Abstract: We investigate the intersection body of a convex polytope using tools from combinatorics and real algebraic geometry. In particular, we show that the intersection body of a polytope is always a semialgebraic set and provide an algorithm for its computation. Moreover, we compute the irreducible components of the algebraic boundary and provide an upper bound for the degree of these components.
Introduction
In the field of convex geometry, the study of "intersection bodies" is central to geometric tomography. An intersection body $IK$ of a star body $K$ is defined by its radial function $\rho_{IK}(u)$, which equals the $(d-1)$-dimensional volume of the central hyperplane section of $K$ perpendicular to the unit vector $u$. These objects gained prominence through the Busemann–Petty problem, which investigates whether the volume of a centrally symmetric convex body is uniquely determined by the volumes of its central sections. While intersection bodies are well-understood in two dimensions, they exhibit complex, often non-convex behavior in higher dimensions. This paper provides a rigorous algebraic characterization of these bodies for polytopes, bridging the gap between discrete geometry and polynomial systems.
The Problem of Algebraic Description
Despite their importance, a formal algebraic description of intersection bodies was previously lacking. While it was known that these bodies exist and can be non-convex, researchers lacked a computational framework to describe their boundaries using polynomial equations. The authors address this by asking whether these shapes can be described as semialgebraic sets—sets defined by finite boolean combinations of polynomial inequalities.
Methods and Combinatorial Insights
The authors' approach relies on the observation that the "combinatorial type" of a section (the structure of the intersection between the polytope and the hyperplane) is not arbitrary. Instead, it remains constant within specific regions of a hyperplane arrangement. This arrangement is defined by the normal fan of a zonotope $Z(P)$ associated with the polytope's vertices.
Within each region of this arrangement, the volume of the section can be computed by triangulating the resulting $(d-1)$-polytope and summing the determinants of its vertex coordinates. Because these coordinates are rational functions of the normal vector $u$, the resulting volume function—and thus the radial function of the intersection body—is piecewise rational. By applying the Tarski-Seidenberg theorem (quantifier elimination), the authors prove that the intersection body $IK$ is a semialgebraic set.
Key Results and Degree Bounds
The paper establishes several landmark results:
- Semialgebraicity (Theorem 3.2): Formally proves that the intersection body of any polytope is semialgebraic.
- Degree Bounds (Theorem 5.6): Provides a theoretical upper bound for the algebraic complexity of the boundary. The maximum degree of the irreducible components of the algebraic boundary is bounded by $f_1(P) - (d - 1)$, where $f_1(P)$ is the number of edges of the polytope.
- Computational Framework: The authors implemented an algorithm in
SageMathandOSCARto compute these bodies. For a 3D cube centered at the origin, they found the boundary consists of 14 irreducible components: 6 linear facets (degree 1) and 8 cubic surfaces (degree 3).
The visualization of the 3-cube's intersection body highlights the counter-intuitive nature of these objects. Unlike the original convex cube, the intersection body is distinctly non-convex, featuring "spikes" and curved patches that correspond to different combinatorial types of sections.
Conclusion and Future Directions
This work provides the first systematic study of the algebraic boundary of intersection bodies of polytopes. By linking the combinatorial properties of a polytope (via zonotopes) to the algebraic properties of its intersection body, the authors have opened new avenues for research in geometric tomography. Future work may focus on tightening the degree bounds, as empirical results often show lower degrees than the theoretical maximum, and extending these insights to non-polytopal convex bodies like cylinders or ellipsoids.
Summary updated