Quasi-Homogeneous Spectral Measures

• Quasi-homogeneous spectral measures are tempered pure-point measures whose uniformly discrete support and Fourier spectrum enforce a finite lattice-coset organization.
• In one dimension, complex tempered measures exhibit a rigid structure by decomposing into finite unions of dual lattice cosets, thus precluding arbitrary aperiodicity.
• In ℝⁿ, positive measures can be expressed as modulated Dirac combs via Poisson summation, linking their structure to model-set quasicrystals and harmonic analysis.

Quasi-homogeneous spectral measures are tempered pure-point measures whose geometric support and Fourier spectrum are both uniformly discrete, and which therefore exhibit a rigid lattice-derived structure rather than arbitrary aperiodic behavior. In the classification developed by Lev and Olevskiĭ, the term refers to measures that are, in essence, finite combinations of Poisson combs on a lattice and its dual; in one dimension this holds for complex tempered measures, and in $\mathbb{R}^n$ it holds for positive tempered measures (Lev et al., 2013). The theory places model-set or quasicrystal measures within a precise harmonic-analytic framework, showing that uniformly discrete support on both sides of the Fourier transform imposes strong quasi-periodic constraints.

1. Definitions and ambient framework

A subset $\Lambda \subset \mathbb{R}$ is uniformly discrete if

$$d(\Lambda):=\inf\{|\lambda-\lambda'|:\lambda,\lambda'\in\Lambda,\ \lambda\neq \lambda'\}>0,$$

and, more generally, $A\subset\mathbb{R}^n$ is uniformly discrete if

$d(A):=\inf\{\|x-y\|:x,y\in A,\ x\neq y\}>0.$

This condition excludes accumulation at small scales and is the basic geometric regularity hypothesis in the classification theory (Lev et al., 2013).

A tempered positive Radon measure $\mu$ on $\mathbb{R}^n$ is called a pure-point or spectral measure if

$$\mu=\sum_{\lambda\in\Lambda} a_\lambda \delta_\lambda \qquad (a_\lambda>0,\ \Lambda\ \text{u.d.})$$

and its Fourier transform is also discrete:

$$\hat\mu(k)=\int_{\mathbb{R}^n} e^{-2\pi i\langle k,x\rangle}\,d\mu(x)=\sum_{\gamma\in\Gamma} b_\gamma \delta_\gamma,$$

with $\Gamma\subset\mathbb{R}^n$ uniformly discrete. In this setting, $\Lambda \subset \mathbb{R}$0 is the support of $\Lambda \subset \mathbb{R}$1 and $\Lambda \subset \mathbb{R}$2 is the support of $\Lambda \subset \mathbb{R}$3.

The prototypical nonperiodic pure-point sets are model sets arising from cut-and-project constructions. If $\Lambda \subset \mathbb{R}$4 is a full-rank lattice such that the projection $\Lambda \subset \mathbb{R}$5 is one-to-one and $\Lambda \subset \mathbb{R}$6 is dense in $\Lambda \subset \mathbb{R}$7, and if $\Lambda \subset \mathbb{R}$8 is bounded with boundary of Lebesgue measure zero, then

$\Lambda \subset \mathbb{R}$9

is uniformly discrete. Measures supported on such sets, or finite linear combinations of their translates and exponential modulations, provide the canonical quasi-crystalline examples.

Within this framework, “quasi-homogeneous” is used as a synonym for model-set-type spectral structure: the measure is not necessarily periodic, but it is forced into a finite lattice-coset organization.

2. One-dimensional classification

In one dimension, Lev–Olevskiĭ prove a strong rigidity theorem for complex tempered measures. If $$d(\Lambda):=\inf\{|\lambda-\lambda'|:\lambda,\lambda'\in\Lambda,\ \lambda\neq \lambda'\}>0,$$0 is a tempered measure on $$d(\Lambda):=\inf\{|\lambda-\lambda'|:\lambda,\lambda'\in\Lambda,\ \lambda\neq \lambda'\}>0,$$1 such that $$d(\Lambda):=\inf\{|\lambda-\lambda'|:\lambda,\lambda'\in\Lambda,\ \lambda\neq \lambda'\}>0,$$2 is uniformly discrete and

$$d(\Lambda):=\inf\{|\lambda-\lambda'|:\lambda,\lambda'\in\Lambda,\ \lambda\neq \lambda'\}>0,$$3

