Near-one oriented-percolation flow from Bernoulli first-passage estimates
Read the paper (PDF) · Download PDF · Supporting ZIP · Zenodo record
Plain-English explanation · Read on this page · Collaboration
Abstract
For the square-lattice oriented Bernoulli edge-flow density, we derive 1 − v(p) ∼ √(2(1 − p)) as p ↑ 1, the asymptotic posed by Aldous and Lenderman. The argument is an explicit corollary of Basdevant–Gouéré–Théret’s 2021 strict-chain scaling and weighted-list estimate. The directed cut dual has zero-cost backward steps, so their ordinary first-passage theorem cannot be substituted directly. Grouping the two outgoing rare-edge marks into independent site marks gives the required event inclusion into their already evaluated counting sum. We supply this comparison, the finite directed-cut reduction and the order of limits. The asymptotic is presented as an attributed application of older estimates; no claim of novelty, historical firstness or present-day openness is made.
The idea in plain English
If closed edges are rare, the amount of flow lost in a large directed square grid is proportional to the square root of the fraction of closed edges. The note explains how to transfer a bound for increasing chains to this particular directed maximum-flow model.
The precise result
Let each north- or east-pointing edge in {0, …, n − 1}2 independently have capacity one with probability p and zero otherwise. Flow enters through the left and bottom boundaries, excluding the top-left and bottom-right sink corners, and leaves through the top and right boundaries. Paths are edge-disjoint; they may share vertices.
Write Vn for the maximum flow and v(p) for the deterministic limiting density Vn/(2n) stated on Aldous’s problem page. The note proves
limq ↓ 0 [1 − v(1 − q)] / √(2q) = 1.
The finite-grid limit is taken at each fixed closed-edge probability q = 1 − p, before q tends to zero. The result is an attributed corollary of earlier estimates, with the comparison to this particular flow model supplied in the note.
Why the result holds
A finite directed-cut identity turns maximum flow into a minimum-cost path problem. Its dual paths pay for forward crossings and can move backward at zero cost. The two outgoing rare-edge marks at a site can be grouped into a single independent site mark with probability 2q − q2.
The note compares the resulting path deficit with strictly increasing chains of those marked sites. Basdevant, Gouéré and Théret’s strict-chain scaling and weighted-list bound then control the loss. The ordinary undirected first-passage theorem alone does not give this comparison because its treatment of backward steps differs.
The source-stated deterministic L1 flow-density limit is an explicit input. The expectation limit used in the calculation also follows from the note’s finite reduction and diagonal superadditivity.
Scope and limitations
The conclusion concerns this edge-capacity model and these boundary conventions. It does not establish the same formula for vertex-disjoint paths, other boundaries or other graph families. Finite verification checks supplement the analytic argument; they do not prove the infinite-volume asymptotic.
No claim of novelty, historical firstness or present-day openness is made. The page records a deduction from older estimates, rather than a new central rare-defect estimate.
Background and prior work
- David Aldous’s problem page records the flow model and the asymptotic formulated with Jason Lenderman in 2004.
- Anne-Laure Basdevant, Jean-Baptiste Gouéré and Marie Théret, First-order behavior of the time constant in Bernoulli first-passage percolation (2022). The operative estimates are in their public 2021 arXiv v1, including §3.3, equation (33), and the strict-site score scaling. The entire asymptotic here is presented as a corollary of those estimates after the explicit directed-model comparison. The manuscript’s equation references concern that version, rather than a retrieved final journal body.
- The archive preserves a longer alternative argument using classical Poisson-chain results of Deuschel and Zeitouni, van den Berg and Kesten’s disjoint-occurrence inequality, and overlapping-region decomposition and endpoint-skeleton ideas also present in Verges’s 2025 work. No originality claim is made for those general mechanisms.
Status and AI use
Unrefereed preprint. AI tools assisted in developing, writing and checking the work. Independent human review is welcome.
Read the paper
Open the published PDF · Download PDF · Original PDF on Zenodo
Supporting materials
Download oriented-flow-verification.zip · Original supporting ZIP on Zenodo
Collaborate on this result
I welcome researchers in percolation, random flows or first-passage models who would like to independently check the directed-cut identity, the model-event comparison with the older weighted-list estimate, or the order of limits. A careful comparison with earlier statements of this particular flow consequence would also be useful. I am open to developing a revised manuscript and journal submission through substantive collaboration.
Authorship and responsibilities would be discussed in relation to the contributions made. Contact Alec.
Cite this paper
Kriebel, Alec (2026). Near-one oriented-percolation flow from Bernoulli first-passage estimates. Version 1.0.0. Zenodo. 10.5281/zenodo.23409451.
Licensing
The deposited work is available under Creative Commons Attribution 4.0.