betaFeedback

WorksMathematics

The Falconer distance conjecture in all dimensions

Abstract

We resolve the Falconer distance conjecture in every dimension. For every integer d ≥ 2, a compact subset of ℝd with Hausdorff dimension greater than d/2d/2 determines a set of Euclidean distances of positive Lebesgue measure.

Introduction

For E⊂RdE \subset\mathbb{R}^{d}, its Euclidean distance set is

Δ(E)={∣x−y∣:x,y∈E}.\Delta(E)=\{|x-y|:x,y\in E\}.

The Falconer distance conjecture asks whether a compact set of Hausdorff dimension greater than half the ambient dimension must determine a set of distances of positive length. We prove this assertion.

Theorem 1.1. Let d≥2d \ge2 be an integer and let E⊂RdE \subset\mathbb{R}^{d} be compact. If dim⁡HE>d/2\dim_{H} E > d/2, then

L1(Δ(E))>0.\mathcal{L}^{1}(\Delta(E))>0.

Theorem 1.1 imposes no regularity beyond the strict Hausdorff dimension assumption. In particular, it does not require equality of Hausdorff and packing dimensions, an Ahlfors regular measure, product structure, or Fourier decay. The assertion concerns positive Lebesgue measure; a distance set of Hausdorff dimension one alone would not suffice. The strict threshold is retained, and no endpoint assertion is made.

History and scope

The problem originates in Falconer’s work [9], which established the sufficient threshold (d+1)/2(d+1)/2. His lattice-type examples at dimension d/2d/2 explain the strict inequality in the positive-measure formulation. Mattila’s spherical-average criterion [16] made Fourier L2L^{2} estimates a central approach to the problem. Bourgain connected improvements in dimensions two and three to Fourier restriction phenomena on spheres [2]. Among the subsequent milestones were Wolff’s planar threshold 4/34/3 [26] and Erdoğan’s threshold d/2+1/3d/2+1/3 for d≥3d \ge3 [8], Theorem 2}; these positive-measure conclusions require strict dimension inequalities. For the parallel finite point-set problem in Rd\mathbb{R}^{d}, d≥3d \ge3, see the distinct-distances theorem in [18], Theorem 1.1}.

Weighted restriction and decoupling led to further improvements. Du, Guth, Ou, Wang, Wilson and Zhang obtained the threshold 9/59/5 in dimension three and improvements in higher dimensions [4], Theorem 1.2}. Du and Zhang obtained the threshold d/2+1/4+1/(8d−4)d/2+1/4+1/(8d-4), improving the preceding bounds for d≥4d \ge4 [7], Theorem 2.6}. The arguments draw on refined restriction estimates and on the decoupling theory of Bourgain and Demeter [3]. For a pin yy, write Δy(E)={∣x−y∣:x∈E}\Delta_{y}(E)=\{|x-y|:x\in E\}. Liu’s L2L^{2} identity connects pinned distance measures with spherical extension estimates and transfers spherical-average bounds to pinned conclusions [13], Theorems 1.4 and 1.9}. The two preceding restriction advances also give L1(Δy(E))>0\mathcal{L}^{1}(\Delta_{y}(E))>0 for some y∈Ey\in E under the corresponding strict dimension inequalities; see [4], Corollary 1.1} and [7], Theorem 2.7}. For compact E⊂R2E\subset\mathbb{R}^{2} with dim⁡HE>5/4\dim_{H}E>5/4, Guth, Iosevich, Ou and Wang proved the same pinned conclusion [11], Theorem 1.2}. Their argument combines radial-projection control of discarded wave packets with refined decoupling for the retained part. Du, Iosevich, Ou, Wang and Zhang obtained the analogous pinned conclusion in even dimensions d≥4d\ge4 when dim⁡HE>d/2+1/4\dim_{H}E>d/2+1/4 [5], Theorem 1.2}. For compact E⊂RdE\subset\mathbb{R}^{d}, d≥3d\ge3, Du, Ou, Ren and Zhang obtained the same existential pinned conclusion when

dim⁡HE>d2+14−18d+4\dim_{H}E>\frac{d}{2}+\frac{1}{4}-\frac{1}{8d+4}

[6], Theorem 1.2}.

