Homogeneous Spectral Measures

• Homogeneous spectral measures are defined by uniform spectral properties, exhibiting spatial invariance or constant multiplicity across different mathematical frameworks.
• They underpin multifractal analysis, Cantor set constructions, and p-adic tiling, providing explicit spectral bases and tunable Hausdorff dimensions.
• In operator and ergodic theories, homogeneity informs scaling laws and constant multiplicity, leading to rigorous spectral classifications and dynamical insights.

Homogeneous spectral measures occur in several mathematically distinct settings, unified less by a single definition than by a recurring principle: some spectral object is spatially uniform, scale-invariant, or multiplicity-constant. In harmonic analysis, a spectral measure is a Borel probability measure $\mu$ for which there exists a countable set $\Lambda$ such that $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$ or $\{e^{2\pi i\lambda\cdot x}:\lambda\in\Lambda\}$ is an orthonormal basis of $L^2(\mu)$; in multifractal analysis, a measure is homogeneously multifractal when every non-degenerate subinterval has the same multifractal spectrum as the whole; in ergodic theory, a transformation has homogeneous spectrum when its spectral multiplicity function is almost everywhere constant; and in spectral theory of operators, the spectrum itself may be homogeneous in Carleson’s sense (Dai et al., 2014, Buczolich et al., 2013, Ryzhikov, 2012, Fillman, 2014). A further operator-theoretic usage concerns joint spectral measures of commuting Hermitian tuples on fractal sets, where Hausdorff measure governs a quasicentral-modulus formula (Voiculescu, 2020).

1. Terminological scope and principal meanings

A recurring source of ambiguity is that both “spectral” and “homogeneous” are used with different technical meanings across subfields. The following table summarizes the principal usages represented in the literature.

Notion	Core definition	Representative source
Homogeneously multifractal measure	Every non-empty subinterval has the same multifractal spectrum as $[0,1]$	(Buczolich et al., 2013)
Spectral measure	$L^2(\mu)$ has an orthonormal basis of exponentials	(Dai et al., 2014)
$p$-homogeneous spectral set in $\mathbb Q_p$	Tree of balls branches with either $1$ or $\Lambda$0 descendants; equivalent to spectrality and tiling for compact open sets	(Fan et al., 2015)
Homogeneous spectrum of multiplicity $\Lambda$1	Spectral multiplicity function is a.e. constant $\Lambda$2	(Ryzhikov, 2012)
$\Lambda$3-homogeneous Cantor spectrum	Carleson lower-density condition on a closed spectral set	(Fillman, 2014)
Fractal spectral-measure formula	$\Lambda$4 for commuting Hermitian tuples with spectrum in $\Lambda$5	(Voiculescu, 2020)

These meanings are not interchangeable. A measure can be spectral without being homogeneously multifractal, and a set can be homogeneous in the Carleson sense without being a spectral set. Likewise, “spectral measure” may mean an exponential-basis measure in harmonic analysis, the maximal spectral type of a unitary operator in ergodic theory, or the joint spectral measure of a commuting operator tuple.

2. Homogeneous multifractal measures

For a Borel probability measure $\Lambda$6 on $\Lambda$7, the local dimension at $\Lambda$8 is

$\Lambda$9

and the multifractal spectrum is

$\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$0

For measures on $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$1, the standard estimate used in the literature is

$\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$2

(Buczolich et al., 2013).

A measure $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$3 supported on $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$4 is called homogeneously multifractal if every non-empty subinterval $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$5 has the same multifractal spectrum as the whole interval: $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$6 for every $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$7 (Buczolich et al., 2013). In this sense, homogeneity is a statement about the spatial invariance of the full multifractal decomposition.

A central rigidity theorem states that for any non-atomic HM measure $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$8 supported on $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$9,

$\{e^{2\pi i\lambda\cdot x}:\lambda\in\Lambda\}$0

must be an interval of the form

$\{e^{2\pi i\lambda\cdot x}:\lambda\in\Lambda\}$1

Equivalently, the support of the spectrum inside $\{e^{2\pi i\lambda\cdot x}:\lambda\in\Lambda\}$2 is connected; there are no gaps in that range (Buczolich et al., 2013). The result is sharp: there exists an HM measure with

$\{e^{2\pi i\lambda\cdot x}:\lambda\in\Lambda\}$3

so isolated exponents above $\{e^{2\pi i\lambda\cdot x}:\lambda\in\Lambda\}$4 can occur (Buczolich et al., 2013).

