Brownian web

Content sourced from Wikipedia, licensed under CC BY-SA 3.0.

The Brownian web is a universal collection of one-dimensional Brownian paths that start from every point in space and time and move forward, merging (coalescing) whenever they meet. It arises as the large-scale limit, under diffusive scaling, of many random walks that also merge when they meet. Imagine every point (x, t) in the plane launching a walker that takes random steps, and whenever two walkers meet they join into one walk. If you zoom out so the steps look like smooth Brownian motion, you get the Brownian web. The challenge is that there are uncountably many starting points, so describing and proving convergence to this object is highly nontrivial.

The idea goes back to Arratia, who was studying the voter model, a simple way to model how opinions spread through a population. In one dimension, if you start with coalescing random walks from finitely many points and scale them correctly, you get a finite number of coalescing Brownian motions. But for the web, you need the limit from every space-time point, which requires new mathematical tools.

To make the Brownian web precise, researchers defined a topology on the space of path collections, turning it into a well-behaved random object (a Polish space). This allows showing that the coalescing random walks converge to the Brownian web and to study its properties. There is also a related object called the Brownian net, which allows branching in addition to coalescing. It was introduced by Sun and Swart, with other constructions provided by Newman, Ravishankar, and Schertzer. For more on these ideas, see surveys by Schertzer, Sun, and Swart.