Gabor systems with Hermite functions of order n n and oversampling greater than n + 1 n+1 which are not frames

Markus Faulhuber · 2025

Read Gabor systems with Hermite functions of order n n and oversampling greater than n + 1 n+1 which are not frames by Markus Faulhuber (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-09510.txt.

Introduction

Markus Faulhuber gives a math result that breaks a hopeful rule of thumb in signal work. Gabor systems build signals from shifted and tuned copies of one wave shape. When the base shape is a Hermite function, a known density rule was thought to be enough to make the system a frame, a stable basis for signals. His 2025 work shows it is not.

Faulhuber proves the point through the zeros of a helper transform, which decide whether such systems hold up as frames. He builds cases with sampling well above the rule's bar that still fail to be frames. The result warns engineers and mathematicians alike: for these wave shapes, density alone cannot promise a sound signal basis.

Catalog record

CREATORMarkus Faulhuber
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-09510

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Gabor systems with Hermite functions of order n n and oversampling greater than n + 1 n+1 which are not frames Thanks: This research was funded in whole or in part by the Austrian Science Fund (FWF) [ 10.55776/P33217 ]. The author thanks Karlheinz Gröchenig for feedback on the initial draft of the paper. The author gratefully acknowledges the comments from the anonymous referees. Markus Faulhuber Abstract . We show that a sufficient density condition for Gabor systems with Hermite functions over lattices is not sufficient in general. This follows from a result on how zeros of the Zak transform determine the frame property of integer over-sampled Gabor systems. 1991 Mathematics Subject Classification 42C15; 33C45 1. Introduction Gabor systems with Hermite functions have some quite satisfying, but just as many mysterious properties. A fundamental result due to Lyubarskii [ 22 ] and Seip and Wallstén [ 25 ] , [ 26 ] is that a Gabor system with a Gaussian window is a frame for L^{2}({\mathbb{R}}) if and only if the sampling rate exceeds 1. Indexing the Gaussian as the 0 -th Hermite function, we have the quite satisfying condition, due to Gröchenig and Lyubarskii [ 15 ] , [ 16 ] , that a Gabor system with the n -th Hermite function over a lattice \Lambda\subset{\mathbb{R}}^{2} is a frame once the density of the lattice is larger than n+1 , i.e., {\rm vol}({\mathbb{R}}^{2}/\Lambda)<1/(n+1) . A main difference to the results in [ 22 ] , [ 25 ] , [ 26 ] is that the results in [ 15 ] and [ 16 ] rely on the lattice structure. In this work, we answer the following question, which has been around for about 20 years, since the first announcement of their result by Gröchenig and Lyubarskii: Can the sufficiency condition be extended beyond the lattice setting? The answer is NO! This paper provides explicit examples of Gabor systems with the first and third Hermite functions and oversampling rates of 3 and 5, respectively, which are not frames. These are derived from the location of known zeros of their Zak transforms (the existence of further zeros is open). Additionally, we show the existence of a Gabor system with the second Hermite function and oversampling rate 5, and one with the third Hermite function and oversampling rate 7, which are not frames. Our approach is not entirely new: The Zak transform and its zeros have often been used to construct lattices which, together with a (class of) Hermite window(s), do not yield a frame, see, e.g., [ 15 , Prop. 3.5] , [ 19 ] , [ 21 ] . The sets we construct still exhibit periodicities and symmetries, but are not lattices. In [ 18 ] , results for Gabor systems with semi-regular lattices \Gamma\times{\mathbb{Z}}\subset{\mathbb{R}}^{2} have been derived. Our examples do not generally show this tensor structure, but we give non-frame examples for the third Hermite function with semi-regular lattices of density 2, 3, and 4. A related problem concerns the frame set structure for Gabor systems with the n -th Hermite functions and lattices of density 1<D(\Lambda)\leq n+1 [ 14 ] . So far, restrictions for odd Hermite functions and rectangular lattices are known [ 23 ] , which have been extended to general lattices [ 6 ] : The Gabor system with an odd Hermite function is not a frame if the lattice density satisfies D(\Lambda)=\frac{N+1}{N} , N a positive integer. For even Hermite windows, there were no further restrictions than the density condition D(\Lambda)>1 until Lemvig provided the first exceptions [ 21 ] . More examples of Gabor systems with Hermite functions and rectangular lattices of density 1<D(\Lambda)\leq n+1 , followed [ 19 ] . For h_{1} , it could still hold that the restriction D(\Lambda)>\frac{N+1}{N} , provided in [ 23 ] , is the only exception [ 14 ] . The results in [ 8 ] broke a pattern occurring for Hermite functions of order up to 3, challenging a question posed by Lemvig [ 21 ] : Is the lattice density condition of Gröchenig and Lyubarskii sharp ? This question remains open. If true, it may be answered by knowing the number and location of the zeros of the Zak transforms of generalized Hermite functions. The author thanks the first referee for providing the following remarks and consequences: Our answer may be surprising. At least from a complex analysis viewpoint hints were making it reasonable to believe that a lower Beurling density D^{-}(\Gamma)>n+1 could be sufficient for separated points sets \Gamma to form a Gabor frame with the n -th Hermite function (cf. [ 4 ] for the case of multiplicity sampling with Hermite functions). Denoting the upper Beurling density of \Gamma by D^{+}(\Gamma) , it can be shown that a Gabor system with Hermite window is a Riesz basis if D^{+}(\Gamma)<1/(n+1) . Using the Weierstrass sigma function, this result was established for lattices in [ 1 , Thm. 1] . However, as in [ 26 ] , the result relies on the growth of the sigma function and can be generalized using a non-uniform version. In the case of a lattice \Lambda\subset{\mathbb{R}}^{2} , we have D^{+}(\Lambda)=D^{-}(\Lambda)=D(\Lambda) , which is called the density of the lattice. Duality results (Ron-Shen [ 24 ] , Wexler-Raz [ 27 ] ) yield equivalent conditions between the Riesz basis property and the frame property: A Gabor system is a frame if and only if the dual system with lattice \Lambda^{\circ}=D(\Lambda)^{1/2}\Lambda is a Riesz basis. Thus, we not only show that the sufficient density result cannot be extended beyond the lattice setting, but our results also prohibit non-uniform versions of the Wexler-Raz or Ron-Shen duality principles. Otherwise, the density result could be deduced from the Riesz basis property. Outline of the paper. We clarify the notation in § 2 : In § 2.1 we introduce our normalization for Hermite functions, which is the same as in the book of Folland [ 11 ] . Time-frequency shifts and multi-window Gabor systems are introduced in § 2.2 , following the book of Gröchenig [ 13 ] . Also, the notion of a periodic structure (the union of finitely many shifted copies of one lattice) is discussed there. We then present the (parameter-free) Zak transform in § 2.3 , which maps functions from {\mathbb{R}} to quasi-periodic functions with period lattice {\mathbb{Z}}^{2} . This is again covered by the book of Gröchenig [ 13 ] . The reader familiar with the concepts and notation may directly jump to the results in § 3 . Finally, in § 4 we pose some open questions on the topic which should be understood as lines of research to advance the status of research. The author thanks the referees for the encouragement to include a discussion on open problems and Hermite windows of low order. 2. Notation In most parts of this article, we will follow the notation from the book of Gröchenig [ 13 ] . The Hilbert space of square-integrable functions on the real line is denoted by L^{2}({\mathbb{R}}) . For technical reasons, it will sometimes be necessary to restrict ourselves to a suitable dense subspace, in which case we will also write that a function

Read the complete text