is a pure-point measure with uniformly discrete support $$d(\Lambda):=\inf\{|\lambda-\lambda'|:\lambda,\lambda'\in\Lambda,\ \lambda\neq \lambda'\}>0,$$4, then there exists a lattice $$d(\Lambda):=\inf\{|\lambda-\lambda'|:\lambda,\lambda'\in\Lambda,\ \lambda\neq \lambda'\}>0,$$5, necessarily of the form $$d(\Lambda):=\inf\{|\lambda-\lambda'|:\lambda,\lambda'\in\Lambda,\ \lambda\neq \lambda'\}>0,$$6, and finite sets $$d(\Lambda):=\inf\{|\lambda-\lambda'|:\lambda,\lambda'\in\Lambda,\ \lambda\neq \lambda'\}>0,$$7 such that

$$d(\Lambda):=\inf\{|\lambda-\lambda'|:\lambda,\lambda'\in\Lambda,\ \lambda\neq \lambda'\}>0,$$8

where $$d(\Lambda):=\inf\{|\lambda-\lambda'|:\lambda,\lambda'\in\Lambda,\ \lambda\neq \lambda'\}>0,$$9 is the dual lattice (Lev et al., 2013).

Equivalently, both the support and the spectrum are finite unions of cosets of mutually dual lattices. This excludes genuinely wild configurations in which a uniformly discrete support and a uniformly discrete Fourier support coexist without hidden periodic scaffolding. The theorem is especially notable because the hypotheses are purely Fourier-geometric: no a priori model-set construction is assumed, yet the conclusion recovers a lattice-coset decomposition.

A plausible implication is that one-dimensional pure-point Fourier duality is far more rigid than generic aperiodic order. Uniform discreteness on both sides already enforces a finite-rank arithmetic organization.

3. Positive measures in $A\subset\mathbb{R}^n$0

For positive tempered measures in higher dimensions, the same lattice-coset conclusion remains valid. If $A\subset\mathbb{R}^n$1 is a positive tempered measure on $A\subset\mathbb{R}^n$2 with $A\subset\mathbb{R}^n$3 and $A\subset\mathbb{R}^n$4 both uniformly discrete, then $A\subset\mathbb{R}^n$5 is contained in finitely many cosets of some lattice $A\subset\mathbb{R}^n$6, and $A\subset\mathbb{R}^n$7 is contained in finitely many cosets of the dual lattice $A\subset\mathbb{R}^n$8 (Lev et al., 2013).

The higher-dimensional positive case also admits an explicit structure theorem. If

$A\subset\mathbb{R}^n$9

then $d(A):=\inf\{\|x-y\|:x,y\in A,\ x\neq y\}>0.$0 can be written as

$d(A):=\inf\{\|x-y\|:x,y\in A,\ x\neq y\}>0.$1

where each $d(A):=\inf\{\|x-y\|:x,y\in A,\ x\neq y\}>0.$2 is a trigonometric polynomial on $d(A):=\inf\{\|x-y\|:x,y\in A,\ x\neq y\}>0.$3, and $d(A):=\inf\{\|x-y\|:x,y\in A,\ x\neq y\}>0.$4 has the same form on $d(A):=\inf\{\|x-y\|:x,y\in A,\ x\neq y\}>0.$5. Thus the measure decomposes into shifted lattice Dirac combs, modulated by exponentials and weighted by trigonometric polynomials.

This normal form clarifies the meaning of quasi-homogeneity in the present setting. The measure need not be a single lattice comb, and it may retain nontrivial phase and finite-coset structure, but its entire Fourier geometry is generated from a lattice template. The nonperiodic component is finite and controlled rather than diffuse.

4. Poisson-type duality and canonical examples

The pure-point analogue of Poisson’s summation formula appears as a duality of weights. Writing

$d(A):=\inf\{\|x-y\|:x,y\in A,\ x\neq y\}>0.$6

one has, for every $d(A):=\inf\{\|x-y\|:x,y\in A,\ x\neq y\}>0.$7,

$d(A):=\inf\{\|x-y\|:x,y\in A,\ x\neq y\}>0.$8

and for each $d(A):=\inf\{\|x-y\|:x,y\in A,\ x\neq y\}>0.$9,

$\mu$0

Each side converges absolutely because $\mu$1 and $\mu$2 are uniformly discrete and the weights grow at most polynomially. Equivalently, for every Schwartz test function $\mu$3,

$\mu$4

(Lev et al., 2013).

