Local Smoothing Conjecture

When a wave propagates through space — for example sound or light — its shape can be described mathematically. There are computational operations, so-called integral operators, that transform such wave functions: they average, smooth, or analyze the function across many points. The Local Smoothing Conjecture makes a precise statement about how “smooth” the result of such an operation can be at most. Smooth here means: how much the values of the function are allowed to change from point to point. The conjecture has remained unproven since the early 1990s — and is considered one of the most difficult open problems in a subfield of mathematics called harmonic analysis, which studies how signals can be decomposed into their constituent parts.

Significance for Mathematics and Computer Science

The conjecture is not an isolated puzzle. It is linked to a number of other unsolved problems, including the so-called Kakeya Conjecture, which describes how much space is at minimum required to rotate a line segment in space in all directions. Anyone who proves the Local Smoothing Conjecture automatically solves an entire chain of related questions along with it. This makes it a strategic hub within mathematics.

In computer science and signal processing, related operators appear constantly in practice: when compressing audio files, sharpening images, or analyzing radar signals. The mathematical theory behind this determines which algorithms are possible at all — and how efficient they can be. A complete proof of the conjecture would yield sharper bounds for such procedures.

What the Conjecture Precisely Claims

Mathematicians measure the “smoothness” and “size” of functions using so-called norms — these are measures, similar to a ruler for functions. The Local Smoothing Conjecture claims that a certain wave operator, applied over a short time interval, makes the function smoother by a very specific amount — measured in these norms. Concretely, this concerns so-called Lp norms, that is, measures indicating how large the p-th power of the function’s values is on average.

The difficulty is that the conjecture can neither be easily proven nor easily disproven. Previous proofs cover only special cases, that is, certain dimensions or certain ranges of values for p. In 2022, the mathematician Hong Wang, together with colleagues, achieved a significant partial breakthrough for the two-dimensional case. However, the general proof is still outstanding.

A common misconception is to dismiss the conjecture as purely abstract. In fact, the bounds in question are directly relevant to the analysis of algorithms: they determine in what sense a procedure behaves stably or unstably when the input data is slightly noisy.

Local Smoothing in Current Debates

The term appears primarily in academic publications and conference proceedings on mathematics and theoretical computer science. When researchers write about the limits of Fourier analysis — that is, methods that decompose signals into oscillations — or about the complexity of image processing algorithms, they frequently cite the current state of the Local Smoothing Conjecture as a benchmark: how far along is the proof, and what would it enable?

In AI research, the direct connection is still indirect. Neural networks that process images or audio signals internally use operations that are structurally similar to the wave operators described here. A better mathematical theory of these operators could, in the long run, help explain why certain network architectures are more stable or efficient than others. The Local Smoothing Conjecture is thus an example of pure mathematics that is on its way toward application — but has not yet arrived.