Alec Kriebel

Quantitative persistence in a fixed-city area equation

· Zenodo record and DOI

In this deterministic model, a fixed set of cities competes for territory in the unit square. Each city's influence increases with its population share and decreases with distance. Its share moves toward the area it wins. The paper gives an explicit condition under which every city keeps a positive share for all time, with a quantitative lower bound for every positive starting state.

Original manuscript abstract

For finitely many fixed distinct interior sites in the unit square, let each site's positive population share grow at the rate of the area where its influence is maximal, minus its current share. Influence is proportional to population to the power α times distance to the power −β. We prove that every positive initial probability vector gives a global trajectory and, when β > 2α, every share has an explicit positive lower bound for all time. An asymmetric three-site instance with α = 2 and β = 8 has all shares between 1/4096 and 2047/2048. This disproves the unrestricted α > 1 collapse alternative in the literal 2007 fixed-city conjecture. The result is a quantitative, all-site persistence statement; it does not establish convergence or resolve the stochastic city-founding model. We credit the original local geometric mechanism and a directly overlapping 2026 two-city counterexample analysis. No firstness claim is made.

Read the paper

Open the original PDF · Download the PDF · BibTeX

Version 1.0.0 and 15 October 2026 are printed in the unchanged manuscript and recorded in the release metadata. The Zenodo record was created on 15 October 2026 UTC.

The fixed-city equation and persistence bound

Fix n ≥ 2 distinct sites xi in the interior of D = [0, 1]2, positive parameters α and β, and positive initial shares summing to one. City i wins the points where its influence is at least every competitor's:

Si(z) = {y ∈ D : ziα |y − xi|−β ≥ zjα |y − xj|−β for every j},
Ai(z) = area Si(z),   dzi/dt = Ai(z) − zi.

Ties and the finitely many sites do not affect the areas. For all positive α and β, the equation has a unique global positive classical trajectory from every positive initial probability vector.

For the quantitative bound, put p = 2α/β. Choose hi > 0 so that the square xi + [−hi, hi]2 lies in D and 8hi2 ≤ |xi − xj|2 for every j ≠ i. Set Ki = 4hi2.

Persistence theorem. If β > 2α, so p < 1, then every coordinate satisfies, for every t ≥ 0,

zi(t) ≥ [Ki + (zi(0)1−p − Ki) exp(−(1−p)t)]1/(1−p).

Writing ηi = min{zi(0), Ki1/(1−p)} > 0 gives

ηi ≤ zi(t) ≤ 1 − ∑j≠i ηj < 1,
lim inft→∞ zi(t) ≥ Ki1/(1−p).

The all-time floor depends on the initial shares; the positive limiting-inferior bound does not. These constants need not be optimal. A limiting inferior bound does not assert that a limiting state exists.

A protected neighborhood inside every winning cell

Taking a positive βth root rewrites the influence comparison as |y − xi| ≤ (zi/zj)α/β|y − xj|. These are classical multiplicatively weighted Voronoi cells. The geometry, and the full vector field, depends only on the ratio α/β, with no change of time.

The separation conditions ensure that a square of half-width hiziα/β around site i lies in its actual winning cell against every competitor. Its area gives Ai(z) ≥ Kizip. When p < 1, the standard substitution ui = zi1−p turns this into

dui/dt ≥ (1−p)(Ki − ui).

An integrating factor gives the displayed time-dependent bound. The paper also proves local Lipschitz regularity of the area map on the positive orthant, including equal weights. Mass stays one, and zi(t) ≥ zi(0)e−t prevents a finite-time loss of positivity, giving global continuation. These are analytic statements about the equation; a finite simulation does not prove them.

An exact asymmetric three-city instance

Take α = 2, β = 8 and

x1 = (1/5, 1/4),   x2 = (3/4, 1/3),   x3 = (2/5, 4/5),
z(0) = (1/6, 1/3, 1/2).

