---
title: Ivashev–Musatov Theorem in Harmonic Analysis
url: https://www.emergentmind.com/topics/ivashev-musatov-theorem
type: topic
---

# Ivashev–Musatov Theorem in Harmonic Analysis

Searching arXiv for recent and classical discussions of the Ivashev–Musatov theorem and related analogues.
The **Ivashev–Musatov theorem** is a theorem in harmonic analysis on the unit circle \(T\) that identifies a sharp borderline between Fourier-side smallness and spatial concentration. In the formulation stated in recent lecture notes and in later work on summable analogues, it asserts that if a positive majorant \(w(n)\) has divergent square sum and satisfies a dyadic regularity condition, then there exists a **probability measure** \(\mu\) supported on a set of Lebesgue measure zero on \(T\) such that
\[
|\widehat{\mu}(n)| \leq w(|n|), \qquad n=\pm1,\pm2,\dots.
\]
The theorem is presented as a sharp manifestation of the uncertainty principle: square-summability of Fourier coefficients would force \(L^2\)-regularity and rule out null support, but failure of \(\ell^2\)-summability—under mild regularity of the majorant—already allows measures supported on null sets [2604.24900]. Later papers treat the theorem as a prototype for threshold phenomena in Fourier analysis and develop topological, weighted, and summable analogues [2505.08980], [2505.15937].

## 1. Statement and basic interpretation

In the formulation given in the 2026 lecture notes, let \((w(n))_{n=0}^\infty\) be a sequence of positive numbers satisfying:

1. **Square divergence**
   \[
   \sum_n w(n)^2 = +\infty.
   \]

2. **Dyadic regularity**
   there exists a constant \(C>1\) such that for every integer \(n\ge 1\),
   \[
   C^{-1} w(n) \leq w(k) \leq C w(n), \qquad n\leq k\leq 2n.
   \]

Then there exists a **probability measure** \(\mu\) supported on a set of Lebesgue measure zero on \(T\), such that
\[
|\widehat{\mu}(n)| \leq w(|n|), \qquad n=\pm1,\pm2,\dots
\]
[2604.24900].

The 2025 paper on summable analogues gives an equivalent historical statement as “Theorem 1.1 (Ivashev–Musatov, 1958),” with the same square-divergence condition and dyadic regularity, plus an additional condition
\[
w_n \downarrow 0 \text{ or } n w_n \to 0,
\]
while also noting that Körner later proved that this extra condition can be removed [2505.08980]. The same paper emphasizes that condition \(\sum_n w_n^2 = \infty\) is absolutely necessary, and that the dyadic regularity condition cannot simply be dropped [2505.08980].

Conceptually, the theorem answers a support-versus-amplitude problem: how small the Fourier coefficients of a measure can be if the measure is supported on a very small set. The answer is that the \(\ell^2\) threshold is critical. This places the theorem among “borderline” uncertainty-principle results [2604.24900].

## 2. The \(\ell^2\) threshold and the uncertainty-principle viewpoint

The theorem is routinely interpreted through the Parseval obstruction. The lecture notes state that if a distribution \(S\) on \(T\) satisfies
\[
\sum_n |\widehat{S}(n)|^2 < \infty,
\]
then \(S\in L^2(T)\), hence cannot be supported on a set of Lebesgue measure zero [2604.24900]. The 2025 paper on summable analogues presents the same principle as the critical scale separating possible pathology from forced regularity, and explicitly interprets the classical theorem as a threshold phenomenon stemming from Parseval’s theorem [2505.08980].

The theorem is therefore understood as an essentially sharp converse to this obstruction. As soon as \(w\notin \ell^2\), meaning
\[
\sum_n w(n)^2 = +\infty,
\]
one can still realize the majorant \(w\) by Fourier coefficients of a probability measure supported on a null set, provided \(w\) has sufficient dyadic regularity [2604.24900].

This threshold interpretation is also central in later analogues. The 2025 weighted-\(\ell^2\) paper frames its main result as a threshold phenomenon in which weighted square-summability may or may not force regularity depending on the behavior of the weight [2505.15937]. The 2025 summable-analogue paper likewise presents the Ivashev–Musatov theorem as a prototype of a Fourier-side uncertainty principle: “small support” and “regularity/smoothness in frequency space” remain compatible right up to the critical \(\ell^2\) boundary [2505.08980].