At the critical dimension, Shmerkin and Wang proved that dim⁡HΔ(E)=1\dim_H \Delta(E)=1 for Borel E⊂RdE \subset\mathbb{R}^d, d≥2d \ge2, whose Hausdorff and packing dimensions both equal d/2d/2 [24].

For Borel E,F⊂R2E,F \subset\mathbb{R}^2, Liu proved that dim⁡HE>1\dim_H E>1, dim⁡HE+dim⁡HF>2\dim_H E+\dim_H F>2, and dim⁡HF=dim⁡PF\dim_H F=\dim_P F imply L1(Δy(E))>0\mathcal{L}^1(\Delta_y(E))>0 for some y∈Fy \in F [14]. Here dim⁡P\dim_P denotes packing dimension. Taking F=EF=E gives the conjectured planar conclusion for sets with equal Hausdorff and packing dimensions greater than one. The regularity of the pin set is a hypothesis of this result, not a consequence of its Hausdorff dimension. Borges, Ou and Pasquariello study positive measure for pinned kk-star distance sets and admissible distance-graph configurations [1]. Their k=1k=1 threshold lies above d/2d/2; they record the sharper earlier one-edge bounds in [1]. These results do not give Theorem 1.1. The result here is unpinned; no corresponding claim about a single pin is asserted at the threshold d/2d/2.

Methodological antecedents

The multiscale analysis of irregular measures has important precedents. Keleti and Shmerkin decompose measures into dyadically regular pieces, control bad projections across scale intervals, and optimize variations of Lipschitz profiles in their work on the dimensions of planar distance sets [12]. These are antecedents for the regular classes, scale-dependent tests and profile accounting used below. Their estimates concern Hausdorff dimension. Their optimization controls a sum of interval drops; Section 10 uses a potential built from two separated depths to pay the present recursion’s costs. Shmerkin and Wang later combined scale profiles with radial projection estimates and slicing [24].

Our angular preparation is related to the multiscale incidence approach of Orponen and Shmerkin [20], the radial thin-tube bootstrap of Orponen, Shmerkin and Wang [21], and the higher-dimensional radial and plate-concentration methods of Ren [22] and Du, Ou, Ren and Zhang [6]. The retained incidence input is the theorem of Orponen and Shmerkin [20], in the tube formulation of [21]. The auxiliary radial estimate is proved here by Orponen’s finite-energy method [19]. It is applied only in a dimension where its energy hypotheses hold.

Auxiliary projections used to prepare radial information, while keeping the final distance map in the original space, also occur in [5]. The separation between approximation of a distance measure and Fourier control of its retained part has a precedent in [11]. Liu’s regular-pin theorem uses a multiscale good–bad decomposition and weighted extension estimates [14]. These comparisons describe relationships between methods, not additional theorem inputs. The present proof supplies its angular tests, oscillatory graph reductions, two-depth potential and monotone shell reconstruction; it does not invoke a decoupling theorem or any earlier distance theorem at an improved threshold.

The two main issues

Put S=d/2S=d/2. An immediate obstacle in higher dimensions is that radial projections of a separated ss-dimensional source cannot acquire dimension greater than ss. Thus, when S<s<d−1S<s<d-1, one cannot assume that these projections have densities relative to surface measure on Sd−1S^{d-1}. The first part of the proof constructs instead the cap bounds that the Fourier argument actually needs.

