Measuring SLE curves beyond their Hausdorff dimension

← All digests

Family 230 · Probability and statistical mechanics
Importance 4/5 · Provisional editorial rating
Sources consulted 2026-10-11

1. The question

Knowing a fractal’s dimension does not always tell us how to measure it. A curve can have dimension $d$, yet receive zero measure when every small covering set is charged its diameter to the power $d$. A slightly different cost may be needed to obtain a finite, nonzero notion of size.

For Schramm–Loewner evolution, or SLE, the dimension is known, and the critical-power Hausdorff measure is known to vanish. Schramm asked whether a deterministic correction could give a meaningful Hausdorff measure of the trace.

Family 230 contains two proposed answers for every fixed $0<\kappa<8$. The first constructs a gauge from deterministic moment integrals without asserting a simple asymptotic formula. The second claims an explicit power multiplied by an iterated-logarithm correction. It also claims that the larger correction suggested by Schramm gives a measure too large even to be sigma-finite on a nontrivial compact positive-time segment.

The issue concerns arbitrary spatial covers. It is distinct from measuring a curve through its time order or through the area of a thin neighborhood.

2. Background and essential definitions

The random curve

Chordal $\mathrm{SLE}_\kappa$ in the upper half-plane runs from zero to infinity. In the source’s time convention, the conformal maps removing its growing hull satisfy

$$\partial_tg_t(z)=\frac2{g_t(z)-W_t}, \qquad W_t=\sqrt\kappa B_t,$$

where $B_t$ is standard Brownian motion. This is half-plane capacity $2t$ parametrization. Capacity time describes how the growing curve changes a conformal map, rather than measuring its ordinary length.

Write the curve as $\gamma$, and its whole trace as $\Gamma=\gamma([0,\infty))$. For $0<\kappa<8$, its almost-sure Hausdorff dimension is

$$d=1+\frac\kappa8, \qquad 1<d<2.$$

Hausdorff gauges and exact measurement

A Hausdorff gauge is a continuous nondecreasing function $h$ with $h(0)=0$ and $h(r)>0$ for $r>0$. For a set $E$, define

$$\mathcal H^h(E)=\lim_{\delta\downarrow0} \inf\left\{\sum_jh(\operatorname{diam}U_j): E\subseteq\bigcup_jU_j,\quad \operatorname{diam}U_j\leq\delta\right\}.$$

The cover can be countable, and its sets can be arbitrary. One chooses the cheapest sufficiently fine cover and then sends the maximum allowed diameter to zero. This digest uses diameters, matching the manuscripts’ convention.

The ordinary $d$-dimensional Hausdorff measure uses $h(r)=r^d$. An exact gauge for the segments in question gives a measure strictly between zero and infinity almost surely. Exactness does not mean that every segment has the same numerical measure.

A measure is sigma-finite on a set if that set can be covered by countably many measurable pieces of finite measure. This is weaker than finite total measure. Failure of sigma-finiteness is therefore stronger than simply saying that one whole segment has infinite measure.

3. Development history and previous partial results

Beffara determined the SLE dimension $1+\kappa/8$. In his overview and collection of problems, Schramm asked for a sigma-finite Hausdorff measure and suggested the gauge $r^d\log\log(1/r)$.

Rezaei subsequently proved that the uncorrected $r^d$ Hausdorff measure is zero. This sharpened the problem: the missing ingredient is a refinement at the critical exponent, not a new dimension.

A parallel line developed a natural length for SLE. Lawler–Sheffield constructed natural parametrization in an initial parameter range, Lawler–Zhou extended it throughout $0<\kappa<8$, and Lawler–Rezaei identified it with $d$-dimensional Minkowski content up to normalization. Minkowski content measures the area within distance $r$ of the trace, rescaled by $r^{d-2}$.

That natural measure can be nonzero even when $\mathcal H^{r^d}$ vanishes. The definitions use different optimization procedures: Minkowski content uses a uniform neighborhood thickness; Hausdorff measure can choose a mixture of spatial scales adapted to the realized set.

Holden and Yuan established a fine-mesh variation characterization of natural parametrization using ordered partitions of the curve. Its logarithmic correction is different again. The family explicitly does not identify its Hausdorff measure with that variation or with Minkowski content.

Multipoint Green-function bounds and higher moments of natural length also have established precedents, notably work of Lawler–Werness and Rezaei–Zhan. M1 uses moment information to construct an exact gauge; M2 seeks quantitative tail and dense-visit estimates strong enough to produce a regular closed formula.

4. What the new manuscripts claim

An existence theorem through a moment-defined gauge

M1 claims that for each fixed $\kappa\in(0,8)$, there is a deterministic gauge $h_\kappa$, depending only on $\kappa$, such that almost surely

$$0<\mathcal H^{h_\kappa}(\gamma([s,t]))<\infty \qquad\text{for every real }0<s<t<\infty.$$

The same gauge works for every sample outside one null event and for every such interval simultaneously. It is not selected after seeing a particular curve.

The construction uses moments of a canonical trace mass in a disk, specified by deterministic integrals. These moments determine finite batches of thresholds and spatial radii. The gauge takes a full infimum over the resulting quadratic covering costs. Its scales may be irregular; M1 does not claim a log-log asymptotic for this gauge.

It also claims finite expected gauge measure for the entire trace in every bounded box of the upper half-plane, including the trace’s real-boundary points. That controls all visits to the box together, rather than merely summing separate finite-time bounds.

An explicit exact correction

M2 gives a separate argument for the small-radius formula

$$h(r)=r^d\bigl(\log\log(1/r)\bigr)^{(2-d)/2}.$$

Choose radii small enough that the expression is positive and increasing, then complete it continuously and nondecreasingly at larger radii. That completion does not affect Hausdorff measure, which uses arbitrarily small covers.