Two classes of examples organize the theory.

Example	Support and spectrum	Characteristic feature
Lattice comb	$\mu$5, $\mu$6	Exact Poisson summation
Model-set quasicrystal	$\mu$7 and $\mu$8 both u.d. but not lattices	Nonperiodic cut-and-project structure

For a full-rank lattice $\mu$9, the Dirac comb

$\mathbb{R}^n$0

has

$\mathbb{R}^n$1

This is the basic periodic case.

For model sets obtained by cut-and-project, the Poisson formula on the larger lattice $\mathbb{R}^n$2 shows that the Fourier transform of $\mathbb{R}^n$3 is again pure point, supported on the projection of $\mathbb{R}^n$4, and the weights are given by the Fourier transform of the characteristic function of the window $\mathbb{R}^n$5. These examples are genuinely nonperiodic while still satisfying uniform discreteness of both support and spectrum (Lev et al., 2013).

5. Proof architecture, Meyer theory, and classification consequences

The proof strategy combines Fourier analysis, density arguments, and additive-geometric structure. For each $\mathbb{R}^n$6, one studies the difference measure

$\mathbb{R}^n$7

Using Paley–Wiener/Beurling–Landau density arguments, or Ingham’s gap theorems in $\mathbb{R}^n$8, one shows that $\mathbb{R}^n$9, $$\mu=\sum_{\lambda\in\Lambda} a_\lambda \delta_\lambda \qquad (a_\lambda>0,\ \Lambda\ \text{u.d.})$$0, $$\mu=\sum_{\lambda\in\Lambda} a_\lambda \delta_\lambda \qquad (a_\lambda>0,\ \Lambda\ \text{u.d.})$$1, and $$\mu=\sum_{\lambda\in\Lambda} a_\lambda \delta_\lambda \qquad (a_\lambda>0,\ \Lambda\ \text{u.d.})$$2 must all have positive lower and upper uniform densities bounded away from zero and infinity (Lev et al., 2013).

The next step uses Meyer set theory. The uniform discreteness of $$\mu=\sum_{\lambda\in\Lambda} a_\lambda \delta_\lambda \qquad (a_\lambda>0,\ \Lambda\ \text{u.d.})$$3 together with the Delone property of $$\mu=\sum_{\lambda\in\Lambda} a_\lambda \delta_\lambda \qquad (a_\lambda>0,\ \Lambda\ \text{u.d.})$$4 forces $$\mu=\sum_{\lambda\in\Lambda} a_\lambda \delta_\lambda \qquad (a_\lambda>0,\ \Lambda\ \text{u.d.})$$5 to be a Meyer set. A classical theorem of Meyer–Lagarias then implies that $$\mu=\sum_{\lambda\in\Lambda} a_\lambda \delta_\lambda \qquad (a_\lambda>0,\ \Lambda\ \text{u.d.})$$6 is contained in finitely many translates of some lattice $$\mu=\sum_{\lambda\in\Lambda} a_\lambda \delta_\lambda \qquad (a_\lambda>0,\ \Lambda\ \text{u.d.})$$7. Once $$\mu=\sum_{\lambda\in\Lambda} a_\lambda \delta_\lambda \qquad (a_\lambda>0,\ \Lambda\ \text{u.d.})$$8 is established, Poisson summation on $$\mu=\sum_{\lambda\in\Lambda} a_\lambda \delta_\lambda \qquad (a_\lambda>0,\ \Lambda\ \text{u.d.})$$9 combined with a small Vandermonde linear-algebra argument yields the finite coset decomposition and the explicit shifted/modulated lattice-comb representation.

These theorems confirm that any tempered or positive pure-point spectral measure with both support and spectrum uniformly discrete is necessarily “quasi-homogeneous,” that is, a finite combination of Poisson combs on a lattice and its dual. They also resolve the conjecture of Lagarias in full generality for $$\hat\mu(k)=\int_{\mathbb{R}^n} e^{-2\pi i\langle k,x\rangle}\,d\mu(x)=\sum_{\gamma\in\Gamma} b_\gamma \delta_\gamma,$$0 and for positive measures in $$\hat\mu(k)=\int_{\mathbb{R}^n} e^{-2\pi i\langle k,x\rangle}\,d\mu(x)=\sum_{\gamma\in\Gamma} b_\gamma \delta_\gamma,$$1 (Lev et al., 2013).

