Genealogical Trees Subuniverse

• Genealogical trees subuniverse is a framework integrating real trees, ultrametric spaces, and combinatorial classes to model ancestral structures across populations.
• It employs universality classes and stochastic processes, such as Kingman and Beta coalescents, to characterize genealogical patterns under varying offspring distributions.
• It bridges theory and practice via algebraic decompositions and formal database schemes, enabling precise enumeration and modeling of complex genealogical structures.

The genealogical trees subuniverse encompasses the comprehensive mathematical, probabilistic, and algorithmic frameworks for representing, characterizing, and analyzing the ancestral structure of populations, incorporating the full range of models from classical Galton–Watson processes and Lévy trees to complex database schemes for real-world data storage. This subuniverse is structured by its foundational objects (real trees, ultrametric spaces, TOM trees), its associated stochastic processes (coalescents, genealogical Markov chains), universality classes, algebraic decompositions, and formal schemes for encoding genealogy.

1. Foundational Objects and Representations

The state spaces underlying the genealogical trees subuniverse are:

• Real Trees and TOM Trees: A real tree $(\tau, d)$ is a complete metric space with uniqueness of geodesic arcs and no cycles. A Totally Ordered Measured Tree (TOM tree) $((\tau, d, \rho), \le, \mu)$ extends this by equipping the real tree with a total order (compatible with geneaological ancestry), a distinguished root $\rho$, and a diffuse Borel measure $\mu$ such that every nontrivial left-interval carries positive mass. TOM trees precisely formalize the genealogies arising in splitting tree and continuum branching models, capturing planarity and chronological structure (Lambert et al., 2016).
• Ultrametric Measure Spaces ($U$-space): A compact ultrametric measure space $(X,r,\mu)$ consists of a complete separable metric space $(X,r)$ satisfying

$$r(x,z)\le\max\{r(x,y), r(y,z)\} \quad \forall x,y,z,$$

with $\mu$ a finite Borel measure. Isomorphism is defined via measure-preserving isometries. Equipped with the Gromov–Prokhorov or Gromov–weak topology, $U$ becomes Polish and is fundamental to the algebraic and probabilistic analysis of random genealogies (Gloede et al., 2016, Grieshammer, 2019).

• Genealogical Tree Lattices and Combinatorial Classes: For finite samples, the combinatorial subuniverse is stratified by plane/non-plane, labelled/unlabelled, ranked/unranked tree shape classes (e.g., ranked histories, cladograms, permutation trees). The space $\mathcal{MT}_N$ of rooted, ranked, unlabeled multifurcating trees is of particular interest in models allowing non-binary ancestry, equipped with a natural lattice structure by edge-collapsing operations (Zhang et al., 12 Jun 2025, Wiehe, 2020).

2. Universality Classes and Coalescent Structures

A central insight in the genealogical trees subuniverse is the emergence of distinct universality classes for sample genealogies depending on the underlying stochastic process:

• Critical Galton–Watson Trees (Universality via Tail Exponent $\alpha$): For offspring distributions satisfying

$$f(s)=s+(1-s)^\alpha\ell(1/(1-s)), \qquad \alpha\in(1,2],$$

with $\ell$ slowly varying, the full subuniverse of sample genealogies for large $T$ collapses to exactly two classes (Harris et al., 2023): - $\alpha=2$: Finite-variance, Kingman class. The sample genealogy is a mixture of time-changed Kingman coalescents with only pairwise mergers. - $\alpha\in(1,2)$: Infinite-variance, Beta class. The scaling limit is a $\Lambda$-coalescent with

$\Lambda(dx)=(\alpha-1)x^{\alpha-2}\,dx,$

permitting simultaneous multiple mergers, characterized by Dirichlet–Lauricella laws and explicit densities for split times and giant birth events.

• Varying Environment GW Models: Even in critical Galton–Watson processes with generation-dependent offspring laws, after a deterministic time-change based on cumulative variance, the genealogy of a large sample always converges to the Kingman coalescent. The topology is purely binary, the split times are mixtures of independent random variables, and this universality is robust under broad regularity conditions (Harris et al., 2022).
• Non-neutral Population Models (SMC, Genetic Algorithms): Under sufficiently rapid mixing (mutation dominates selection), the entire subuniverse of genealogical trees for resampling-based models (e.g., Sequential Monte Carlo, genetic algorithms) converges after time-rescaling to the standard Kingman coalescent. Deviations require infinite-variance offspring distributions or persistent fitness inheritance, otherwise the binary-coalescent structure is universal (Koskela et al., 2024).