The constructive theory is organized by an admissible class $\{e^{2\pi i\lambda\cdot x}:\lambda\in\Lambda\}$5 of functions $\{e^{2\pi i\lambda\cdot x}:\lambda\in\Lambda\}$6 that can be written as a supremum of countably many step functions supported on closed intervals $\{e^{2\pi i\lambda\cdot x}:\lambda\in\Lambda\}$7, with

$\{e^{2\pi i\lambda\cdot x}:\lambda\in\Lambda\}$8

For every $\{e^{2\pi i\lambda\cdot x}:\lambda\in\Lambda\}$9, there exists an HM Borel probability measure $L^2(\mu)$0 supported on $L^2(\mu)$1 such that

$L^2(\mu)$2

$L^2(\mu)$3

and the set of points where $L^2(\mu)$4 has Lebesgue measure $L^2(\mu)$5 (Buczolich et al., 2013). This realizes a broad class of prescribed spectra, subject to the interval-support constraint forced by homogeneous multifractality.

The same paper also constructs HM strictly increasing continuous functions with prescribed spectrum and, by wavelet methods, HM non-monotone functions whose spectra are affine images of HM measure spectra. This indicates that homogeneous multifractality is stable under several distinct constructive frameworks, including Cantor-type insertions and wavelet series (Buczolich et al., 2013).

3. Spectral measures on homogeneous Cantor sets and affine self-similar systems

In harmonic analysis, a compactly supported probability measure $L^2(\mu)$6 is spectral if there exists a countable $L^2(\mu)$7 such that

$L^2(\mu)$8

is an orthonormal basis for $L^2(\mu)$9. A standard criterion is the Jorgensen–Pedersen identity: if $[0,1]$0, then $[0,1]$1 is a spectrum for $[0,1]$2 iff

$[0,1]$3

(Dai et al., 2014).

A major class of examples is given by Riesz product measures on homogeneous Cantor sets. With sequences $[0,1]$4, $[0,1]$5 satisfying

$[0,1]$6

the homogeneous Cantor set is

$[0,1]$7

The associated Riesz product measure $[0,1]$8 is defined by

$[0,1]$9

(Dai et al., 2014).

Under the arithmetic assumptions

$L^2(\mu)$0

the explicit set

$L^2(\mu)$1

is a spectrum of $L^2(\mu)$2 (Dai et al., 2014). This gives concrete spectral measures supported on homogeneous Cantor sets that are generally not self-similar unless $L^2(\mu)$3 and $L^2(\mu)$4 are constant.

The Hausdorff dimension is tunable. If

$L^2(\mu)$5

then

$L^2(\mu)$6

Hence for every $L^2(\mu)$7 there exists a spectral measure $L^2(\mu)$8 with $L^2(\mu)$9, including non-atomic zero-dimensional spectral measures and one-dimensional singular spectral measures (Dai et al., 2014).

A parallel operator-algebraic framework treats self-similar affine measures generated by

$p$0

where $p$1 is expansive and $p$2 is finite with $p$3. If there is a second digit set $p$4 with $p$5 such that

$p$6

is unitary, then $p$7 is a Hadamard pair, and the operators

$p$8

satisfy the Cuntz relations (Dutkay et al., 2010). The dual IFS, the transfer operator

$p$9

and the geometry of $\mathbb Q_p$0-extreme $\mathbb Q_p$1-cycles then organize reducing subspaces of $\mathbb Q_p$2. In favorable cases, cycle-generated exponentials form a spectrum; in higher dimensions, more general invariant sets may be needed (Dutkay et al., 2010). This framework shows that homogeneous affine scaling by a single matrix $\mathbb Q_p$3 can support highly structured spectral measures.

4. $\mathbb Q_p$4-homogeneity, tiling, and spectrality in $\mathbb Q_p$5

For a compact open set $\mathbb Q_p$6 with $\mathbb Q_p$7, the normalized Haar measure is

$\mathbb Q_p$8

The set $\mathbb Q_p$9 is spectral if there exists $1$0 such that

$1$1

is an orthonormal basis of $1$2 (Fan et al., 2015).

