Introduction
In 1951, Thøger Bang proved a celebrated conjecture in discrete geometry: if a $d$-dimensional convex body is covered by a collection of "planks" (the region between two parallel hyperplanes), the sum of the widths of those planks must be at least the minimal width of the body itself. This result, known as Bang’s plank problem, has served as a cornerstone of the field for decades. However, the geometric community has long lacked a "dual" version of this theorem—one that addresses the constraints of packing rather than covering.
This paper establishes the first packing analogues to Bang’s problem. Rather than asking how much width is required to cover a shape, the authors investigate the upper limits of how many cylinders or planks can be packed inside a convex body. Furthermore, they extend these results to "$r$-fold" packings, a generalization where limited overlap is permitted, allowing any point in the host body to be covered by at most $r$ cylinders.
The Packing Framework
The authors focus on $k$-codimensional cylinders—objects that are "thin" in $k$ dimensions and "long" in the remaining $d-k$ dimensions. To quantify the space these objects occupy relative to their host body $K$, the paper utilizes the concept of "relative crossectional volume" ($ ext{crv}_K$).
In a standard packing, the cylinders must be disjoint. In the more flexible $r$-fold packing, the cylinders can overlap, provided that no point in the host body is contained in more than $r$ of them. This framework allows for a more robust analysis of geometric density and redundancy.
Methodology
The paper’s findings are rooted in rigorous proofs in convex and discrete geometry, employing several sophisticated analytical techniques:
- Density Functions: To derive bounds for ellipsoids, the authors use a specific weight function, $p(x) = 1/\sqrt{1-|x|^2}$. By integrating this function over the body, they can relate the measure of the host body to the sum of the measures of the packed cylinders.
- Banach-Mazur Distance: To generalize results from ellipsoids to arbitrary convex bodies, the authors employ the Banach-Mazur distance, which measures how closely a general convex shape can be approximated by an ellipsoid.
- Cauchy’s Surface Area Formula: This classical tool is used to estimate the total volume of the bases of 1-codimensional cylinders (planks) packed into general convex bodies.
- Sectional Integrals: Following methods introduced by Falconer, the authors use sectional integrals to analyze "NS-domains"—specific 2D arrangements of non-separable disks.
Key Results
1. The Ideal Case: Ellipsoids
Ellipsoids represent the most symmetric and mathematically "clean" case for packing. The authors prove (Theorem 4.2) that for a $d$-dimensional ellipsoid $K$, any $r$-fold packing of planks $C_i$ satisfies: $$\sum_{i=1}^{N} \text{crv}_K(C_i) \le r$$ This result shows that for ellipsoids, the sum of relative crossectional volumes is directly bounded by the overlap factor $r$, providing a remarkably simple and tight bound.
2. General Convex Bodies
Moving beyond ellipsoids, the authors establish bounds for any $d$-dimensional convex body $K$. In Theorem 5.1, they show that the total volume of the bases of packed 1-codimensional cylinders is bounded by: $$\sum \text{vol}_{d-1}(B_i) \le c_d \cdot r \cdot \text{max\_projection\_area}(K)$$ Here, the constant $c_d$ scales with the square root of the dimension ($c_d \sim \sqrt{\pi d / 2}$). This indicates that as the dimensionality of the space increases, the packing efficiency relative to the maximum projection area changes predictably.
3. 2D NS-Domains
For 2D shapes formed by non-separable disks (NS-domains), the authors provide a packing version of Falconer’s covering results. Theorem 6.1 states that the sum of the widths of planks in an $r$-fold packing is at most $r \cdot \text{diam}_{NS}(K)$, where $\text{diam}_{NS}$ is the NS-diameter.
Discussion and Future Directions
While this work provides the first comprehensive look at packing analogues for the plank problem, several questions remain open. A primary area for future research is the refinement of the constant $c_d$ in Theorem 5.1. The authors suggest that it would be significant to determine if $c_d$ could be replaced by an absolute constant independent of the dimension $d$.
Additionally, the authors pose a characterization problem: which specific convex bodies, other than ellipsoids, satisfy the packing bounds with a constant of exactly 1? Finally, while the paper provides some lower bounds (Theorem 4.8), a significant gap remains between the known upper and lower bounds for general $k$-codimensional cylinders, leaving room for further theoretical tightening.
Summary updated