3. Algebraic Decompositions and Infinite Divisibility

The genealogical subuniverse admits a rich algebraic structure:

• Concatenation Semigroups for Ultrametric Spaces: For each $h>0$, ultrametric measure spaces admit a commutative topological semigroup structure under $h$-concatenation, where any forest is uniquely (up to order) a concatenation of irreducible $h$-trees (representing maximal subfamilies at kinship $h$). This semigroup is Delphic and underpins all infinite-divisibility representations (Gloede et al., 2016).
• Infinite Divisibility and Lévy–Khintchine Representation: Random genealogies (as $U$-valued objects) are infinitely divisible if every $h$-top can be represented as the concatenation of i.i.d. $h$-forests. The Laplace transform of an infinitely divisible law satisfies

$$-\ln \mathbb{E}\left[e^{-\Phi_h(U)}\right] = \int_{U(h)^\sqcup\setminus\{0\}} (1-e^{-\Phi_h(u)})\,\lambda_h(du),$$

for a unique family of consistent Lévy measures $\lambda_h$, yielding a Poisson cluster (PPP) decomposition at each depth (Gloede et al., 2016).

• Family Size Decomposition and Homeomorphism: The map assigning to each ultrametric measure space the path of family sizes at each depth (by taking maximal disjoint $h$-balls) is perfect onto its image and homeomorphic on dense “identifiable” subspaces, providing a complete reconstruction of the genealogy in processes such as the Fleming–Viot (Grieshammer, 2019).

4. Genealogical Tree Subuniverses in Stochastic Processes

Various branching processes induce distinct genealogical trees subuniverses:

• Splitting Trees and TOM Trees: The class of continuum genealogies arising from splitting trees is exactly the class of (locally compact) TOM trees with the splitting property. Their contour functions are excursions of spectrally positive Lévy processes, and the celebrated backbone (prolific-skeleton) decomposition in the supercritical regime assembles infinite genealogical lines with grafted subcritical compact splitting trees (Lambert et al., 2018, Lambert et al., 2016).
• Crump–Mode–Jagers Explosive Trees: In non-homogeneous branching processes admitting explosion (finite time to infinite population), the genealogical tree at explosion can (under sharp criteria) contain either a unique “star” node of infinite degree or a unique infinite path. The dichotomy is controlled by asymptotic properties of birth times and offspring rates (Iyer et al., 2023).
• Branching Models with Competition: In models with general (possibly density-dependent) competition, the height and mass of the genealogical forest are finite if and only if certain integrals of the competition function converge. Thus, the subuniverse is either bounded (finite genealogical complexity) or unbounded (potentially infinite historical depth and complexity) depending on biological parameters (Le et al., 2014).

5. Combinatorics and Lattice-Structured Tree Spaces

• Enumeration and Generating Functions: Binary, multifurcating, and imbalance-constrained trees are rigorously classified, with explicit enumeration via recurrences, generating functions, and lattice posets. For instance, the space of rooted, ranked, multifurcating tree shapes with $N$ leaves forms a bounded lattice under edge-collapsing, facilitating MCMC exploration and inference for non-binary coalescent models (Zhang et al., 12 Jun 2025, Wiehe, 2020, Disanto et al., 2013).
• Statistical Invariants Under Constraints: Even strong combinatorial constraints such as bounded node imbalance (Ω-trees) preserve first and second moments (e.g., expected number of cherries $(n+1)/3$ and variance $2(n+1)/45$) found in unconstrained Yule trees, emphasizing the robustness of certain summary statistics across subuniverses (Disanto et al., 2013).

6. Formal Data Models and Implementation Frameworks

• Abstract Database Schemes: The genealogical trees subuniverse admits formal encoding in database theory via object-type sets (COUNTRIES, CITIES, DYNASTIES, TITLES, RULERS, MARRIAGES, REIGNS) and structural functions, equipped with a suite of integrity constraints (age, acyclicity, event temporal consistency, etc.) formulated as Horn clauses for enforcement at the application or database level (Mancas et al., 15 Jan 2026).
• Translation to Relational Schemes: The Algorithm M–R provides a systematic process to translate abstract mathematical genealogy schemes into relational tables and procedural business rule enforcement, with guarantees of soundness, completeness, and optimality. This bridges the gap between abstract genealogy models and robust, schema-constrained data representations required for empirical and historical applications (Mancas et al., 15 Jan 2026).

7. Synthesis and Significance

The genealogical trees subuniverse, in all its formalizations, comprises a precisely delineated collection of spaces and probability laws for genealogies arising in stochastic population models. It is stable under truncations and algebraic operations, admits universal or near-universal large-sample limits, and supports exact combinatorial enumeration and efficient data modeling. Structures such as TOM trees, ultrametric semigroups, coalescent process universality classes, and relational database encodings serve not merely as tools but as canonical building-blocks delineating the possible landscape of genealogies in both mathematical and empirical contexts (Harris et al., 2023, Gloede et al., 2016, Lambert et al., 2016, Zhang et al., 12 Jun 2025, Mancas et al., 15 Jan 2026).