After restriction to two separated Frostman measures, we form decreasing sets of admissible pairs, denoted by ΓN\Gamma_N. Their intersection has positive product mass. At each scale 2−N2^{-N}, a further exceptional set of pairs of power-small mass can be deleted so that every remaining directional fiber has a cap bound of exponent SS, with an arbitrarily small power loss. Here a directional fiber is the opposite measure, restricted by the pair condition at a fixed endpoint and then pushed to the sphere by the radial map; it is not normalized separately. Integer exponents come from supercritical radial integrability after projection to a smaller space. In odd dimensions, a thin-tube improvement supplies the additional half dimension. Its linear gain is proved by slicing to the plane; the integer endpoint requires a separate transverse-triple argument.

The auxiliary projections concern directions and incidences only. Every distance and every oscillatory distance phase in the rest of the proof is taken in the original space Rd\mathbb{R}^{d}.

The second issue is to exploit these fractal angular laws in a distance estimate. A cap test controls parallel tubes, whereas a scalar projection test involves belts on the sphere. The required belt estimate follows from a Fourier calculation: the annular power is Bd−2SB^{d-2S}, which is exactly one at S=d/2S=d/2. This gives the angular tests used by the recursive reductions.

From profiles to a distance measure

At a fixed dyadic frequency R=2NR=2^{N}, decompose each Frostman measure into almost uniform dyadic classes. If a class has mass profile mm, write

An=m(n/N)−Sn/N.A_n=m(n/N)-Sn/N.

Thus its occupied depth-nn cells have mass approximately R−m(n/N)R^{-m(n/N)}, up to the controlled regularization loss. The excess profile AA records decay beyond the critical exponent SS. The two recursive quantities are a count of pairs with nearly equal distances, denoted by D, and a product of spherical Fourier energies with independent angular variables, denoted by H. The high-frequency part of D is split into distance-frequency shells. Stationary phase first produces the square of an angular inner product HFH_F, which Cauchy–Schwarz bounds by H. The graph estimate then uses the parallel tests to advance to finer cells. Other reductions return to positive distance counts or shorten the interval of scales. The definitions and the conversions appear in Sections 7–9.

An angular state carries tests selected on an interval containing the one currently being estimated. Each test is retained until the recursion reaches its associated depth. A potential formed from two values of AnA_n at sufficiently separated depths is evaluated on the current interval. It pays the positive step costs and closes the recursion.

For two profiles, let cc be the last depth at which either excess profile is at most a fixed small level β>0\beta>0. The graph estimate passes the independent angular energy to cells at depth cc. Cauchy–Schwarz in the radial variable bounds this by the geometric mean of two single-profile states, so the recursive bound applies. Each potential is at least 2β−Od(1/N)2\beta-O_d(1/N), while the entry cost is at most β\beta up to the regularization loss. The profile increments cancel, leaving a strictly decaying Fourier shell bound once the losses are chosen sufficiently small; see Lemma 11.1 and Proposition 11.3.

The coarse pair restrictions ΓN\Gamma_N need not converge at a quantitative rate. We therefore use the following form of the scalar shell principle. If positive measures αN\alpha_N decrease to α∞\alpha_\infty, if

∥αN−τN∥TV≲2−ηN,∫2N−1≤∣r∣≤2N+1∣τN^(r)∣2 dr≲2−γN,\lVert\alpha_N-\tau_N\rVert_{\mathrm{TV}}\lesssim2^{-\eta N}, \qquad \int_{2^{N-1}\leq|r|\leq2^{N+1}}|\widehat{\tau_N}(r)|^2\,dr\lesssim2^{-\gamma N},

and if η,γ>0\eta,\gamma>0, then α∞\alpha_\infty is absolutely continuous. The proof uses summable positive increments αn−αn+1\alpha_n-\alpha_{n+1} and telescoping smooth Fourier cutoffs; it is given in Lemma 11.4. Our limiting measure is nonzero and carried by Δ(E)\Delta(E), which proves Theorem 1.1 in Section 11.

The components that may be useful separately are the thin-tube upgrade (Proposition 5.4), the critical fractal angular Fourier estimate (Lemma 6.3), the graph estimate with independent angular variables (Proposition 8.1), and the monotone-measure shell principle (Lemma 11.4).