For this explicit gauge, M2 claims the same simultaneous positive-finite conclusion for all $0<s<t<\infty$. For every deterministic radius $R<\infty$, it further claims

$$\mathbb E\,\mathcal H^h(\Gamma\cap B(0,R))<\infty.$$

The expectation covers the whole infinite-time trace inside the disk. It implies almost-sure local finiteness on a countable exhaustion by bounded regions. The positivity statement itself is stated for nontrivial compact positive-time segments.

For example, at $\kappa=4$, one has $d=3/2$ and the correction exponent is $1/4$. At $\kappa=6$, the dimension is $7/4$ and the exponent is $1/8$. These are much smaller corrections than multiplying by one full power of $\log\log(1/r)$.

The exponent-one candidate is not sigma-finite

Let

$$g(r)=r^d\log\log(1/r)$$

at small radii. M2 claims that $\mathcal H^g$, restricted to any of the positive-time segments above, is not sigma-finite.

The source explains this through gauge comparison. Since $(2-d)/2<1$, the ratio $h(r)/g(r)$ tends to zero. Every set of finite $\mathcal H^g$-measure consequently has zero $\mathcal H^h$-measure. A countable collection of such sets cannot cover a segment with positive $\mathcal H^h$-measure.

Thus the explicit claimed answer both supplies a working correction and rules out the proposed exponent-one correction. It does not state that M1’s moment gauge is asymptotic to M2’s formula; their asymptotic relationship remains unspecified.

5. Why the advance matters

Importance: 4/5. The family addresses a long-standing geometric question for the full non-space-filling SLE parameter range. It would refine dimension into an exact spatial measure and determine an explicit iterated-logarithm exponent.

SLE describes scaling limits of many planar interfaces. A finer understanding of its trace measure therefore enriches the continuum geometry underlying those models, even though transferring such a measure back to a lattice requires additional work.

The result also clarifies why several plausible notions of fractal length differ. Critical-power Hausdorff measure, corrected Hausdorff measure, Minkowski content, and ordered variation respond differently to dense local visits and the freedom to choose covering scales.

The rating reflects a major contribution to random fractal geometry, with scope narrower than a general theorem for all random curves or an automatic identification of every natural parametrization.

6. The main idea

To show positive Hausdorff measure, a useful strategy is to find a positive mass supported on the trace whose mass in every small ball is at most a constant times $h(r)$. Any cover must then pay enough total gauge cost to account for that mass.

To show finite Hausdorff measure, one needs the opposite kind of construction: economical covers. Regions where the curve visits unusually densely can be covered cheaply relative to the trace mass they contain. One selects such regions across scales and covers the missed residue separately.

M1 arranges both tasks using all finite moments of a local trace mass. Moderate thresholds give repeated opportunities for useful covering regions; much larger thresholds have summably small costs. The moments choose finite batches of tests, and a quadratic envelope makes their costs compatible with subdividing planar sets. The argument uses conditional restarting and angle mixing, not independence of successive curve pieces.

M2 makes the relevant tail quantitative. Put $p=2-d$. In normalized local coordinates, it claims a bound of the form

$$\mathbb P\{\text{local stopped mass}>u\} \leq C\exp(-cu^{2/p})$$

for large $u$, uniformly under the permitted stopped histories and mesh approximations. At spatial scales $r=2^{-k}$, a threshold proportional to $(\log k)^{p/2}$ makes exceptional mass losses summable. Since $\log k$ has the scale of $\log\log(1/r)$, this explains the correction in the candidate gauge.

For economical upper covers, the paper uses rare dense visits arranged along a serpentine ribbon. A ribbon with order $m$ strands supplies mass of order $m^p$, at a probability cost exponential in $m^2$. Repeated conditional opportunities across scales balance this cost against the number of trials. Selecting disjoint charging regions allows one whole-trace neighborhood measure to pay for the cover.

Separate boundary and cleanup estimates retain visits near the real axis and avoid overlooking points missed by the favorable tests. The lower and upper estimates together yield the claimed exactness; either one alone would leave half the measurement problem unresolved.

7. Scope, limitations, and open questions

The statements concern ordinary chordal SLE in the upper half-plane for each fixed $0<\kappa<8$. The probability-one event can depend on that fixed parameter; no simultaneous assertion over an uncountable range of $\kappa$ is made. The space-filling endpoint and larger parameters are excluded.

The gauges use spatial diameters and arbitrary countable covers. Their correction should not be substituted into an ordered-variation formula without another theorem. Neither manuscript identifies its Hausdorff measure with natural parametrization, and M2 does not establish the asymptotic equivalence of the two constructed gauges.

The all-time bounded-region estimate gives local finiteness, not finite total measure of the unbounded trace. No convergence theorem for lattice step counts is asserted here. Such questions require a normalization and a separate discrete-to-continuum argument.

The remaining relationships among these measures, their exact proportionality constants where applicable, and extensions to other SLE variants are distinct questions beyond the source claims.

8. Sources and further reading

Schramm’s problem collection, Beffara’s dimension theorem, and Rezaei’s zero-measure theorem frame the question. Lawler–Rezaei and Holden–Yuan explain the related but different notions of natural length.

M1 below supplies the moment-based existence construction; M2 supplies the explicit gauge and its comparison with the exponent-one candidate. Both mathematical claims are reported as stated.

Source manuscripts

• M1. OpenAI, An exact Hausdorff gauge for SLE: A moment-integral and finite-batch construction (September 25, 2026).
• M2. OpenAI, An explicit exact Hausdorff gauge for SLE: A regular formula from quantitative tails and dense visits (September 26, 2026).

Sources consulted on 11 October 2026 at repository revision fd4aeeb. Download the manuscript citations.