A large fractal must determine many distances

← All digests

Family 073 · Real and complex analysis
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-14

1. The question

Place a collection of points in Euclidean space and measure every distance between two of them. How large must the collection be before its possible distances occupy a positive amount of the real line?

For a solid ball, the answer is immediate: every distance from zero up to the diameter occurs. The interesting case is a fractal, which can have zero ordinary volume while still having substantial dimension. Falconer’s conjecture predicts that a fractal occupying more than half the ambient dimension already determines a set of distances of positive length.

The manuscript in family 073 claims this conclusion in every dimension at least two, with no regularity assumption beyond compactness and the strict Hausdorff-dimension threshold. Its conclusion concerns distances between all pairs. It does not claim that one fixed point sees a positive-length distance set at the same threshold.

2. Background and essential definitions

Three meanings of size

For $E\subset\mathbb R^d$, define its distance set by

$$\Delta(E)=\{|x-y|:x,y\in E\}.$$

Here $|x-y|$ is the ordinary Euclidean distance. The set $\Delta(E)$ lies in $[0,\infty)$, regardless of the dimension of $E$.

Lebesgue measure on the real line is length. A set of positive length need not contain an interval, but it cannot be covered by intervals whose total lengths are arbitrarily small. This is stronger than saying that the set is uncountable or has Hausdorff dimension one.

Hausdorff dimension measures how efficiently a set can be covered at small scales. For a proposed exponent $s$, cover $E$ by sets of diameter at most $\delta$ and add their diameters raised to the power $s$. The threshold at which the best possible sums change from large to arbitrarily small is $\dim_H E$. A smooth curve has dimension one, a surface has dimension two, and a fractal can have a noninteger value.

Packing dimension is another small-scale size notion. It can exceed Hausdorff dimension for irregular sets. Equality of the two is a meaningful extra hypothesis, not a consequence of knowing only $\dim_H E$.

Turning a fractal into a distribution of mass

A useful bridge from dimension to analysis is a Frostman measure. If $s<\dim_H E$, one can put a probability measure $\mu$ on compact $E$ such that

$$\mu(B(x,r))\le C r^s.$$

The ball $B(x,r)$ has center $x$ and radius $r$. The inequality says that no small ball captures too much of the mass. It is an upper bound; it does not say every ball centered on $E$ carries mass comparable to $r^s$.

Choose two independent points according to $\mu$, and record their distance. The resulting distance measure $\nu$ is characterized by

$$\int g(t)\,d\nu(t)=\iint g(|x-y|)\,d\mu(x)\,d\mu(y)$$

for continuous test functions $g$. Its support lies in $\Delta(E)$. If a nonzero such measure has a density with respect to ordinary length, then the distance set must have positive length. This converts a geometric question into a question about the concentration of a measure.

3. Development history and previous partial results

Falconer’s 1985 paper, On the Hausdorff dimensions of distance sets, proved the conclusion when $\dim_H E>(d+1)/2$. Lattice-based constructions explained why half the ambient dimension is the natural lower threshold in the conjecture.

Mattila developed a criterion involving spherical averages of the Fourier transform of a measure on $E$. This connected distance geometry with oscillatory cancellation: a measure sufficiently spread out in frequency can force its distances to be spread out on the line.

Wolff reached the planar threshold $4/3$, and Erdoğan obtained $d/2+1/3$ in dimensions at least three. Later weighted restriction and decoupling methods improved these bounds further. They exploit the way frequency pieces pointing in different directions interact.

Guth, Iosevich, Ou and Wang proved a stronger, pinned conclusion in the plane when $\dim_H E>5/4$: some point of $E$ sees a positive-length distance set. Du, Ou, Ren and Zhang obtained the corresponding conclusion in dimensions $d\ge3$ above

$$\frac d2+\frac14-\frac1{8d+4}.$$

Other advances approached the conjecture through regularity assumptions or weaker conclusions. Shmerkin and Wang established full dimension of the distance set for certain critical-dimensional sets whose Hausdorff and packing dimensions agree. Liu’s regular-pin theorem reaches the planar positive-measure conclusion above dimension one when the relevant pin set has equal Hausdorff and packing dimensions.

These distinctions explain the remaining gap addressed by the manuscript. Full dimension need not mean positive length, and a result assuming equal Hausdorff and packing dimensions does not cover an arbitrary compact fractal.

4. What the new manuscripts claim

The principal theorem is

$$E\subset\mathbb R^d\text{ compact},\quad d\ge2,\quad \dim_H E>\frac d2 \quad\Longrightarrow\quad \mathcal L^1(\Delta(E))>0.$$

The symbol $\mathcal L^1$ denotes one-dimensional Lebesgue measure. No product structure, Fourier decay, equality of dimensions, or uniform lower mass estimate is required.