Figure 1 summarizes the roles of the auxiliary and original spaces.

Figure 1. The main construction. Auxiliary projections prepare directional bounds and pair filters; the cap bounds hold after an additional power-small pair deletion. These data return to the original space, where the argument uses the Euclidean distance of the original pairs. The decreasing filters and shell-dependent approximants meet in the final scalar monotone-shell argument.

Conventions and preliminary facts

Measures, scales and directions

All measures are finite Borel measures. A subprobability has mass at most one. Restrictions are left unnormalized unless normalization is explicitly specified. We write

Is(μ)=∬∣x−y∣−s dμ(x) dμ(y),μ^(ξ)=∫e−ix⋅ξ dμ(x).I_s(\mu)=\iint|x-y|^{-s}\,d\mu(x)\,d\mu(y), \qquad \widehat{\mu}(\xi)=\int e^{-ix\cdot\xi}\,d\mu(x).

For measures on the line we use the same Fourier convention. The norm ∥⋅∥TV\lVert\cdot\rVert_{\mathrm{TV}} is total variation. The notation f≲gf\lesssim g means f≤Cgf\leq Cg, with the dependence of CC specified by the ambient discussion.

A measure has an ss-ball bound with loss KK, down to scale δ\delta, if

μ(B(z,r))≤Krsfor every z and δ≤r≤1.\mu(B(z,r))\le Kr^s \quad\text{for every } z \text{ and } \delta\le r\le1.

A cap bound on a sphere has the analogous meaning for its Euclidean chordal metric. Passing between this metric, angular distance and fixed smooth charts changes constants only. We may use dyadic radii, and enlarge the constant to recover all radii in the specified interval.

For x≠yx\ne y, put

πx(y)=y−x∣y−x∣,D(x,y)=∣x−y∣.\pi_x(y)=\frac{y-x}{|y-x|},\qquad D(x,y)=|x-y|.

A tube is a neighborhood of an affine line; a plate is a neighborhood of an affine plane of the dimension indicated. Directions are oriented unless an antipodal pair of caps is explicitly used. All configurations lie in bounded sets. Fixed rescalings and a finite number of coordinate charts have constant cost.

The original ambient dimension is always denoted by dd, and S=d/2S=d/2. We use R=2NR=2^N for the reciprocal final scale. In odd dimension we write d=2k−1d=2k-1; the linear angular gain is then used in Rk+1\mathbb{R}^{k+1}, while S=k−1/2S=k-1/2 still refers to the original dimension. Letters denoting dimensions inside a preliminary lemma are local to that lemma.

Finite energy and projection

We record elementary consequences of the usual Frostman and projection principles, including the restriction operation needed later.

Lemma 2.1. Let μ\mu be supported in a bounded subset of Rd\mathbb{R}^d. (i) If μ(B(z,r))≤Crs\mu(B(z,r))\le Cr^s for 0<r≤10<r\le1, then It(μ)<∞I_t(\mu)<\infty for every 0<t<s0<t<s.

(ii) If It(μ)<∞I_t(\mu)<\infty, then for every ε>0\varepsilon>0 there is a compact restriction of mass at least μ(Rd)−ε\mu(\mathbb{R}^d)-\varepsilon that satisfies a tt-ball bound with a finite constant.

(iii) If 0<t<m≤d0<t<m\le d, It(μ)<∞I_t(\mu)<\infty, and PVP_V is orthogonal projection to a Haar-random mm-plane, then It((PV)#μ)<∞I_t((P_V)_\#\mu)<\infty for almost every VV.

Proof. For (i), split the integral at distance one and sum dyadic annuli. The contribution at distance at most 2−j2^{-j} is bounded by C2−j(s−t)C2^{-j(s-t)}, up to a constant depending on the total mass.

