The packing spectrum: interpolation between Hausdorff and packing dimensions
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.
Ada: We're looking at the paper: "The packing spectrum".
Emmy: Here's the gist of the paper. The paper introduces a family of metric outer measures whose critical exponents form a packing spectrum interpolating between Hausdorff and packing dimensions.
Ada: This spectrum provides finer bounds than classical Hausdorff–packing inequalities by characterizing the behavior over intermediate ranges of scales.
Emmy: Defining the Packing Spectrum The packing spectrum dimPθ is defined using metric outer measures obtained by covering a set with carriers and measuring packings within each carrier over a prescribed scale window.
Ada: The function theta→dimPθ F is the packing spectrum of F.
Title and authors: Ada: Now, let's talk about the paper itself, "The packing spectrum: interpolation between Hausdorff and packing dimensions." The authors are Peizhi Liu, and they lay out this entire framework for connecting the Hausdorff dimension and the classical packing dimension through this new family of measures.
Emmy: It’s interesting because it moves beyond just knowing two fixed numbers; it gives us a continuous spectrum that interpolates between them as we vary the scale parameter theta from zero to one.
Kolya: So, if you were looking at a set, you wouldn't just get one dimension; you'd get this whole spectrum describing its behavior across different scaling windows.
Ada: Right. The core idea is that for analytic sets, they characterize each spectral value as the supremum of the Hausdorff dimension of a product with a compact testing set minus the testing set’s upper intermediate dimension at the complementary parameter.
Emmy: That sounds like they are using these compact sets to probe every spectral value, which is a significant structural insight. It suggests that these spectrum values are determined by how large you can make a Hausdorff product while controlling the intermediate dimension at that specific scale.
Kolya: So, it’s not just about measuring the set; it’s about characterizing the set's structure through these product relationships and intermediate dimensions.
The paper's summary: Ada: To summarize what they found in "The packing spectrum: interpolation between Hausdorff and packing dimensions," they provide a variational principle using windowed local mass exponents to prove local Lipschitz continuity below the packing endpoint, with a possible jump at one.
Emmy: That variational principle is key because it implies that the function describing the spectrum is locally Lipschitz on the interval from zero up to one minus an infinitesimal amount. But they admit there could be a sudden jump when you hit exactly one.
Kolya: So, the structure of this function isn't perfectly smooth everywhere; it has these points where it might not be continuous in that precise way, which is important for applying these bounds later on.
Ada: Then they characterize all attainable profiles—meaning all possible functions that could arise as a packing spectrum—by demanding monotonicity, continuity below the packing endpoint, and a sharp upper right Dini-derivative bound.
Emmy: That characterization tells us exactly what properties this function must have to be considered an attainable packing spectrum in R d. And they show that every admissible profile can actually be realized by some compact binary digit set whose uniform measure achieves the finite window and classical packing variational suprema.
Kolya: So, they’ve connected the abstract mathematical definition of a spectrum to something concrete you can actually construct using these digit sets and measures.
The paper's improvements: Ada: One major improvement they point out is the product bound, Theorem C, which states that for bounded sets E and F, the Hausdorff product dimension is less than or equal to the minimum of dimP theta E plus dim one minus theta F and dimP theta F plus dim one minus theta E.
Emmy: That inequality refines the classical Hausdorff-packing inequalities by introducing this parameter dependence on theta, which allows for more precise comparisons between dimensions of different sets.
Kolya: And they also have a bound for the packing spectrum itself, Theorem four point seven, which states that dimP theta of a product set is less than or equal to the minimum of dimP theta E plus dimP F and dimP theta F plus dimP E <ref:2610.10375#pg1>.
Ada: That second bound is particularly interesting because it shows how the spectrum behaves under products—it respects some form of additivity when you consider the packing spectrum itself.
Emmy: And they also have a way to characterize complementary digit sets, saying that their total dimension sums up to one dimension in d and their uniform measure relates directly to the Lebesgue measure on the unit cube.
Kolya: So, this whole set of product bounds suggests that these spectrum measures are behaving quite nicely when you combine different sets together.
Conclusion: Ada: To wrap things up on "The packing spectrum: interpolation between Hausdorff and packing dimensions," the authors show that if a function f satisfies certain conditions—monotonicity, continuity below the endpoint, and that sharp Dini-derivative bound—then there is a compact binary digit set ES whose packing spectrum dimP theta ES is exactly that function f(theta) for all theta from zero to one.