A standard example arises from polynomial majorants. The lecture notes state that taking
\[
w(n)\asymp n^{-\alpha}
\]
with \(0<\alpha\le \tfrac12\) gives
\[
\sum_n w(n)^2 = \sum_n n^{-2\alpha} = \infty,
\]
and dyadic comparability is satisfied, so the theorem yields a probability measure supported on a null set with
\[
|\widehat{\mu}(n)| \lesssim |n|^{-\alpha}.
\]
The same notes remark that the theorem proves something stronger than the mere existence of positive measures with polynomial Fourier decay [2604.24900].

## 3. Proof architecture and constructive mechanism

The lecture notes present the theorem through a “Principal IM-Lemma.” Under the same hypotheses on \(w\), for every \(0<\varepsilon<1\) there exists \(\psi_\varepsilon\in C^\infty(T)\) such that:

1. \(\psi_\varepsilon \ge 0\) on \(T\),
2. \(\displaystyle \int_T \psi_\varepsilon\,dm = 1\),
3. \(\displaystyle m(\operatorname{supp}(\psi_\varepsilon\,dm))\le \varepsilon\),
4. \[
   |\widehat{\psi_\varepsilon}(n)| \le w(|n|), \qquad n\neq 0.
   \]

From this lemma, one sets
\[
d\mu_\varepsilon = \psi_\varepsilon\,dm
\]
and passes to a weak-\(*\) cluster point. The limit is a probability measure, its support has Lebesgue measure zero, and the Fourier majorant survives in the limit [2604.24900].

The notes describe two proof traditions. The “conventional approach,” traced to Ivashev–Musatov and early Körner work, starts with a smooth function \(f\), composes with a highly oscillatory unimodular map \(U\), and uses cancellation of low Fourier modes; this approach involves oscillatory integrals and stationary phase [2604.24900]. Körner later found a cleaner method that avoids these oscillatory-integral complications, and the notes state that their proof is largely based on Körner’s 1986 paper [2604.24900].

A key preparatory step converts dyadic regularity into polynomial comparability. The notes state that there exists \(M=M(w)>1\) such that
\[
w(m)\le \min\!\left(\frac{m}{n},\frac{n}{m}\right)^M w(n), \qquad m,n\ge 0.
\]
This bounds the oscillation of the majorant across scales and is crucial in summing Fourier estimates [2604.24900].

The principal local building block in the notes is a smooth function \(f_{I,N}\in C^\infty(T)\), supported on a closed arc \(I\subset T\), with controlled sign, controlled integral, and Fourier coefficients satisfying, for \(n\neq 0\),
\[
|\widehat{f_{I,N}}(n)| \le C(M)\,m(I)\,N^{-1/2}
\min\!\left(
\left(\frac{|n|m(I)}{N}\right)^M,
\left(\frac{|n|m(I)}{N}\right)^{-M}
\right).
\]
By patching such blocks over partitions of \(T\), the proof constructs nonnegative smooth functions \(\psi_{\varepsilon,\delta}\) with normalized integral, many zeros on every sufficiently large arc, and Fourier bounds
\[
|\widehat{\psi_{\varepsilon,\delta}}(n)| \le \varepsilon\, w(|n|), \qquad n\neq 0.
\]
An inductive multiplication scheme then shrinks the support while keeping the Fourier majorant below \(w\) [2604.24900].

A plausible implication is that the theorem is constructive in a strong approximation sense: it first builds absolutely continuous approximants with arbitrarily small support measure and then obtains the null-supported measure by compactness. That constructive feature is explicit in the notes’ proof skeleton [2604.24900].

## 4. Role of regularity, sharpness, and counterexamples

The dyadic regularity hypothesis is not decorative. The lecture notes explicitly state that the proof needs it in order to derive the polynomial comparability estimate
\[
w(m)\le \min\!\left(\frac{m}{n},\frac{n}{m}\right)^M w(n),
\]
and they then stress that the assumption is not something that can simply be ignored [2604.24900].