For example, in the plane a compact set of Hausdorff dimension $1.01$ satisfies the hypothesis. In three-dimensional space the threshold is strictly greater than $1.5$; in four dimensions it is strictly greater than two. These examples illustrate the stated theorem, not additional regularity conclusions about the fractal.

The manuscript constructs a nonzero absolutely continuous measure carried by the distance set. Absolute continuity means that every set of length zero has measure zero for this new measure. This is the mechanism by which the claimed analytical estimates imply positive length.

The theorem does not assert a quantitative lower bound for that length depending only on dimension. Rescaling $E$ already changes all its distances, so any quantitative statement would need additional normalization and geometric information.

5. Why the advance matters

Importance: 5/5. If correct, the theorem resolves a central problem linking fractal geometry, harmonic analysis and incidence geometry. The advance is the sharp strict dimension threshold for arbitrary compact sets in every dimension, rather than an improvement restricted to a particular family of fractals.

Distances are among the simplest nonlinear geometric measurements. A universal theorem about them shows that enough small-scale complexity in a set forces a substantial range of outcomes under that measurement. The threshold is particularly striking because the original set can occupy no ordinary volume at all.

The proposed methods also address a recurring difficulty in analysis: information may be good at each scale without being uniformly good along one fixed limiting object. Preserving enough positive mass while combining estimates across infinitely many scales is relevant well beyond the distance problem itself.

6. The main idea

Separate directions from distances

Start with two separated portions of a Frostman measure. Separation avoids the singular geometry of pairs whose endpoints nearly coincide. The direction from $x$ to $y$ is

$$\frac{y-x}{|y-x|}\in S^{d-1},$$

where $S^{d-1}$ is the unit sphere. The distribution of these directions controls how pairs concentrate in narrow tubes.

A subtle point arises in higher dimensions. A measure supported on a set of dimension just above $d/2$ cannot automatically acquire a smooth density on the $(d-1)$-dimensional sphere. The paper instead seeks upper bounds on the mass in small spherical caps. These are weaker than smoothness, but are tailored to the later Fourier estimates.

Auxiliary projections into lower-dimensional spaces help prepare these directional bounds. The actual distance $|x-y|$ remains the original Euclidean distance in $\mathbb R^d$. Projected distances are not substituted into the final conclusion.

Track how mass changes with scale

The manuscript divides space into successively smaller dyadic cubes, whose side lengths halve at each step. It groups portions of the measure with similar mass behavior and records a profile describing the mass in occupied cubes at each depth.

The useful information is the profile’s excess over the critical dimension $d/2$. That excess may appear unevenly across scales. The described recursion alternates between counting pairs with almost equal distances and controlling angular Fourier energies. A bookkeeping quantity involving two separated depths is used to pay for the losses in these alternations.

The strict hypothesis $\dim_H E>d/2$ supplies a positive margin. According to the manuscript, this margin eventually gives decay on successive frequency shells, meaning bands of frequencies with magnitudes comparable to powers of two.

Keep the limiting measure nonzero

At successive scales, the argument discards exceptional pairs with badly concentrated directions. Discarding independently at every scale would risk losing all the useful mass or changing the object being estimated without control.

The paper organizes a decreasing sequence of positive measures with a nonzero limit. It combines that monotonicity with approximating measures whose errors and Fourier-shell contributions are summable. Its stated reconstruction lemma then yields absolute continuity of the limit.

The overall strategy therefore has two complementary requirements: enough oscillatory cancellation to prevent concentration of distances, and enough retained positive mass to make the eventual conclusion substantive. Neither requirement alone would establish the theorem.

7. Scope, limitations, and open questions

The inequality in the dimension hypothesis is strict. The manuscript makes no universal positive-length assertion at $\dim_H E=d/2$.

A pinned distance set is

$$\Delta_x(E)=\{|x-y|:y\in E\}.$$

The claimed theorem concerns the union over all pairs and does not prove that one fixed $x$ has $\mathcal L^1(\Delta_x(E))>0$ at the conjectured threshold. Earlier pinned results above larger thresholds remain stronger in this different respect.

Positive length also does not guarantee an interval of distances, every sufficiently small distance, or a prescribed finite configuration inside $E$. Such conclusions require their own arguments. Finally, the theorem uses Euclidean distance; replacing it with an arbitrary norm or a distance on a curved space changes the problem.

8. Sources and further reading

The manuscript’s introduction compares its claim with the positive-measure, pinned and full-dimension results separately. Useful primary references are the planar pinned theorem, the higher-dimensional improvement, and Shmerkin and Wang’s critical-dimension work.

The Hausdorff dimension and Frostman’s lemma articles offer optional background for the size and mass notions used above.

Source manuscripts

• M1. OpenAI, The Falconer distance conjecture in all dimensions (September 23, 2026).

Sources consulted on 14 October 2026 at repository revision 1581d4c. Download the manuscript citations.