The study of how well rational numbers (fractions) approximate points on curves is a classical problem in number theory. Specifically, researchers want to count how many fractions with denominators up to some size Q land very close to the graph of a given function. A long-standing intuition says that this count should simply reflect the area of a thin strip around the curve, meaning the rational points scatter roughly uniformly near it. For smooth, well-behaved curves, mathematicians confirmed this intuition rigorously over decades of work, culminating in precise results from the 2000s and 2010s that required the curve to be at least twice differentiable.
This paper asks what happens when the curve is as irregular as possible, specifically when it is a typical path of Brownian motion, the famously jagged random process that models everything from particle diffusion to financial prices. Brownian motion is continuous but nowhere differentiable, placing it far outside the regime covered by previous results. The authors prove that, for almost every Brownian path, the count of nearby rational points still follows the area prediction across nearly all parameter ranges. The only exception is a narrow critical window where the behavior remains open. Crucially, they first had to establish how the area of the thin strip around a Brownian path grows, which is itself a delicate geometric problem because the path's wild oscillations cause the strip to overlap itself in complicated ways.
Beyond resolving a question that traces back to work by Sprindzhuk in 1979, the paper points toward something conceptually exciting. The accuracy of the area-based count turns out to be tied to the local roughness of the curve, measured by quantities studied in a field called multifractal analysis. This suggests that how densely rational points cluster near a curve is not just a number-theoretic question but is fundamentally linked to the curve's fine geometric texture. The authors propose that this connection could seed a unified theory bridging number theory and the analysis of fractal-like irregularity, two areas that have largely developed independently.