Hilbert transforms along Lipschitz directions

Hilbert transforms along any Lipschitz direction field in the plane obey one energy bound at short lengths.

1 paper September 25, 2026 Explained by AI

A directional Hilbert transform compares a function on the two sides of each point, along a line through it. Here each point of the plane picks its own direction, and the directions change at a controlled rate. This paper proves that such transforms never increase energy by more than a fixed factor, at short lengths (Lean proof listed). The factor does not depend on the field or on how small a window around the point is cut out. This answers Stein's question in its short-scale form.

First step

—

Significance

Major

If it holds, it answers a long-standing question of Stein that earlier work settled only in special cases.

Trust

Formal proof listed

Field

Real and complex analysis

1 paper

The explanation

Picture a straight line with a bright stripe painted on it, from position 0 to position 1. Stand at a point on the line. Add up the paint on one side, weighted by 1 over its distance. Subtract the same sum for the other side. The answer is the Hilbert transform at that point. It measures how lopsided the paint looks from where you stand. With no limit on distance, the answer at point x works out to the logarithm of x/(x − 1), ignoring sign. At the middle point 0.5 the two sides cancel exactly, giving 0. At 0.75 it is about 1.10. At 2 it is about 0.69. Near the stripe's edge it climbs: about 4.62 at 1.01, and about 9.21 at 1.0001.

That climb is the danger. Nearby paint gets a huge weight, so the sum depends on delicate cancellation between the two sides. On one fixed line, mathematicians have long controlled this well. The standard yardstick is energy, the total of the squared values over all points. A good bound says the output energy is at most a fixed multiple of the input energy.

Now move to a flat plane, and give every point its own direction. This rule is a vector field (an arrow of length 1 at each point). At each point, you take the Hilbert transform along the straight line through it, in that point's own direction. Different points use different lines. The field is Lipschitz with constant K when two arrows a distance d apart differ by at most K times d. For example, let the arrow at height y point at angle y radians. That field has constant 1.

Elias Stein asked whether such transforms keep a uniform energy bound. The question, in the form Lacey and Li wrote down, allows lines up to a length set by 1/K. It asks for a weak bound, which limits how much area can carry large output values. The constant must not depend on the field. Lacey and Li trace the question back to an older conjecture of Zygmund on averages along short segments.

Much partial progress came before. In 2006 Lacey and Li proved a bound for one frequency range at a time, for any measurable directions. Bateman and Thiele, in 2013, handled full transforms when the direction depends on only one coordinate. The paper calls that one-coordinate condition a substantive restrictionSource 1. Bourgain, and later Stein and Street, used the extra structure of analytic fields. Di Plinio, Guo, Thiele and Zorin-Kranich showed how bounds on single frequency bands reassemble into the full bound.

The obstacle is that the line moves with the point. Keeping the lines short limits how much the field turns along them. With K equal to 1 and length one millionth, the arrow turns at most about 0.0000573 degrees. Yet the paper stresses that this does not reduce the problem to frequency methodsSource 2. Lacey and Li also recall that smoothness slightly weaker than Lipschitz does not give the matching bound for the averaging version.

This paper proves a strong energy bound for every Lipschitz field in the planeSource 3. The field may depend on both coordinatesSource 4. The integration length runs up to a fixed absolute constant divided by K. The bound also holds uniformly when a tiny window around the point is cut out, however small that window isSource 5. A strong energy bound implies the weak bound Stein asked forSource 3. So, inside this scale, no Lipschitz field can make the output energy blow up.

The result stops at that scale. The paper separates it from estimates at arbitrary lengths for varying fields, which it does not giveSource 6. For a constant field, any length reduces to the one-line caseSource 7. The constants are not given as explicit numbers in the paperSource 8.

How sure is this? The Lean scope note says a computer proof checker called Lean has checked a formal version. It covers every cut-off range from ε up to R, with R at most 1/(1,000,000 K), on smooth rapidly decaying inputs. It also covers the weak bound at a fixed short scale for fields with constant 1. No specialist review has been published.

What it doesn't show. This does not give a bound at arbitrary lengths for fields whose direction varies.

Where it came from

This result comes from an unreleased OpenAI model, published in the openai/math release on 6 October 2026.

Hype check

The Lean scope note lists a formal proof of the short-scale bounds. The paper's constants are not explicit, and no specialist review has been published. It covers only short lengths.

Words used

Hilbert transform: A weighted difference between a function's values on the two sides of a point, with weight 1 over distance.
Vector field: A rule that attaches an arrow, here of length 1, to every point of the plane.
Lipschitz: Changing at a controlled rate: two values a distance d apart differ by at most a fixed constant times d.
Energy (L² norm): The total of a function's squared values, which measures its overall size.
Weak bound: A limit on how much area can carry output values above a given level.
Truncation: Cutting the integral to distances between an inner and an outer length.

The original paper

A uniform Hilbert transform estimate for Lipschitz directions (PDF, opens in a new tab)
September 25, 2026

Lean notes

All quoted lines (8)

The one-coordinate hypothesis remains a substantive restriction on the direction field.
A uniform Hilbert transform estimate for Lipschitz directions Back to text
The scale restriction limits the turning of the field, but it does not make the operator a Fourier multiplier: the line used by the integral still varies with the output point.
A uniform Hilbert transform estimate for Lipschitz directions Back to text
… gives an affirmative answer with a strong $L^2$ bound.
A uniform Hilbert transform estimate for Lipschitz directions Back to text
Both coordinates may enter the field arbitrarily subject to the Lipschitz bound.
A uniform Hilbert transform estimate for Lipschitz directions Back to text
The supremum in the theorem is outside the norm, so its first assertion is a uniform operator bound for each inner truncation.
A uniform Hilbert transform estimate for Lipschitz directions Back to text
These conventions distinguish the short-scale assertion from an estimate at arbitrary outer lengths for varying fields.
A uniform Hilbert transform estimate for Lipschitz directions Back to text
Rotation and Fubini reduce its transform, for any finite outer length, to the one-dimensional Hilbert transform.
A uniform Hilbert transform estimate for Lipschitz directions Back to text
There are absolute constants $a_*\in(0,1/2)$ and $C_*<\infty$ such that every Lipschitz field $v:\R^2\to S^1$ with $\Lip(v)\le1$ satisfies $\sup_{0<\epsilon<a_*} \norm{H_{v,a_*}^{\epsilon}f}_{L^2(\R^2)} \le C_*\norm{f}_{L^2(\R^2)} \qquad(f\in\mathcal S(\R^2)).$
A uniform Hilbert transform estimate for Lipschitz directions Back to text