This point is sharpened by a counterexample due to Körner, also stated in the notes. There exists a decreasing sequence of positive numbers \((w(n))_n\) such that
\[
\sum_n w(n)^2=+\infty,
\]
but every nontrivial distribution \(S\) on \(T\) satisfying
\[
|\widehat{S}(n)| \le C\,w(|n|), \qquad n=0,\pm1,\pm2,\dots
\]
for some \(C>0\), must have **full support** in \(T\) [2604.24900]. Thus failure of \(\ell^2\)-summability alone is not enough.

The same theme appears in the 2025 summable-analogue paper, which states that the dyadic regularity \(w_n/w_{2n}\asymp 1\) “cannot simply be dropped” [2505.08980]. The lecture notes also discuss a one-sided variant: there exists a decreasing sequence \(\Phi(n)\) with
\[
\sum_n \Phi(n)^2=\infty,
\]
but any nontrivial measure \(\mu\in M(T)\) satisfying
\[
|\widehat{\mu}(n)|\le \Phi(n), \qquad n=1,2,3,\dots
\]
must have full support in \(T\) [2604.24900].

This combination of a positive theorem and sharp counterexamples explains why the Ivashev–Musatov theorem is regarded as a borderline result rather than a universal characterization. The correct interpretation is not that every non-\(\ell^2\) majorant is admissible, but that the \(\ell^2\) barrier is essentially sharp once a mild scale-regularity condition is imposed [2604.24900].

## 5. Later refinements and summable analogues

Recent work treats the Ivashev–Musatov theorem as the prototype of a broader family of threshold phenomena. The 2025 paper “Summable analogous to the Ivashev-Musatov Theorems” shifts from pointwise majorants
\[
|\widehat S(n)|\le C\,w_{|n|}
\]
to summability conditions in weighted \(\ell^1(w)\) and Orlicz spaces \(\ell^\Phi\) [2505.08980]. Its main measure-theoretic analogue states:

- if \(\Phi\) is continuous and
  \[
  \frac{\Phi(x)}{x^2}\downarrow0 \qquad (x\downarrow0),
  \]
  then there exists a probability measure \(\sigma\) supported on a compact set of Lebesgue measure zero in \(T\), such that
  \[
  \sum_{n=-\infty}^\infty \Phi\big(|\widehat \sigma(n)|\big)<\infty;
  \]

- if \((w_n)_n\) is positive with
  \[
  \inf_n w_n=0,
  \]
  then there exists a probability measure \(\sigma\) supported on a set of Lebesgue measure zero, such that
  \[
  \sum_n |\widehat\sigma(n)|\,w_{|n|}<\infty.
  \]
Both are presented as sharp summable analogues of the classical theorem [2505.08980].

The same paper also develops topological analogues in which “support of measure zero” is replaced by “support has empty interior.” Under suitable assumptions, there exists a positive essentially bounded function \(f\) on \(T\), whose support contains no interior, with Fourier coefficients summable in \(\ell^1(w)\) or \(\ell^\Phi\) [2505.08980]. More strongly, the subcollection of such pairs \((f,E)\) is generic in a complete metric space, so the pathology is Baire-generic rather than merely existential [2505.08980].

A Hilbertian version appears in “A generic threshold phenomena in weighted \(\ell^2\),” which gives a “summable Baire category version of Körner’s topological Ivashev–Musatov theorem” [2505.15937]. If \((\lambda_n)_n\) is positive and satisfies
\[
\sum_{n=0}^\infty \frac{1}{\lambda_n}=+\infty
\]
together with a dyadic comparability condition
\[
C^{-1}\lambda_n \le \lambda_k \le C\lambda_n \quad\text{whenever } n\le k\le 2n,
\]
then there exists a positive function \(f\in L^\infty(T,dm)\) such that \(\operatorname{supp}(f\,dm)\) has empty interior, but
\[
\sum_{n\in\mathbb Z} |\widehat f(n)|^2 \lambda_{|n|}<\infty.
\]
The same paper proves that such support pathology is generic in a complete metric space of admissible pairs \((f,E)\) [2505.15937].

