Kakeya conjecture

The Kakeya conjecture is a problem in geometry and analysis asking how small a set can be if it contains a line segment pointing in every direction. It states that any such set in n-dimensional Euclidean space must have full Hausdorff dimension n, even though it may have zero Lebesgue measure. The conjecture is central to geometric measure theory and harmonic analysis. It is settled in the plane (1971) and in three dimensions (2025) and remains open for dimensions four and higher.

Definitions

A Kakeya set in $\mathbb{R}^n$ is a set $K$ such that, for every direction $e$ on the unit sphere $S^{n-1}$, $K$ contains a unit line segment pointing in direction $e$. Many authors require $K$ to be compact, and some work with Borel sets; the choice does not change the conjecture in an essential way. A set of this kind is also called a Besicovitch set.

The conjecture is usually stated with Hausdorff dimension, and often also with Minkowski (box-counting) dimension. Since the Hausdorff dimension of a set never exceeds its Minkowski dimension, a result for Minkowski dimension implies the corresponding Hausdorff result, so the Minkowski version is the stronger statement.

The conjecture is a restatement of a classical question. Sōichi Kakeya asked in 1917 for the smallest area of a planar region in which a needle can be turned through a full rotation. Abram Besicovitch later showed that the area can be made arbitrarily small, which shows that such sets can have measure zero while still containing segments in every direction. The Kakeya conjecture asks how small these sets are in terms of dimension rather than measure.

Historical development

The planar case

In the plane ($n = 2$), the conjecture was proved by Roy Davies in 1971. His result shows that a Besicovitch set in $\mathbb{R}^2$ must have Hausdorff dimension 2. A few years later, Antonio Córdoba proved a quantitative version, known as the Kakeya maximal function estimate, which has become a standard tool in analysis.

Lower bounds in higher dimensions

Before the recent resolution, progress came mainly through lower bounds. Jean Bourgain introduced the "bush argument" in 1991 and showed that Kakeya sets in $\mathbb{R}^3$ have Hausdorff dimension at least $7/3$. He extended this method to higher dimensions by induction.

In 1995 Thomas Wolff introduced the "hairbrush argument", which gave a lower bound of $(n+2)/2$ for every $n$ and raised the three-dimensional bound to $5/2$. Nets Katz, Izabella Łaba and Terence Tao (1999) studied "sticky" Kakeya sets, which have approximate self-similarity across scales, and obtained a small improvement on the Minkowski dimension in $\mathbb{R}^3$. Later work by Katz and Tao improved the general lower bounds in dimensions five and above. Katz and Zahl then pushed the three-dimensional bound slightly above $5/2$, and gave a bound of about 3.059 for $n = 4$.

The three-dimensional resolution

In 2022, Hong Wang and Joshua Zahl proved the conjecture for sticky Kakeya sets in $\mathbb{R}^3$. In February 2025 they posted a long preprint, "Volume estimates for unions of convex sets, and the Kakeya set conjecture in three dimensions", which proves the full conjecture in $\mathbb{R}^3$. Their argument shows that every Kakeya set in $\mathbb{R}^3$ has Hausdorff and Minkowski dimension 3. The proof combines the sticky-set analysis with a multiscale study of thin tubes, and uses an induction on scales scheme in which a statement about unions of tubes is proved for all scales. The authors also developed a stronger statement about tube unions that is easier to prove by induction than the original conjecture. Terence Tao described the result as spectacular progress in the field.

Finite-field and p-adic analogues

The conjecture has analogues in other settings where the proofs are easier. In 2008 Zeev Dvir proved the finite-field version: any Kakeya set in $\mathbb{F}_q^n$, where $\mathbb{F}_q$ is a finite field with $q$ elements, has at least $\binom{q+n-1}{n}$ elements, which is asymptotically $q^n / n!$. The proof uses the polynomial method: a set that is too small admits a nonzero polynomial of low degree vanishing on it, and a dimension count rules this out in the required direction.

The analogue over the $p$-adic numbers $\mathbb{Q}_p$ has also been proved, showing that bounded subsets of $\mathbb{Q}_p^n$ containing a unit segment in every direction have full Hausdorff and Minkowski dimension $n$. Results of this kind are often seen as models for the real problem.

Variants and related problems

Several closely related questions are studied alongside the conjecture:

• Kakeya maximal function conjecture. This is a quantitative $L^p$ estimate for averages over tubes pointing in different directions. It implies the dimension statement and is considered a stronger, more analytic form of the problem. Wang and Zahl's work also bears on this version.
• Restricted Kakeya sets. These are Kakeya-type sets in which the midpoints of the segments are constrained to lie in a set of prescribed dimension. Recent work gives lower bounds for these restricted sets that depend on the dimension of the constraint set.
• Sticky Kakeya sets. These are sets with approximate multiscale self-similarity. Since they were central to early counterexample-type attempts, a proof for this class was an important step toward the general result.
• Assouad dimension. Wang and Zahl also studied the Assouad dimension of Kakeya sets in $\mathbb{R}^3$, a stronger notion of dimension measuring the worst local behaviour of a set.

Connections to other fields

The conjecture is linked to several major open problems. In harmonic analysis it is related to the Fourier transform-based restriction conjecture and to Bochner–Riesz summability, because estimates for Kakeya-type averages control how functions with Fourier support on a sphere behave. The method of induction on scales has also been used in other problems involving partial differential equations, such as wave propagation.

In combinatorics and incidence geometry, the problem is treated as a question about how families of tubes or lines can overlap. Techniques from additive combinatorics, such as sum-product estimates and the study of arithmetic structure, are important tools. Several results of this type are also used to study projections of fractal sets.

Status

As of 2026, the conjecture is confirmed for $n = 1, 2$ and $3$. For $n \ge 4$ it remains open, and the best known lower bounds are strictly below $n$, with the gap widening as $n$ grows. The three-dimensional proof is long and technical, and its detailed analysis continues to be studied and extended in survey articles and follow-up papers. Whether the methods developed for $\mathbb{R}^3$ will extend to higher dimensions is an active research question.

References

1. Sticky Kakeya sets and the sticky Kakeya conjecturearxiv.org
2. Hausdorff dimension of restricted Kakeya setsarxiv.org
3. arXiv:2108.03750v3 [math.NT] 1 Nov 2021 THE P-ADIC KAKEYA CONJECTUREarxiv.org
4. Dvir’s proof of the finite field Kakeya conjectureterrytao.wordpress.com
5. Mathematicians Move the Needle on Decades-Old Problemnyu.edu
6. The three-dimensional Kakeya conjecture, after Wang and Zahlterrytao.wordpress.com
7. Kakeya Set Conjecture in ℝ³emergentmind.com
8. Hong Wangen.wikipedia.org
9. A Survey of the Kakeya conjecture, 2000-2025 Joshua Zahl∗arxiv.org
10. Volume estimates for unions of convex sets, and the Kakeya setdeepoo.net
11. Mathematicians move the needle on the Kakeya conjecture, a decades-old geometric problemphys.org
12. Nonstandard Hausdorff dimension in $\mathbb{R}^n$ and an application to the Kakeya conjecture in $\mathbb{R}^3$arxiv.org