a non-completely-monotone survival function solving the distributional front equation, separating the class of completely monotone solutions from general nonincreasing ones.

Further, some pending "certificate targets" (e.g., in mixed Poisson and fractional diffusion-wave contexts) are identified but left for future work pending full symbolic or interval arithmetic verification.

Implications and Prospects

The methods unify treatment across disparate contexts—exchangeable coalescents, copula construction, subordination in stochastic processes, and renewal theory—by reducing functional and probabilistic questions to explicit measure representations and monotonicity analysis. The approach establishes broad applicability for finite-simplex kernel certificates in analyzing occupancy and concentration functionals, and for inverse-ODE sign induction in classifying implicit generator regimes for copulas and subordinators.

Future research may significantly advance systematic classifications of generator invertibility (particularly for rational and algebraic settings), discriminant-phase diagrams in subordination theory, and concrete analytic certificates for boundary and exceptional cases (such as the certificate targets outlined in the appendices). The combination of analytic recognition principles with constructive probabilistic realization positions this methodology as a fundamental toolkit for positivity, monotonicity, and dependence modeling in applications ranging from random partitions to high-dimensional dependence structures.

Conclusion

This paper synthesizes measure-theoretic, analytic, and probabilistic techniques into a systematic calculus for positivity and monotonicity in stochastic and functional analytic contexts. The affirmative theorems settle previously open conjectures and source problems, providing concrete tools for verifying and constructing completely monotone and Bernstein functions, with direct applications to coalescent processes and the construction of multidimensional Archimedean copulas. The explicit characterization of the analytical regime for Bernstein structure and the systematic boundary analysis demarcate both the scope and limitations of current recognition calculus, setting a solid foundation for future expansion in analytic probability and dependence modeling.