🔢 mathematics

Fourier series for Henstock-Kurzweil integrable functions

This paper establishes the convergence and localization of Fourier series for functions that are Henstock-Kurzweil integrable but not Lebesgue integrable due to singular behavior at the origin, provided they satisfy specific Dini-type or bounded-variation conditions.

Original authors: Manuel Bello-Hernandez, Pablo Saenz-Lopez-Guerra

Published 2026-08-06

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Manuel Bello-Hernandez, Pablo Saenz-Lopez-Guerra

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Great Wave Hunt: Chasing Signals in the Chaos

Imagine you are trying to listen to a specific song on the radio, but the signal is buried under a storm of static. In the world of mathematics, this "song" is a function—a rule that describes how things change—and the "static" is the noise or irregularity that makes it hard to understand. For over a century, mathematicians have used a powerful tool called Fourier series to break these complex functions down into simple, smooth waves (like sine and cosine waves). Think of it like taking a chaotic, jagged mountain range and describing it as a stack of perfectly smooth, rolling hills. If you can stack enough of these smooth hills, you can rebuild the mountain exactly.

However, there's a catch. For a long time, this method only worked if the mountain wasn't too wild. If the function had a "singularity"—a point where it spikes infinitely high or wiggles so violently that it breaks the rules of standard math (called Lebesgue integration)—the Fourier series would often fail, giving up and producing nonsense. It was like trying to listen to a song through a wall of screaming static; the math said, "I can't hear anything, so I'll just stop." But nature doesn't always follow the rules of standard math. Some functions are wild, oscillating wildly near a point, yet they still have a hidden order. The question is: Can we tune our radio to hear the song even when the static is screaming?

The Paper's Discovery: Taming the Wild Wiggles

This paper, written by Manuel Bello-Hernández and Pablo Saenz-Lopez-Guerra, tackles exactly that problem. They investigate a specific type of "wild" function that is Henstock-Kurzweil integrable. In plain English, this is a special kind of math that allows us to handle functions that are too messy for standard rules but still have a "cancellation" effect. Imagine a crowd of people running in a circle: if they all run at the same speed, they create a chaotic mess. But if they run in a pattern where half run clockwise and half run counter-clockwise at just the right speeds, the chaos cancels out, and the net movement is zero. These wild functions have that same "cancellation" property near their spikes, which makes them integrable in this advanced sense, even though they are technically "broken" by standard rules.

The authors focus on a specific model of these wild functions: a curve that spikes near zero (like $x^{-a}$ ) while vibrating incredibly fast (like $\cos(x^{-b})$ ). They ask: If we try to rebuild this wild curve using Fourier series (stacking smooth waves), does it actually work?

The Main Finding:
The paper proves that yes, the Fourier series does converge to the correct value, but only under a very specific set of conditions. The key is the relationship between how high the spike gets (controlled by the number $a$ ) and how fast it wiggles (controlled by the number $b$ ). The authors show that if the wiggles are fast enough to cancel out the spike, but not too fast, the Fourier series will successfully reconstruct the function at any point away from the spike. Specifically, they prove that if the condition $a < 1 + b/2$ is met, the series converges to the average of the left and right values of the function. If the function is continuous there, it converges to the function's actual value.

What They Rule Out:
Crucially, the paper shows that this "magic" has a hard limit. If the spike gets too high relative to the wiggles—specifically, if $a \ge 1 + b/2$ —the Fourier series breaks down. The authors construct a specific example where $a = 1.5$ and $b = 1$ (which hits the limit exactly). In this case, the Fourier coefficients (the building blocks of the waves) stop getting smaller and start staying large. As a result, the series fails to converge at certain points, no matter how many waves you add. It's like trying to balance a tower of blocks where the bottom blocks are too heavy; no matter how carefully you stack the top, the whole thing collapses.

How Sure Are They?
The authors don't just guess or simulate; they provide rigorous mathematical proofs. They use a toolkit of advanced estimates (like the van der Corput lemma, which is a way of measuring how much a wiggly wave cancels itself out) to show exactly why the series works in the "safe" zone and why it fails in the "danger" zone. They also prove a "localization principle," which means that what happens far away from the wild spike doesn't mess up the convergence near the spike, provided the function behaves nicely nearby. However, they also show that this localization principle can fail if the function is too wild (outside the standard $L^1$ class), meaning that in the most extreme cases, a problem in one part of the function can ruin the reconstruction everywhere else.

In short, the paper draws a precise map of where the Fourier series works for these chaotic, oscillating functions. It tells us that as long as the "wiggles" are strong enough to tame the "spikes" but not so strong that they create a new kind of chaos, we can successfully listen to the song through the static. But cross the line, and the music stops.

