A new mathematical preprint finds that removing diffusion from a confined model of moving and tumbling components can leave a separate store of mass at the boundary. In the limiting description, the left-moving component includes a concentrated mass at the edge, while the right-moving component approaches the corresponding bulk solution. The result comes from a one-dimensional family of run-and-tumble models on the half-line, not from observations of animals, people or physical particles.
The paper, an arXiv preprint labeled version 1 and dated 20 August 2026, examines whether nonlinear boundary layers—the near-edge parts of the model that must be resolved before diffusion is removed—can be treated rigorously. It asks what boundary rule emerges in the zero-diffusivity limit and how quickly the diffusive system approaches that limit.
The edge becomes part of the model
The central finding is a measure-valued limit, a mathematical description that can keep track of both spread-out bulk concentration and mass concentrated at a point. As κ tends to zero, the right-moving measure approaches the inviscid bulk measure, while the left-moving limit contains the inviscid bulk measure plus a Dirac mass at the boundary, weighted by the boundary-mass variable b−(t).
That boundary contribution cannot be determined from the inviscid equations alone. The paper says the nonlinear boundary law must be recovered by examining the diffusive boundary layer before diffusion is taken away. With zero boundary velocity and nonlinear tumbling, the boundary mass follows a nonlinear ordinary differential equation: it loses mass through a term involving the outgoing end state c∞ and gains mass from the incoming bulk concentration c− at the boundary.
The limiting description therefore includes a boundary-mass variable and a dynamic boundary condition, so the edge has its own state as the bulk evolves. The result is conditional on the theorem’s assumptions and on the maximal interval for which the solution exists.
A quantitative estimate with a narrow scope
The analysis does more than identify the form of the limit. It gives a Fortet–Mourier error bound of order κ^(1−θ), for θ in (0,1] and for times T before the stated maximal existence time. The Fortet–Mourier norm is a way of comparing distributions through how they act on well-behaved test functions. The theorem provides a power-law estimate for the gap between the diffusive and limiting descriptions, although the constant hidden in that estimate is not numerically specified.
In the paper’s second case, the matched-asymptotic approximation has an L1 error bounded by the same κ^(1−θ) power over the stated time interval and under the stated solvability assumptions. The approximation combines an outer, bulk solution with a rescaled boundary-layer solution. The remaining stability estimate is closed with Grönwall’s inequality.
The inviscid bulk problem is also shown to have a unique nonnegative solution on a maximal existence interval under the paper’s assumptions. The construction combines characteristic methods for transport, contraction mapping for semilinear terms and iteration for the quasilinear drift. The result is local in time: the analysis does not establish global existence for every permitted initial condition or parameter choice.
A numerical example points to competing states
The paper also displays a numerical bifurcation example for the incoming boundary-layer ordinary differential equation. The example shows two saddle-node bifurcations and an intermediate regime with two stable solutions. That pattern is consistent with possible phase-transition-like changes and hysteresis, meaning that the state reached could depend on the direction in which a parameter is varied.
But this part is exploratory numerical evidence, not a proof that the full limiting model has hysteresis. The convergence analysis assumes a selected solution branch for the incoming boundary-layer equation, while the suggested non-uniqueness of that equation is not rigorously resolved. The supplied text also does not report numerical uncertainty, solver details, discretization details or replication information.
What the result does not settle
The rigorous convergence theorem covers the constant-tumbling case or the case in which K(0,y)=0. It also requires the boundary transport signs to remain valid and does not cover switching, pinning, depinning or grazing scenarios. These restrictions define the range in which the stated boundary-layer analysis applies.
The models are one-dimensional, and the result does not establish the corresponding theorem for the higher-dimensional Doi–Saintillan–Shelley system. A hydrodynamic kernel discussed later in the paper is outside the technical coverage of the convergence theory because the required Lipschitz condition fails. The paper identifies switching, non-unique incoming branches, non-saturated tumbling, broader stability assumptions and higher-dimensional active-suspension models as open directions.
The study is a theorem-based and numerical contribution within idealized deterministic PDE and ODE models. It supplies no experimental or observational validation in humans, animals or real particle systems, and it does not establish a causal claim about biological active-matter behavior. Its relevance is chiefly to researchers studying singular limits, boundary layers and theoretical active matter.
The document is an arXiv preprint, version 1, dated 20 August 2026. The supplied record reports no conflicts of interest.
Paper data and sources
Original title: Boundary layers and vanishing diffusivity in run-and-tumble models
Authors: Dallas Albritton, Laurel Ohm, Timur Yastrzhembskiy
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text