The sites form a scalene, noncollinear triangle. Choosing hi = 1/16 gives p = 1/2 and Ki = 1/64. Every share along the actual trajectory satisfies

1/4096 ≤ zi(t) ≤ 2047/2048   for i = 1, 2, 3 and all t ≥ 0.

Exact rational arithmetic checks the geometric hypotheses and constants. The theorem carries this conclusion through the entire trajectory, rather than only sampled times.

The literal 2007 conjecture and the overlapping work

Aldous's 2007 rough notes give this fixed-city equation. Their literal Conjecture 32(a) predicts collapse to one city when α > 1 or 0 < β < 2α, with an unspecified general-position proviso. The theorem refutes the unrestricted α > 1 alternative on the open region α > 1, β > 2α: every distinct interior configuration and every positive initial probability vector has every share bounded away from both zero and one.

Aldous also supplied the protected-neighborhood idea and the exponent 2α/β in an unfinished argument with stochastic and environmental conditions. Neither is presented here as new. Weighted Voronoi geometry, the scalar substitution and the integrating-factor step are standard.

Ferudun's unrefereed preprint of 14 October 2026 directly overlaps the negative answer. It explicitly gives robust two-city counterexamples in another parameter range. Its proved derivative and position-continuity lemmas also imply qualitative two-city noncollapse in the present β > 2α regime: for p < 1, the symmetric area-map derivative can be 7p/12 < 1. That conclusion receives credit even though its separate quantitative extension there is labeled a sketch.

The contribution here is quantitative persistence for all sites and all positive initial probability vectors. It is not advertised as the first negative answer, a new geometric mechanism or an exhaustive priority finding. Ratio invariance also makes the literal conjecture's two clauses incompatible in some cases, but that observation alone does not identify which clause fails; the persistence theorem directly excludes collapse.

What the theorem leaves unresolved

Persistence does not establish convergence, uniqueness of a positive equilibrium or independence of a limiting state from its starting point. No persistence theorem is claimed at β = 2α or β < 2α. Zero initial shares, coincident sites, boundary sites and the one-city case are outside the stated theorem. Positive-orthant uniqueness does not assert uniqueness on its closed boundary.

The later Aldous–Huang model allows stochastic founding of new cities and declares 0 < α ≤ 1. This fixed-city result does not refute that different process or resolve its behavior.

Original files and citation

The unchanged files appear in their native Zenodo order:

  1. city-ode-persistence-note.pdf (81,811 bytes) · Zenodo original PDF
  2. city-ode-persistence-verification.zip (568,282 bytes) · Zenodo original ZIP

Alec Kriebel. Quantitative persistence in a fixed-city area equation. Version 1.0.0, 15 October 2026. doi:10.5281/zenodo.23393139. Download BibTeX.

The native Zenodo record and authored release metadata specify Creative Commons Attribution 4.0 International (CC BY 4.0).

Status and collaboration

OpenAI tools were used extensively in solving, proof development, drafting, coding, reproduction and adversarial checking. This is an unrefereed preprint; completed external human peer review and formal proof-assistant verification are not claimed. The author is responsible for the claims.

Independent mathematical verification and substantive work toward a journal submission would be useful contributions. Collaboration and any authorship would reflect actual scholarly contributions.

Credited literature

  • D. J. Aldous, Technical notes on a spatial model of city growth (descriptive title), rough notes dated 25 April 2007, especially §6.4, Conjecture 32, p. 40, and the earlier local geometric mechanism. Author's notes.
  • F. Aurenhammer and H. Edelsbrunner, An optimal algorithm for constructing the weighted Voronoi diagram in the plane, Pattern Recognition 17 (1984), 251–257. Publisher DOI.
  • D. Aldous and B. Huang, A Spatial Model of City Growth and Formation, arXiv:1209.5120v1 (2012), §2.
  • A. Ferudun, On Conjecture 1 of Aldous for a Spatial Model of City Growth: A Proof of Part (a), and Counterexamples to Parts (b) and (c) for 2α/√3 < β < 2α, EulerSolve AMR-096-0023, version 1.0, 14 October 2026, unrefereed preprint, especially §4 and §8. Zenodo DOI.