The central classification theorem states that for compact open sets $1$3, the following are equivalent:

1. $1$4 is a spectral set.
2. $1$5 is $1$6-homogeneous.
3. $1$7 tiles $1$8 by translation.

Here $1$9-homogeneity is encoded by the tree of balls associated with $\Lambda$00: each vertex has either $\Lambda$01 or $\Lambda$02 descendants, and the number of descendants depends only on the ball’s size (Fan et al., 2015). In this setting homogeneity is completely equivalent to both Fourier spectrality and translational tiling.

If $\Lambda$03 is $\Lambda$04-homogeneous with admissible order set $\Lambda$05, then, up to isometry of $\Lambda$06, the spectrum is uniquely determined by

$\Lambda$07

while the tiling complement is uniquely determined by

$\Lambda$08

(Fan et al., 2015). The same equivalence appears in finite cyclic groups: for $\Lambda$09, spectrality, tiling, and $\Lambda$10-homogeneity are equivalent, and one criterion is

$\Lambda$11

for suitable integers $\Lambda$12 (Fan et al., 2015).

The paper also constructs singular spectral measures as weak limits of normalized Haar measures on nested $\Lambda$13-homogeneous compact open sets. If $\Lambda$14 are disjoint infinite subsets with $\Lambda$15, the resulting measure $\Lambda$16 is spectral with spectrum

$\Lambda$17

(Fan et al., 2015). In periodic cases, $\Lambda$18 becomes self-similar. Thus the $\Lambda$19-adic theory yields an exact equivalence between a combinatorial homogeneity condition and spectrality, together with explicit singular spectral measures.

5. Operator-theoretic spectral measures on fractals and homogeneous multiplicity in dynamics

A different operator-theoretic use of spectral measures arises for commuting Hermitian tuples $\Lambda$20. The quasicentral modulus relative to a normed ideal $\Lambda$21 is

$\Lambda$22

where $\Lambda$23 denotes finite-rank positive contractions and $\Lambda$24. For the $\Lambda$25-Lorentz ideal $\Lambda$26, one writes $\Lambda$27 (Voiculescu, 2020).

The principal structural result is the ampliation homogeneity theorem: $\Lambda$28 This scaling law is the key input in the fractal spectral analysis of the paper (Voiculescu, 2020).

The geometric setting is a Cantor-like totally disconnected self-similar set $\Lambda$29 generated by

$\Lambda$30

with $\Lambda$31, self-similarity

$\Lambda$32

and disjointness

$\Lambda$33

Its Hausdorff dimension is

$\Lambda$34

and the $\Lambda$35-dimensional Hausdorff measure $\Lambda$36 of $\Lambda$37 is finite and nonzero (Voiculescu, 2020).

For a commuting $\Lambda$38-tuple of Hermitian operators with $\Lambda$39, the joint spectral measure is denoted $\Lambda$40, and the multiplicity function by $\Lambda$41. If $\Lambda$42 is singular with respect to $\Lambda$43, then

$\Lambda$44

If $\Lambda$45 denotes the canonical tuple of multiplication by coordinate functions on $\Lambda$46, then for $\Lambda$47,

$\Lambda$48

The main formula is

$\Lambda$49

valid for commuting Hermitian tuples with $\Lambda$50 and $\Lambda$51 (Voiculescu, 2020). Consequently,

$\Lambda$52

In this setting the quasicentral modulus detects precisely the $\Lambda$53-absolutely continuous part of the spectral measure.

In ergodic theory, the phrase homogeneous spectrum refers instead to multiplicity. If $\Lambda$54 is an automorphism of a Lebesgue probability space and $\Lambda$55 is the maximal spectral type of $\Lambda$56 on the zero-mean subspace $\Lambda$57, then homogeneous spectrum of multiplicity $\Lambda$58 means that the spectral multiplicity function is almost everywhere constant and equal to $\Lambda$59 (Ryzhikov, 2012). A central theorem states that if $\Lambda$60 is ergodic and

$\Lambda$61

weakly for some sequence $\Lambda$62 and some $\Lambda$63, then

$\Lambda$64

and if $\Lambda$65 has simple spectrum, then $\Lambda$66 has homogeneous spectrum of multiplicity $\Lambda$67 (Ryzhikov, 2012). The same conclusion holds for the geometric-series weak limit

$\Lambda$68

The paper also constructs a mixing staircase transformation with

$\Lambda$69

(Ryzhikov, 2012). Thus homogeneous spectral multiplicity can emerge from convolution disjointness of spectral measures.

