A mathematical preprint offers a way to estimate when a single modeled population begins to break into groups that are no longer freely connected by mating. Its central result is a closed-form matrix formula for the critical genome length, Lc, at which fragmentation begins in the finite-genome Derrida–Higgs model. The predicted threshold is not governed by one simple rule: at large population size with a fixed mutation rate, the leading length becomes effectively independent of population size and follows a mutation-controlled μ−2 scale. In regimes where transient genealogical variance—the temporary spread in relatedness among lineages—dominates, the length instead grows as M to the 3/2 power or as the square root of M divided by μ, depending on how M and μ change together.
That makes the work a mathematical baseline, not an observation of species formation in an organism. The document is an arXiv preprint, arXiv:2608.25995v1, dated 26 Aug 2026. Its conclusions are confined to the model, its moment equations and finite-range simulations.
The model’s mating rule
The model fixes the population at M and gives each member a haploid, biallelic genome of length L. In the species-formation version, reproduction uses two parents, recombination and mutation, but mating is allowed only when the pair’s overlap reaches qmin. Overlap is the model’s compatibility score: it determines whether two genomes can mate. The question is where genome length sits when those permitted pairings begin to fragment.
To find that point, the analysis follows the average overlap through time and the spread around it. At the crossing, it compares δq, the deterministic distance covered in one generation, with ΔL, the extra width associated with finite genome length. The proposed boundary is where those two quantities are equal. The variance bookkeeping separates v∞, the genealogical variance in the infinite-genome limit, from v1, the 1/L finite-locus correction; the authors describe this decomposition as exact rather than merely a large-L approximation.
The calculation starts from an unrestricted homogeneous reference and a full moment hierarchy, then reduces it to a smaller matrix system. The reduced calculation drops a higher-order variable called dt; the feedback it would provide, and the resulting effect on the critical length, are estimated at order M−2. This reduction produces the explicit matrix prediction while retaining a benchmark against the fuller hierarchy.
One threshold, several regimes
The asymptotic picture changes with the balance among population size, mutation and transient width. With M sufficiently large while μ stays fixed, the leading critical length loses its dependence on M and follows μ−2. When transient genealogical variance is the dominant contribution, the predicted length grows as M to the 3/2 power in one joint limit, and as the square root of M divided by μ in another. These are scaling laws for limiting regimes of the model, not empirical measurements of how real genomes determine species formation.
Near the boundary where qmin approaches q0, the model predicts a further change. Retaining the transient width produces a linear divergence in the critical length, rather than the quadratic divergence from the narrower treatment. In ordinary terms, the gap between the mating threshold and the model’s equilibrium overlap changes how quickly the required genome length rises.
Checks within the model
The reduced formulas were checked against the full moment hierarchy. At M = 1,000, the matrix result was almost indistinguishable from the hierarchy, with a relative error below 3 × 10−3%. The simpler scalar expression for Lc had an error below 0.09% at M = 1,000 and below 0.014% at M = 5,000. These are internal model benchmarks: they measure consistency within the calculation, not agreement with biological data.
Direct tests used the restricted species-formation model and estimated fragmentation from connected-component counts. Each run lasted 3τ generations, with each duration rounded up to a whole generation; the first 2τ generations were treated as transient, and counts were averaged over the final τ generations. Crossings were interpolated locally in log L. At each tested length, the design used ten pilot replicates and 20 refined-grid replicates, with 95% confidence intervals built from 1,000 nonparametric bootstrap samples.
Across the finite ranges tested, the simulations followed the main predicted trends: dependence on M was weak in the fixed-mutation regime, one width-dominated regime lay near the M-to-the-3/2 scaling, and the boundary behavior stayed near a −1 exponent rather than −2. But the available ranges did not fully reach the asymptotic limits, particularly in the fixed-mutation and boundary tests. The simulated crossing, Lsimc, is therefore an operational estimate tied to the run duration, tested length grid and definition of fragmentation, not an exact stationary species count.
A controlled result with clear limits
The scope is deliberately narrow. The framework uses a fixed-size population of haploid, biallelic genomes and a mating threshold; it does not derive an exact critical length from the full nonlinear restricted dynamics. The transient criterion examines width at one characteristic time, so it does not provide the full distribution of separation times.
The result is best read as a controlled mathematical baseline. The matrix expression is a closed-form prediction, and the scalar expression was highly accurate at the reported large population sizes. The checks support consistency within this model; they do not establish direct applicability beyond the stated setup or replace empirical evidence.
Paper data and sources
Original title: What sets the critical genome length for sympatric speciation? A closed form and asymptotic theory
Authors: Dan Braha, Marcus A. M. de Aguiar, Vitor M. Marquioni
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text