These papers show that the Ivashev–Musatov phenomenon extends beyond pointwise majorization. A plausible implication is that the classical theorem has become a model example for studying the exact regularity threshold at which Fourier-side control still permits severe support or range pathology.

## 6. Relation to Schaeffer–Salem theory, dynamics, and common confusions

The theorem also appears in the literature on singular measures with near-critical Fourier decay. The 2012 paper on “Singular Schaeffer-Salem measures of dynamical system origin” places Ivashev–Musatov alongside Wiener and Wintner, Schaeffer, Salem, and Zygmund in the study of singular measures with Fourier coefficients close to the critical decay exponent \(-1/2\) [1205.6964].

That paper states that Schaeffer proved the existence of a singular \(\sigma\) with
\[
\widehat{\sigma}(n)=O\bigl(r(|n|)\cdot |n|^{-1/2}\bigr)
\]
for any given increasing sequence \(r(n)\to\infty\), and hence
\[
\widehat{\sigma}(n)=O(|n|^{-1/2+\varepsilon}) \quad \text{for any } \varepsilon>0.
\]
It then says that Ivashev–Musatov “got a further improvement of Schaeffer's result” by finding singular measures satisfying
\[
\widehat{\sigma}(n)=O(\rho(n)\cdot |n|^{-1/2+\varepsilon}),
\]
with \(\rho(n)\to 0\) but \(\rho(n)\gg |n|^{-\varepsilon}\) for any \(\varepsilon>0\) [1205.6964]. In that line of work, Ivashev–Musatov is associated with refined endpoint decay within the same critical regime.

The same 2012 paper proves that certain spectral measures of random dynamical systems satisfy
\[
\widehat{\sigma_f}(t)=O(|t|^{-1/2+\varepsilon})
\]
for any \(\varepsilon>0\), almost surely, for a dense set of functions \(f\) [1205.6964]. However, it explicitly says that it cannot determine whether these spectral measures are almost surely singular. Thus the paper gives a dynamical analogue of the decay side of Schaeffer–Salem theory, but not a full dynamical analogue of the classical singular-measure theorem [1205.6964].

A separate source of confusion is terminological. The 2014 paper “On the Closed Graph Theorem and the Open Mapping Theorem” does **not** explicitly mention, name, or cite an “Ivashev–Musatov theorem” [1411.5500]. Its focus is instead the relation between the Closed Graph Theorem and the Open Mapping Theorem, formalized in Theorem 14 via the properties \(\mathfrak{G}_C\) and \(\mathfrak{O}_C\) for topological groups [1411.5500]. Accordingly, the name “Ivashev–Musatov theorem” should not be attached to that theorem.

In the modern uncertainty-principle literature, the standard meaning is the harmonic-analytic theorem on the unit circle about Fourier majorants for measures supported on null sets [2604.24900], together with later topological and summable analogues [2505.08980], [2505.15937].

## 7. Historical placement and significance

The 2026 lecture notes describe the theorem as originating from the “profound work of Ivashev–Musatov” and later being “refined and improved by T. W. Körner” [2604.24900]. The 2025 summable-analogue paper likewise treats the 1958 theorem as classical and places it in a line of work continued by Körner, Kahane, Katznelson, and de Leeuw [2505.08980].

Its significance is twofold. First, it gives a sharp support-versus-Fourier-amplitude theorem on \(T\): null-supported measures can have remarkably small Fourier coefficients, right up to the \(\ell^2\) threshold [2604.24900]. Second, it has become a template for modern threshold results, in which one studies exactly when weighted \(\ell^1\), weighted \(\ell^2\), or Orlicz control forces continuity and when it still allows empty-interior support, unbounded range, or failure of Fourier–Stieltjes representability [2505.08980], [2505.15937].

In that sense, the Ivashev–Musatov theorem occupies a central position in the harmonic-analysis side of the uncertainty principle. It shows that “small support” and “small Fourier coefficients” are compatible much farther than naive intuition suggests, but only up to a sharp boundary dictated by \(\ell^2\) and tempered by scale regularity [2604.24900].

Source: https://www.emergentmind.com/topics/ivashev-musatov-theorem