Emmy: That's the realization part; they proved that every valid theoretical profile corresponds to a physical construction using these digit sets, and the uniform measure on this set attains both the finite window and classical packing variational suprema simultaneously.
Kolya: From a probabilistic standpoint, it’s interesting because they link this spectrum to the measure-theoretic characterization where dimP theta E equals the supremum of probability over a set A of points where dmu theta x is greater than some value t.
Ada: And they also establish that the limiting invariant, called dimqP, is countably stable and biLipschitz invariant, which means it’s a stable measure you can rely on when taking limits as theta goes to one.
Emmy: The flat and fast realizations in Example six point two show that even when the Hausdorff and packing dimensions are identical for some sets, they can have different limiting values for this packing spectrum depending on how you approach it <ref:2610.10375#pg1>.
Kolya: So, what we’ve seen here is a very structured way to interpolate between these two fundamental dimensions using these measures, giving us tools to bound products and characterize the shape of sets.
Ada: This paper, "The packing spectrum: interpolation between Hausdorff and packing dimensions," gives us a finer way to understand set dimension by providing this continuous spectrum, which opens up new avenues for studying complex geometric objects.
math.MG
Submitted: 2026-10-07
Updated: 2026-11-11
Comments: minor revisions and text clarify; comments welcome
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 89/100
The gist: The paper introduces a family of metric outer measures whose critical exponents form a packing spectrum interpolating between Hausdorff and packing dimensions.
Key concepts
- Packing Spectrum (dimPθ)
- This is a function that measures the dimension of a set based on covering it with carriers and measuring packings within those carriers over specific scale windows. It interpolates between Hausdorff dimension (at theta=0) and packing dimension (at theta=1), providing intermediate scaling information.
- Attainable Profiles
- These are the possible functions that can be realized as the packing spectrum of some set in $\mathbb{R}^d$. The paper proves these profiles must be monotonic, continuous below the endpoint, and satisfy a specific Dini-derivative bound. Every such profile is guaranteed to correspond to a compact binary digit set.
- Measure-Theoretic Characterization
- This links the packing spectrum directly to probability measures. It states that $\dimP\theta E$ equals the supremum of the probability that a point sampled from a measure $\mu$ falls into $E$, weighted by a local exponent $d\mu\theta(x)$ related to how fast the measure decays at scale.
- Product Bounds
- The paper establishes inequalities for the packing spectrum of products of sets, such as $\dimP\theta (E \times F)$. These bounds relate the spectrum of a product to the individual spectra and complementary dimensions, offering refined ways to estimate the dimension of combined geometric structures.
Terminology
Summary
The paper introduces a family of metric outer measures whose critical exponents form a packing spectrum interpolating between Hausdorff and packing dimensions. p. 0 This spectrum provides finer bounds than classical Hausdorff–packing inequalities by characterizing the behavior over intermediate ranges of scales. p. 0
Defining the Packing Spectrum
The packing spectrum dimPθ is defined using metric outer measures obtained by covering a set with carriers and measuring packings within each carrier over a prescribed scale window. The function theta→dimPθ F is the packing spectrum of F. p. 1 p. 1
For s ≥ 0, θ ∈ [0, 1], and a nonempty labelled carrier (E, R), define Φsθ(E, R) = sup (X N i=1 (2ri) s: xi ∈ E, R1/(1−θ) ≤ ri ≤ R, xi − xj ≥ ri + rj (i ≠ j)). p. 0
For F ⊂ R d and 0 < δ 0].
The packing spectrum is countably stable, biLipschitz invariant, and satisfies the usual Hölder image bound. p. 1
Characterization of Attainable Profiles
Theorem D characterizes all attainable profiles by monotonicity, continuity below the packing endpoint, and a sharp upper right Dini-derivative bound.
Every admissible profile is realized by a compact binary digit set whose uniform measure attains the finite-window and classical packing variational suprema. p. 0
Theorem 5.8 states that there is a nonempty bounded set E ⊂ R d with dimPθ E = f(θ) for every θ ∈ [0, 1] if and only if f is nondecreasing on [0, 1], continuous on [0, 1), and satisfies (5. p. 221).