Open problems listed in the same development are threefold: removing the positivity assumption in higher dimensions, weakening uniform discreteness to mere discreteness, and extending the theory to other locally compact abelian groups. These problems delineate the boundary between the current classification and more general Fourier-quasicrystal phenomena.

6. Related usages and terminological distinctions

The phrase “spectral measure” is not uniform across the literature. In the Lev–Olevskiĭ setting, it denotes a pure-point measure whose Fourier transform is also a pure-point measure. In a later self-similar setting, a compactly supported probability measure $$\hat\mu(k)=\int_{\mathbb{R}^n} e^{-2\pi i\langle k,x\rangle}\,d\mu(x)=\sum_{\gamma\in\Gamma} b_\gamma \delta_\gamma,$$2 on $$\hat\mu(k)=\int_{\mathbb{R}^n} e^{-2\pi i\langle k,x\rangle}\,d\mu(x)=\sum_{\gamma\in\Gamma} b_\gamma \delta_\gamma,$$3 is called spectral if $$\hat\mu(k)=\int_{\mathbb{R}^n} e^{-2\pi i\langle k,x\rangle}\,d\mu(x)=\sum_{\gamma\in\Gamma} b_\gamma \delta_\gamma,$$4 admits an orthogonal basis of exponentials $$\hat\mu(k)=\int_{\mathbb{R}^n} e^{-2\pi i\langle k,x\rangle}\,d\mu(x)=\sum_{\gamma\in\Gamma} b_\gamma \delta_\gamma,$$5, with completeness tested by

$$\hat\mu(k)=\int_{\mathbb{R}^n} e^{-2\pi i\langle k,x\rangle}\,d\mu(x)=\sum_{\gamma\in\Gamma} b_\gamma \delta_\gamma,$$6

(Liu et al., 31 Oct 2025). These are related Fourier-analytic notions, but they are not identical.

A further extension of the adjective “quasi-homogeneous” appears in the study of self-similar measures with $$\hat\mu(k)=\int_{\mathbb{R}^n} e^{-2\pi i\langle k,x\rangle}\,d\mu(x)=\sum_{\gamma\in\Gamma} b_\gamma \delta_\gamma,$$7-periodic alternating contraction ratios. In that context, the alternation of contraction sign produces a “quasi-homogeneous” iterated function system, and the associated measure can be conjugated to a uniform-ratio system. The resulting characterization states that the self-similar measure $$\hat\mu(k)=\int_{\mathbb{R}^n} e^{-2\pi i\langle k,x\rangle}\,d\mu(x)=\sum_{\gamma\in\Gamma} b_\gamma \delta_\gamma,$$8 is spectral if and only if $$\hat\mu(k)=\int_{\mathbb{R}^n} e^{-2\pi i\langle k,x\rangle}\,d\mu(x)=\sum_{\gamma\in\Gamma} b_\gamma \delta_\gamma,$$9 and $\Gamma\subset\mathbb{R}^n$0 (Liu et al., 31 Oct 2025). This is a distinct usage of quasi-homogeneity: it concerns arithmetic and Fourier structure of an IFS rather than the lattice-coset classification of uniformly discrete support and spectrum.

Another nearby but non-equivalent notion comes from multifractal analysis. A diffuse Borel probability measure on $\Gamma\subset\mathbb{R}^n$1 is homogeneously multifractal if its restriction to every non-degenerate subinterval has the same multifractal spectrum as the whole measure, and an $\Gamma\subset\mathbb{R}^n$2-quasi-homogeneous variant relaxes this by requiring

$\Gamma\subset\mathbb{R}^n$3

for every sufficiently large subinterval $\Gamma\subset\mathbb{R}^n$4 (Buczolich et al., 2013). This notion concerns local dimension spectra rather than Fourier support, and it should not be conflated with quasi-homogeneous spectral measures in the Lev–Olevskiĭ sense.

Taken together, these later usages suggest a broad terminological family centered on controlled nonuniformity. In the classification of pure-point Fourier measures, however, the precise content of quasi-homogeneity is sharply arithmetic: finite unions of lattice cosets, explicit Poisson-type duality, and a structure generated by lattice combs and their duals.