Cantor measures with odd base do not admit Fourier frames

Abstract: We prove that the Cantor measure with base $b$ does not admit a Fourier frame whenever $b > 1$ is an odd integer. In particular, this answers a question of Strichartz on the existence of a Fourier frame for the middle third Cantor measure. A formalization of our main result in Lean 4 is also provided.

• The paper establishes that for any odd integer base b > 1, the Cantor measure μ_b lacks a Fourier frame in L²(μ_b).
• It employs self-similarity, trigonometric identities, and Beurling density arguments to rigorously demonstrate the incompatibility of frame bounds.
• The result decisively resolves a longstanding open problem, refining the dichotomy between even and odd bases in fractal harmonic analysis.

Nonexistence of Fourier Frames for Cantor Measures with Odd Base

Introduction

The paper "Cantor measures with odd base do not admit Fourier frames" (2607.08656) addresses a central open problem in the interplay between fractal measures and Fourier analysis: whether Cantor measures with an odd integer base $b > 1$ admit Fourier frames in $L^2(\mu_b)$. This problem, originating in classical studies on spectral and frame-spectral measures and made explicit by Strichartz, remained unresolved for several decades. The authors provide a precise and definitive answer, establishing that for any odd base $b > 1$, the associated self-similar Cantor measure $\mu_b$ admits no Fourier frame.

Background and Problem Formulation

Let $\mu$ be a finite Borel measure on $\mathbb{R}$. It is termed spectral if $L^2(\mu)$ admits an orthonormal basis of complex exponentials, i.e., there exists $\Lambda \subset \mathbb{R}$ such that $\{e^{2\pi i \lambda x} : \lambda \in \Lambda\}$ is orthonormal and complete in $L^2(\mu)$. A less restrictive requirement is that of a Fourier frame: there exist constants $L^2(\mu_b)$0 such that for all $L^2(\mu_b)$1

$L^2(\mu_b)$2

Spectrality implies the frame property; the converse is false. The frame property is significant for signal processing due to stability and redundancy.

Self-similar Cantor measures $L^2(\mu_b)$3 are supported on Cantor sets constructed via the base-$L^2(\mu_b)$4 iterated function system with digits $L^2(\mu_b)$5. It is a classical result (Jorgensen and Pedersen) that $L^2(\mu_b)$6 is spectral if and only if $L^2(\mu_b)$7 is even. For odd $L^2(\mu_b)$8, the construction of an orthonormal exponential basis is impossible, but the existence of Fourier frames remained an open problem due to the lack of general obstructions in this setting.

Main Results

The primary result is the following:

Theorem: If $L^2(\mu_b)$9 is an odd integer, then the Cantor measure $b > 1$0 does not admit a Fourier frame in $b > 1$1.

This settles Strichartz's question regarding the middle third Cantor set ($b > 1$2) and confirms that the spectral dichotomy between even and odd bases persists at the level of frames. Consequently, $b > 1$3 admits a Fourier frame if and only if $b > 1$4 is even, establishing a sharp boundary in the class of self-similar Cantor measures.

Techniques and Proof Outline

The analysis leverages several structural properties of self-similar measures, Fourier analysis, and trigonometric polynomial duality. The main novelty lies in demonstrating an intrinsic obstruction to the frame property for odd-base measures—even beyond orthogonality constraints.

Key technical steps include:

1. Structure of Cantor Measures: The authors rigorously develop the measure-theoretic and functional-analytic properties of $b > 1$5, exploiting affine contraction identities and the self-similarity equation to construct explicit orthonormal bases for certain hierarchical subspaces (Haar-type functions).
2. Fourier Transform Recurrence: The explicit recurrence $b > 1$6 for $b > 1$7 is employed iteratively to factor the Fourier transform along dyadic and base-$b > 1$8 scaling.
3. Nonexistence of Frame Spectra: The crux of the proof is to show that for any candidate system of exponentials, the upper and lower frame bounds become incompatible. This is achieved by constructing a sequence of Haar functions $b > 1$9 (using dual trigonometric polynomials) whose "energy" under the frame inequalities must be bounded below uniformly in $\mu_b$0, while an explicit computation shows the corresponding sum vanishes in the limit $\mu_b$1. The key is the exploitation of the delicate trigonometric identities and self-similarity at the heart of $\mu_b$2.
4. Beurling Density Arguments: Beurling-type density conditions for exponential frames are invoked in an adapted form for singular measures. The argument shows that necessary density estimates are violated for odd $\mu_b$3 due to algebraic constraints in the self-similar structure.
5. Formal Verification: The main theorem and all essential constructions are formalized in Lean 4, guaranteeing complete correctness and reproducibility of the complex combinatorial details underpinning the measure, trigonometric identities, and frame inequalities.

Numerical and Theoretical Strength

The paper's conclusion is sharp and unconditional: for all odd integers $\mu_b$4, the frame property for exponentials fails for $\mu_b$5. The proof rules out not just orthonormal bases or Riesz bases, but all frames of exponentials—a significantly stronger obstruction than previously available for fractal measures.

This result is in contrast to the case $\mu_b$6 and other even bases, where frames, orthonormal bases, and richer harmonic structures exist; for instance, explicit exponential Parseval frames were constructed for the Cantor-4 measure [Picioroaga & Weber, 2017].

Moreover, the work subsumes earlier partial results, such as the existence of high-density Riesz sequences without full frames for these measures [Dutkay, Emami & Lai, 2021], by establishing that the frame property itself is categorically obstructed.

Implications and Future Directions

This theorem decisively clarifies the landscape of Fourier analysis on self-similar singular measures, showing that the obstructions for frame spectrality for Cantor measures are not merely a technical artifact of orthogonality, but a manifestation of deep combinatorial and arithmetic incompatibilities between the dilation structure (parity of $\mu_b$7) and the Fourier system.

Theoretical implications include:

• The sharp dichotomy refines spectral theory for self-similar and fractal measures, pointing to parity as an invariant of fundamental harmonic importance.
• The negative answer to Strichartz's question rules out frame-based signal decompositions for important classes of singular measures, influencing theory and applications in analysis, mathematical physics, and fractal geometry.
• The proof methodology suggests new algebraic and trigonometric obstructions potentially applicable to other classes of iterated function system (IFS) measures, such as Bernoulli convolutions and more general non-integer base constructions.

Practical implications are more subtle:

• For applications in signal processing or harmonic analysis involving fractal or highly singular measures, the result delineates when frame-based expansions (which underlie stable representation and reconstruction) are a priori impossible and must be replaced by fundamentally different analytic tools.
• The strong nonexistence result informs computational harmonic analysts and numerical practitioners to avoid certain "natural" frame-based approaches for odd-base Cantor-type supports.

Future directions pointed out by the result include:

• Extension of these dichotomy obstructions to broader classes of self-affine or dynamically defined measures, perhaps in higher dimensions or with digit sets beyond $\mu_b$8.
• Investigation of whether other forms of frames (weighted, almost-Parseval, or non-exponential) can exist in these contexts.
• Further exploration of the connection between parity-based arithmetic structure and harmonic theoretical properties in singular measure spaces.

Conclusion

The paper provides a comprehensive, definitive resolution to a fundamental question on Fourier frames for Cantor measures, establishing that such frames do not exist when the underlying base is odd. The classification result unifies several threads in spectral theory for fractal measures, highlights the subtlety of harmonic structure in the singular setting, and sets the stage for further progress in the analysis and formalization of non-classical frame phenomena in fractal harmonic analysis (2607.08656).