Product Bounds and Complementary Sets
Theorem C provides the product bound: dimH(E × F) ≤ min(dimPθ E + dim1−θF, dimPθ F + dim1−θE).
The converse characterization (1. p. 04) identifies the spectrum as the optimal value of these product tests. p. 2
Theorem 4.7 gives the product bound for the packing spectrum itself: dimPθ(E × F) ≤ min(dimPθ E + dimP F, dimPθ F + dimP E). p. 2
The complementary digit sets satisfy ES + ESc = [0, 1]d, µ∗ ν = L d↾ [0, 1]d. p. 3
Measure-Theoretic Characterization
Theorem 3.2 gives the measure-theoretic characterization: dimPθ E = sup µ probability supp µ⋐E ess inf x∼µ dµθ(x). p. 1
Proposition 3. p. 17 states that if A subset of supp µ is Borel and dµθ(x) > t on A, then dimPθ A ≥ t. p. 1
Corollary 3. p. 39 shows that P−θ(µ) = ess inf x∼µ dµθ(x), P + θ(µ) = ess sup x∼µ dµθ(x). p. 1
Attainable Profiles and Realization
The uniform digit measure µS is the distribution of P k∈S 2−kUk.
The dimensions of ES are dimPθ ES = d lim inf N→∞ max k∈W1/(1−θ) AS(k) k (0 ≤ θ < 1). p. 1
The complementary set satisfies dimHESc = d − dimPθ ES (0 ≤ θ ≤ 1). p. 3 p. 3 p. 3
Comparison with Other Dimensions
Example 6.5 distinguishes the packing spectrum from Angelini’s packing intermediate dimension. p. 1
The complementary-window bound gives 1/2 ≤ dimH(E × F) ≤ 3/4−θ4 (0 < θ < 1). p. 2
Proposition 6. p. 14 shows that dimAPθ K = dimθK = dimmodθ K (0 < θ ≤ 1) under the spatial mass comparison assumption (5. p. 71).
Product Bounds for the Packing Spectrum
Theorem 4.7 bounds the supremum involving Hausdorff dimension and the upper intermediate dimension of the testing set. p. 0
The lower bound is established by choosing appropriate exponents and measures, leading to dimPθ(E × F) ≥ dimPθ E + t. p. 0 p. 0
Example 6.1 attains the optimized bound as θ ↑ 1. p. 0 p. 0
Classification of Attainable Profiles
Theorem 5.8 establishes the necessary and sufficient conditions for a function f to be an attainable packing spectrum in R d. p. 6
The proof shows that if f satisfies its conditions, there is a compact binary digit set ES ⊂ [0, 1]d with dimPθ ES = f(θ) for all θ ∈ [0, 1]. p. 0
Geometric Criteria for Dimension Equality
Proposition 6.9 relates the Hausdorff product dimension to the packing spectrum and modified intermediate dimensions. p. 0
The proof shows that if dimH(E × F) − dimH F ≥ dimP E − ε, then dim0+F − dimH F ≥ dimmod 0+ F − dimH F ≥ dimP E - limqP E - ε. p. 2
The limiting invariant is countably stable and biLipschitz invariant. p. 1
Complementary Digit Pairs
Lemma 5.7 shows that for a nonempty compact set E, dimH(ES × F) = d. p. 0
This suggests an additive version of product testing. p. 0
Limiting Invariants
The limiting invariant dimqP is countably stable and biLipschitz invariant. p. 1
The flat and fast realizations in Example 6. p. 2 p. 22 have limiting values h and d, respectively, despite their identical Hausdorff and packing dimensions. p. 3 p. 3
Attainability of the Spectrum
Corollary 5.11 states that f is an attainable packing spectrum if and only if h(η) = d − f(1−η) is an attainable upper intermediatedimension profile in R d. p. 0
Further Questions
Question 6.6 asks which ordered pairs of nonempty compact sets E ⊂ R m and F ⊂ R n satisfy dimH(E × F) = inf 0≤θ≤1 dimPθ E + dim1−θF. p. 2
Question 6.7 asks if dimPθ E = n − inf B∈C(E) dim1−θB? p. 2
Question 6.8 asks for which natural classes of nonempty analytic sets E ⊂ R d does dimAPθ E = dimmodθ E (0 < θ < 1) hold? p. 1
Summary of Findings
The paper establishes a packing spectrum interpolating between Hausdorff and packing dimensions. It provides variational principles in terms of windowed local mass exponents and sharp parameter estimates implying local Lipschitz continuity. p. 0 It characterizes all attainable profiles by monotonicity, continuity below the packing endpoint, and a sharp upper right Dini-derivative bound. p. 0
Pruning redundant reasoning tokens
The packing spectrum dimPθ is defined using metric outer measures obtained by covering a set with carriers and measuring packings within each carrier over a prescribed scale window. The function theta→dimPθ F is the packing spectrum of F.
