MathLabs

Worked solution: The three-dimensional case: every Kakeya set in $\mathbb{R}^3$ has dimension 3 (Wang–Zahl, 2025)

Step 1 of 5: Reframing the conjecture as a volume bound, and the road to d=3d=3
In plain words

Discretize: let T\mathbb{T} be ≈δ−2\approx\delta^{-2} tubes of dimensions δ×δ×1\delta\times\delta\times 1 in R3\mathbb{R}^3, pointing in δ\delta-separated directions. Write K(d)K(d) for the statement that their union always has volume ≳δ3−d\gtrsim \delta^{3-d}. A single tube trivially gives K(1)K(1); a classical 1970s L2L^2 argument of Córdoba (the same idea behind the planar Kakeya theorem) gives K(2)K(2); Wolff's 1995 'hairbrush' argument gives K(2.5)K(2.5), and progress stalled there for three decades despite deep work by Bourgain, Katz, Tao, and others. Wang and Zahl's new preprint proves K(3−ε)K(3-\varepsilon) for every ε>0\varepsilon>0, which is exactly the statement that every Kakeya set in R3\mathbb{R}^3 has Hausdorff and Minkowski dimension 33.

∣⋃T∈TT∣≳δ3−d(K(d)),Goal: K(3−ε) for every ε>0\Big|\bigcup_{T\in\mathbb{T}} T\Big| \gtrsim \delta^{3-d} \quad (K(d)),\qquad \text{Goal: } K(3-\varepsilon) \text{ for every } \varepsilon>0
Detailed analysis

Wang and Zahl's paper (2502.17655, §1) states the main theorem as this volume bound for arbitrary δ×δ×1\delta\times\delta\times 1 tube families satisfying the 'Katz–Tao convex Wolff axioms' (a technical generalization of plain direction-separation needed to make the induction close). Terence Tao's expository blog post on the paper explains that the proof strategy is induction on scales: assume the bound K(d)K(d) holds at every scale for tube families of every size, and try to bootstrap it into K(d+α)K(d+\alpha) for some small α>0\alpha>0; iterating finitely many times then drives dd up to 3−ε3-\varepsilon for every ε>0\varepsilon>0.

Terms in this step
Induction on scales
A proof strategy, pioneered by Bourgain and Wolff, that tries to bootstrap K(d)⇒K(d+α)K(d) \Rightarrow K(d+\alpha) for some small α>0\alpha>0 by regrouping thin tubes into fat ones at an intermediate scale, applying the hypothesis at both the coarse and fine scale, and combining the results.