For (ii), let Utμ(x)=∫∣x−y∣−t dμ(y)U_t^\mu(x)=\int|x-y|^{-t}\,d\mu(y). This is finite for μ\mu-almost every xx, so a sufficiently large level MM discards arbitrarily little mass. Choose a compact subset K⊂{Utμ≤M}K\subset\{U_t^\mu\le M\} losing at most the remaining prescribed mass. If K∩B(z,r)≠∅K\cap B(z,r)\ne\varnothing, choose xx in this intersection. Then

μ∣K(B(z,r))≤μ(B(x,2r))≤(2r)tUtμ(x)≤M(2r)t.\mu|_K(B(z,r))\le\mu(B(x,2r))\le(2r)^tU_t^\mu(x)\le M(2r)^t.

The same bound is immediate if the intersection is empty.

For (iii), rotational invariance and the integrability of the negative tt-moment of the length of an mm-dimensional projection give

∫G(d,m)∣PVz∣−t dV≤Cd,m,t∣z∣−t(z≠0).\int_{G(d,m)}|P_Vz|^{-t}\,dV\le C_{d,m,t}|z|^{-t}\qquad(z\ne0).

For m<dm<d, this follows by integrating the density of the first mm coordinates of a uniform unit vector near zero: the radial integral there is bounded by a constant times ∫01rm−1−t dr<∞\int_0^1r^{m-1-t}\,dr<\infty. For m=dm=d, the identity projection makes the assertion immediate. Integrating in (x,y)(x,y) and applying Tonelli proves the claim. □