6. Homogeneous spectra of operators and geometric contrasts

For one-dimensional discrete limit-periodic operators, homogeneity can be a property of the spectral set itself. A closed set $\Lambda$70 is homogeneous in Carleson’s sense if there exist $\Lambda$71 and $\Lambda$72 such that for every $\Lambda$73 and every $\Lambda$74,

$\Lambda$75

If one wishes to emphasize the constant, one says that $\Lambda$76 is $\Lambda$77-homogeneous (Fillman, 2014).

Within the class of one-dimensional discrete limit-periodic operators, a dense subset has spectra that are homogeneous Cantor sets in this sense. More precisely, fixing a periodic positive off-diagonal sequence $\Lambda$78, letting $\Lambda$79 denote the space of real-valued limit-periodic sequences, and letting $\Lambda$80 be the set of $\Lambda$81 such that $\Lambda$82 is a $\Lambda$83-homogeneous Cantor set and $\Lambda$84 has purely absolutely continuous spectrum, one has:

For every $\Lambda$85, $\Lambda$86 is dense in $\Lambda$87.

The proof uses periodic approximants, Floquet theory, and controlled gap opening near $\Lambda$88-break points. Quantitative control comes from the Hausdorff continuity estimate

$\Lambda$89

and band-length estimates such as

$\Lambda$90

for each band $\Lambda$91 of a $\Lambda$92-periodic Jacobi operator. The limiting lower-density estimate is

$\Lambda$93

with $\Lambda$94, which yields Carleson homogeneity; superexponential convergence of periodic approximants then gives purely absolutely continuous spectrum via Egorova’s theorem (Fillman, 2014).

A related contrast is provided by the planar measure

$\Lambda$95

supported on the union of two perpendicular unit line segments. This measure is spectral precisely when there exists a countable $\Lambda$96 such that

$\Lambda$97

forms an orthonormal basis of $\Lambda$98 (Kolountzakis et al., 20 Jan 2025). The classification is nearly complete: for $\Lambda$99, $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$00 is not spectral; for all irrational $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$01, $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$02 is not spectral; for rational $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$03, earlier work implies that $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$04 is spectral iff $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$05; and the only unresolved case is $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$06, the plus-space case (Kolountzakis et al., 20 Jan 2025). In all known spectral cases the spectrum is contained in a line, and there is a necessary and sufficient projection criterion for such line spectra. A plausible implication is that geometric symmetry alone is much weaker than the precise arithmetic or tree-like homogeneity that appears in the $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$07-adic and homogeneous Cantor constructions.

7. Conceptual synthesis

Across these theories, homogeneity controls spectral behavior in several distinct but structurally related ways. In homogeneous multifractality, it enforces spatial invariance of the multifractal spectrum and forces connected support on $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$08 (Buczolich et al., 2013). In homogeneous Cantor constructions, it supplies the recursive combinatorics needed for explicit Fourier spectra and for precise tuning of Hausdorff dimension (Dai et al., 2014). In affine IFS theory, common linear scaling and Hadamard duality support Cuntz-algebraic decompositions of $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$09 into cycle- or invariant-set-driven spectral components (Dutkay et al., 2010). In $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$10, tree homogeneity is exactly equivalent to both spectrality and tiling, and it survives in singular limit measures with explicit digit spectra (Fan et al., 2015). In operator theory on fractals, ampliation homogeneity of quasicentral modulus matches multiplicity scaling and leads to the Hausdorff-measure formula

$\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$11

for joint spectral measures (Voiculescu, 2020). In ergodic theory, homogeneous spectrum refers to constant multiplicity and is closely tied to convolution disjointness such as $\{e^{-2\pi i \lambda x}:\lambda\in\Lambda\}$12 (Ryzhikov, 2012). In limit-periodic spectral theory, homogeneity is a lower-density property of the spectral set itself and is compatible with Cantor geometry and purely absolutely continuous spectral type (Fillman, 2014).

The term “homogeneous spectral measures” therefore designates a family of ideas rather than a single doctrine. What unifies them is the presence of an exact uniformity principle—across intervals, scales, tree levels, multiplicities, or local spectral neighborhoods—that makes Fourier bases, tilings, quasicentral formulas, or multiplicity statements rigid enough to classify.