In July, I wrote about Falconer’s distance set problem in the plane, focusing on a breakthrough by Larry Guth, Alex Iosevich, Yumeng Ou, and Hong Wang. OpenAI’s October 2026 mathematics release includes a preprint (OpenAI 2026) proving the full conjecture in every dimension. The theorem has also been formalized in Lean.
Let us recall the problem. For a compact set , its distance set is where is the Euclidean distance between and . In the finite Erdős distinct distances problem, we ask how many different distances a set of points must determine. Falconer’s question is about a different kind of size: when must the set of distances have positive Lebesgue measure, that is, positive length?
The size of itself is measured by its Hausdorff dimension, denoted by . To define it, cover by countably many balls of radii and take the infimum of over all such covers. The Hausdorff dimension is the infimum of the exponents for which this quantity tends to zero as . It can be fractional: a compact set in the plane can have Hausdorff dimension strictly between and while having zero area.
In 1985, Kenneth Falconer (Falconer 1985) proved that guarantees that has positive length. He also constructed a compact set of dimension whose distance set has length zero. The idea behind his example comes from a grid. A -dimensional grid with points along each axis and spacing contains points, but every squared distance is an integer multiple of between and . Thus there are at most distinct distances. If we replace each grid point by a ball of radius much smaller than , these distances spread into intervals of small total length. Falconer’s construction turns this observation into a fractal example by using grids at successively finer scales.
Could a set of dimension even slightly greater than still determine a set of distances of length zero? Falconer’s conjecture predicted that it could not. In the plane, Guth, Iosevich, Ou, and Wang lowered the sufficient threshold to (Guth et al. 2020). In dimensions , Du, Ou, Ren, and Zhang brought it down to (Du et al. 2023). The new result reaches itself as the threshold.
Theorem. Let be an integer and let be compact. If , then has positive Lebesgue measure.
Thus, in the plane, a compact set of Hausdorff dimension already determines a set of distances of positive length. The inequality in the theorem must be strict, as Falconer’s example shows. The conclusion uses all pairs of points in . Asking whether a single point determines a positive-length set at the same threshold is a stronger question.
One way to understand the proof is to imagine sampling pairs of points and recording their distances. The resulting distribution might, in principle, be concentrated on a set of length zero. The proof keeps a positive mass of pairs while controlling how strongly their directions can cluster. It then uses Fourier estimates at successive scales to control pairs with nearly equal distances. The strict inequality gives enough room to combine these estimates and obtain a nonzero measure on with a density with respect to length.