References

  1. [1]Tainara Borges, Yumeng Ou, and Marcus Pasquariello, From weighted paraboloid restriction to k-stars and distance graphs, 2026, Preprint, arXiv:2607.10574v1, submitted 12 July 2026.arxiv.org/abs/2607.10574
  2. [2]Jean Bourgain, Hausdorff dimension and distance sets, Israel Journal of Mathematics 87 (1994), no. 1-3, 193–201, Preprint consulted: IHES/M/92/61, August 1992.DOI
  3. [3]Jean Bourgain and Ciprian Demeter, The proof of the l² decoupling conjecture, Annals of Mathematics 182 (2015), no. 1, 351–389.arxiv.org/abs/1403.5335
  4. [4]Xiumin Du, Larry Guth, Yumeng Ou, Hong Wang, Bobby Wilson, and Ruixiang Zhang, Weighted restriction estimates and application to Falconer distance set problem, American Journal of Mathematics 143 (2021), no. 1, 175–211, Preprint version consulted: arXiv:1802.10186v1, 27 February 2018.arxiv.org/abs/1802.10186
  5. [5]Xiumin Du, Alex Iosevich, Yumeng Ou, Hong Wang, and Ruixiang Zhang, An improved result for Falconer’s distance set problem in even dimensions, Mathematische Annalen 380 (2021), no. 3-4, 1215–1231, Preprint version consulted: arXiv:2006.06833v2, 30 March 2021.arxiv.org/abs/2006.06833
  6. [6]Xiumin Du, Yumeng Ou, Kevin Ren, and Ruixiang Zhang, New improvement to Falconer distance set problem in higher dimensions, 2024, Preprint, arXiv:2309.04103v2, 22 October 2024; first submitted 8 September 2023.arxiv.org/abs/2309.04103
  7. [7]Xiumin Du and Ruixiang Zhang, Sharp L² estimates of the Schrödinger maximal function in higher dimensions, Annals of Mathematics 189 (2019), no. 3, 837–861.DOI
  8. [8]M. Burak Erdoğan, A bilinear Fourier extension theorem and applications to the distance set problem, International Mathematics Research Notices 2005 (2005), no. 23, 1411–1425, Author version consulted, dated 22 February 2005.
  9. [9]K. J. Falconer, On the Hausdorff dimensions of distance sets, Mathematika 32 (1985), no. 2, 206–212, Published version.DOI
  10. [10]Gerald B. Folland, Real analysis: Modern techniques and their applications, 2 ed., Pure and Applied Mathematics, John Wiley & Sons, New York, 1999.
  11. [11]Larry Guth, Alex Iosevich, Yumeng Ou, and Hong Wang, On Falconer’s distance set problem in the plane, Inventiones Mathematicae 219 (2020), 779–830, Preprint version consulted: arXiv:1808.09346v1, 28 August 2018.arxiv.org/abs/1808.09346
  12. [12]Tamás Keleti and Pablo Shmerkin, New bounds on the dimensions of planar distance sets, Geometric and Functional Analysis 29 (2019), 1886–1948.DOI
  13. [13]Bochen Liu, An L²-identity and pinned distance problem, Geometric and Functional Analysis 29 (2019), no. 1, 283–294, Preprint version consulted: arXiv:1802.03050v3, 23 June 2018.
  14. [14]———, Lebesgue measure of distance sets with regular pins and multi-scale Mizohata–Takeuchi-type estimates, 2026, Preprint, arXiv:2603.15328v3, 16 July 2026; first submitted 16 March 2026.arxiv.org/abs/2603.15328
  15. [15]L. H. Loomis and H. Whitney, An inequality related to the isoperimetric inequality, Bulletin of the American Mathematical Society 55 (1949), no. 10, 961–962.
  16. [16]Pertti Mattila, Spherical averages of Fourier transforms of measures with finite energy; dimensions of intersections and distance sets, Mathematika 34 (1987), no. 2, 207–228.DOI
  17. [17]———, Geometry of sets and measures in Euclidean spaces: Fractals and rectifiability, Cambridge Studies in Advanced Mathematics, vol. 44, Cambridge University Press, 1995.
  18. [18]OpenAI, The higher-dimensional Erdős distinct-distances conjecture, OpenAI Math Release preprint OAI:The-higher-dimensional-Erdos-distinct-distances-conjecture-September-23-2026, 2026.
  19. [19]Tuomas Orponen, On the dimension and smoothness of radial projections, Analysis & PDE 12 (2019), no. 5, 1273–1294, Preprint version consulted: arXiv:1710.11053v3, 31 July 2018.DOI
  20. [20]Tuomas Orponen and Pablo Shmerkin, On the Hausdorff dimension of Furstenberg sets and orthogonal projections in the plane, Duke Mathematical Journal 172 (2023), no. 18, 3559–3632, Version consulted: arXiv:2106.03338v3, 5 September 2023.arxiv.org/abs/2106.03338
  21. [21]Tuomas Orponen, Pablo Shmerkin, and Hong Wang, Kaufman and Falconer estimates for radial projections and a continuum version of Beck’s Theorem, Geometric and Functional Analysis 34 (2024), no. 1, 164–201, Preprint version consulted: arXiv:2209.00348v1, 1 September 2022.arxiv.org/abs/2209.00348
  22. [22]Kevin Ren, Discretized radial projections in Rᵈ, 2023, Preprint, arXiv:2309.04097v1, 8 September 2023.
  23. [23]Pablo Shmerkin, A non-linear version of Bourgain’s projection theorem, Journal of the European Mathematical Society 25 (2023), no. 10, 4155–4204, Published version.DOI
  24. [24]Pablo Shmerkin and Hong Wang, On the distance sets spanned by sets of dimension d/2 in Rᵈ, Geometric and Functional Analysis 35 (2025), no. 1, 283–358, Preprint version consulted: arXiv:2112.09044v3, 13 August 2024.
  25. [25]Leon Simon, Introduction to geometric measure theory, Tsinghua lecture notes, 2014, Author’s December 2014 revision.
  26. [26]Thomas Wolff, Decay of circular means of Fourier transforms of measures, International Mathematics Research Notices 1999 (1999), no. 10, 547–567.

Paper details

vlnc:W_rq39bk9f
Version
v1 · latest
Published
License
LicenseRef-OpenAI-public
Usage

views · 15

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

unreleased internal OpenAI model

View statement
Looking for related papers…
Contents