The theorem establishes that P sθ is a metric outer measure, and it satisfies P s 0 = 2sHs, P s ≤ Ps1 ≤ MdP s, where Hs is Hausdorff measure with the diameter-cost normalization. p. 1
The packing spectrum is locally Lipschitz on [0, 1), although a jump at 1 is possible. p. 1
How the training signal is built
The measure-theoretic characterization involves the extended nonnegative local exponent dµθ(x) = lim inf R↓0 sup Rp≤r≤R log µ(B(x, r)) log r.
Theorem 3.2 states that dimPθ E = sup µ probability supp µ⋐E ess inf x∼µ dµθ(x). p. 1
The packing spectrum admits a local mass description, which will also be used in the compact testing construction below. p. 1
Where the gains come from
Theorem 4.1 bounds Hausdorff product dimension using the packing spectrum of one factor and the upper intermediate dimension of the other. p. 0
The lower bound is established by choosing appropriate exponents and measures, leading to dimPθ(E × F) ≥ dimPθ E + t.
Example 6.1 attains the optimized bound as θ ↑ 1.
The paper establishes that for every f satisfying its conditions, there is a compact binary digit set ES ⊂ [0, 1]d with dimPθ ES = f(θ) for all θ ∈ [0, 1]. p. 0
The complementary set satisfies dimHESc = d − dimPθ ES (0 ≤ θ ≤ 1).
The limiting invariant dimqP is countably stable and biLipschitz invariant.
The flat and fast realizations in Example 6.2 have limiting values h and d, respectively, despite their identical Hausdorff and packing dimensions.
The complementary set satisfies dimHESc = d − dimPθ ES (0 ≤ θ ≤ 1).
The addition of the digit set and the uniform digit measure yields X + Y = P k≥1 2−kUk is uniformly distributed on [0, 1]d. p. 0
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved dimension characterization for analytic sets by using compact testing sets to characterize every spectral value as
the supremum of the Hausdorff dimension of a product with a compact testing set minus the testing set’s upper intermediate dimension at the complementary parameter.
-
Enhanced bounds on Hausdorff product dimensions by using windowed local mass exponents, which can give
exact Hausdorff additivity when both factors have distinct endpoint dimensions
and refine classical inequalities. -
Robustness in determining attainable profiles by characterizing all admissible profiles by
monotonicity, continuity below the packing endpoint, and a sharp upper right Dini-derivative bound,
ensuring every admissible profile is realized by a compact binary digit set. -
Improved product bounds using complementary windows, where
dimH(E × F) ≤ min(dimPθ E + dim1−θF, dimPθ F + dim1−θE)
and an improved version for arbitrary sets withdimmod 1−θ.
-
Verification of the packing spectrum through a measure-theoretic characterization, where
dimP θ E = sup µ probability supp µ⋐E ess inf x∼µ dµθ(x).
-
Realization of admissible profiles via binary digit sets, where the set ES satisfies
dimPθ ES = f(θ) for all θ ∈ [0, 1]
and the uniform digit measure also attains the classical packing supremum. -
Direct determination of limiting dimension invariants by showing that
dimqP is countably stable and bi-Lipschitz invariant,
allowing for computation of the limit of the packing spectrum as theta goes to one.
Sources
- New Hausdorff type dimensions and optimal bounds for bilipschitz invariant dimensions
- Equivalences between different forms of the Kakeya conjecture and duality of Hausdorff and packing dimensions for additive complements
Related papers
- Axis-parallel square packing with an exact formula for optimal packing of seven squares
- The microscopic weighting on a metric space
- Unrectifiable surfaces with intrinsic regularity in Carnot groups
- Monge's transport problem in synthetic Lorentzian spacetimes
- Quasiconformal uniformization and extension for planar trees
- The Observable Wasserstein Distance