Technical Summary: Fourier Series for Henstock–Kurzweil Integrable Functions

Problem Statement
The convergence of Fourier series for functions that are Henstock–Kurzweil (HK) integrable but not Lebesgue integrable remains a significant challenge in harmonic analysis. While classical $L^1$ theory guarantees almost-everywhere convergence for $L^p$ ( $p>1$ ) functions (Carleson-Hunt theorem), it fails for general $L^1$ functions (Kolmogorov's counterexample). The HK integral extends the Lebesgue integral to include highly oscillatory improper integrals and derivatives that are not absolutely integrable. However, Fourier analysis in this setting is complicated by the fact that Fourier coefficients of HK-integrable functions need only be $o(n)$ , rather than $o(1)$ , and standard localization principles may fail. This paper investigates the pointwise and uniform convergence of Fourier series for a specific class of functions exhibiting singular oscillatory behavior at the origin, designed to lie beyond the Lebesgue $L^1$ framework.

Methodology
The authors analyze functions of the form $f(x) = g(x) \cos(x^{-b} - \theta)$ on $(0, 1)$ , where the singularity at the origin is characterized by a power-law decay $x^{-a}$ and rapid oscillation $x^{-b}$ . The core methodology involves:

Decomposition of Partial Sums: The symmetric partial sums $S_n(f, x_0)$ are decomposed into three regions: a singular neighborhood near the origin, a local neighborhood around the convergence point $x_0$ , and a regular intermediate region.
Oscillatory Integral Estimates: The authors employ refined versions of the van der Corput lemma and Abel–Stieltjes integration by parts. Specifically, they utilize second-order van der Corput estimates with amplitudes of bounded variation (BV) to control the singular contributions.
Stationary Phase Analysis: A critical component is the asymptotic evaluation of integrals where the phase function possesses a stationary point that approaches the singular endpoint as the frequency tends to infinity. The authors derive a "moving-scale" stationary phase formula (Lemma 9) for BV amplitudes to determine the decay rate of Fourier coefficients.
Uniformity Arguments: To establish uniform convergence on compact subsets, the authors prove that the constants in the van der Corput and Abel–Stieltjes estimates remain uniform with respect to the evaluation point $x_0$ within a compact set $K$ away from the singularity.

Key Results

Pointwise Convergence (Theorem 1): For functions $f(x) = g(x)\cos(x^{-b}-\theta)$ satisfying specific conditions (bounded variation of $x^a g(x)$ near 0, local BV near $x_0$ , and the parameter constraint $a < 1 + b/2$ ), the Fourier series converges pointwise to the midpoint value $\frac{f(x_0+) + f(x_0-)}{2}$ . If $g$ is continuous near $x_0$ , the convergence is uniform on compact subsets.
Dini-Type Convergence (Theorem 18): An analogous result is proven where the local bounded variation assumption at $x_0$ is replaced by a Dini-type condition, provided the integral of the difference quotient converges.
Localization Principle (Theorem 19): The paper establishes a conditional localization principle for this specific class of HK-integrable functions: if two such functions (satisfying the specific oscillatory singularity hypotheses and the $a < 1 + b/2$ constraint) agree on an open interval, their Fourier partial sums converge uniformly to each other on any compact subset of that interval.
Sharpness of Conditions and Failure of General Localization (Section 5):
- The condition $a < 1 + b/2$ is proven to be sharp. At the endpoint $a = 1 + b/2$ , the stationary phase contribution to the Fourier coefficients does not decay to zero (it remains of order $O(1)$ ), leading to divergence of the Fourier series at certain points.
- For the supercritical case $a > 1 + b/2$ , the Fourier coefficients grow unbounded, and the series diverges almost everywhere.
- Crucially, the paper constructs an explicit example of a general HK-integrable function where the localization principle fails, demonstrating that the principle is not valid for the entire class of HK-integrable functions, even when Fourier coefficients are well-defined. The localization result in Theorem 19 is restricted to the specific model functions with controlled oscillatory singularities.

Significance and Claims
The authors position this work as a modest but concrete advancement beyond the general Lebesgue framework. Rather than claiming a general Carleson-type theorem for all HK-integrable functions, the paper isolates a specific family of oscillatory singularities where ordinary Fourier partial sums can be controlled.

The primary significance lies in identifying the precise threshold ( $a < 1 + b/2$ ) where the oscillation near the singularity is sufficient to cancel the singularity's effect on Fourier convergence, despite the lack of absolute integrability. The results clarify the structural limitations of nonabsolute integration in orthogonal expansions, showing that while HK integration recovers coefficients for convergent series, pointwise convergence of the series itself requires stricter regularity conditions than mere HK integrability. Furthermore, the paper highlights that localization principles do not automatically extend to the full HK space, requiring specific structural constraints on the singularities. The paper concludes that these convergence theorems extend to functions with finitely many such admissible singularities, provided the local hypotheses hold at the points of interest.

The Great Wave Hunt: Chasing Signals in the Chaos

The Paper's Discovery: Taming the Wild Wiggles

More like this