🔢 mathematics

Projections of convex polytopes to a line and higher univariate Prony systems

Motivated by the inverse moment problem for convex polytopes, this paper investigates the pushforward of Lebesgue measure to a line, characterizing the resulting spline densities and their moments through higher univariate Prony systems, fixed-knot spline cones, and their connection to Hankel determinantal varieties.

Original authors: Boris Shapiro

Published 2026-05-18

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Boris Shapiro

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you have a mysterious, solid 3D object (like a complex crystal or a polyhedral gem) sitting in a dark room. You cannot see the object itself, but you have a powerful flashlight that can shine a beam of light through it from any angle.

When the light passes through the object, it hits a screen on the other side. What you see on the screen isn't the object itself, but a shadow. However, this isn't just a flat, black-and-white shadow. Because the object has volume, the shadow has "thickness" or density. Where the object is thick, the shadow is dark; where it is thin, the shadow is light.

This paper is about a mathematical detective story: Can we figure out the shape of the original 3D object just by looking at these 1D shadows and their densities?

Here is a breakdown of the paper's main ideas using simple analogies:

1. The "Shadow" is a Spline (The Smooth Curve)

In the old days, mathematicians studied objects made of just a few distinct points (like a few marbles). If you shine a light on marbles, you get a few distinct dots on the screen. This is the "classical" problem.

But this paper looks at solid shapes (polytopes). When you shine a light on a solid shape, the shadow isn't a few dots; it's a smooth, continuous curve that rises and falls.

The Analogy: Imagine slicing a loaf of bread. If you look at the slices from the side, the width of the bread changes smoothly. The paper proves that the "density" of this shadow is always a specific type of smooth curve called a spline.
The Rule: If the object is a 3D shape, the shadow curve is made of pieces of quadratic curves (like parabolas). If it's a 4D shape, the pieces are cubic curves, and so on. The paper calls this a "spline density."

2. The "Higher Prony System" (The Recipe for Reconstruction)

The mathematician de Prony invented a famous recipe in 1795 to figure out where marbles were located based on their shadows. This paper creates a "Higher" version of that recipe for solid shapes.

The Problem: You are given a list of numbers (called "moments") that describe the shadow's shape. These numbers are like the ingredients in a recipe.
The Solution: The paper shows that even though the shadow is a smooth curve, these numbers still hold a secret code. By using a specific mathematical formula (a "recurrence relation"), you can decode the numbers to find:
1. The Knots: The specific points where the shadow's shape changes (these correspond to the vertices of the original 3D shape projected onto the screen).
2. The Amplitudes: How much "weight" or volume is associated with each section of the shadow.

Think of it like listening to a chord on a piano. Even though the sound is a smooth wave, if you analyze the frequencies (the moments), you can figure out exactly which keys were pressed (the nodes) and how hard they were hit (the amplitudes).

3. The "Spline Cone" vs. The "Real Polytope" (The Filter)

Here is a crucial distinction the paper makes:

The Spline Cone: This is a collection of all possible smooth curves that could be generated by our mathematical recipe. It's like a bag of all possible smooth shapes you could draw.
The Real Polytope: This is the much smaller group of shapes that are actually valid shadows of a real, solid, convex 3D object.

The Analogy: Imagine a bag of all possible smooth clay shapes (the Spline Cone). Most of these shapes are impossible to make from a single solid block of clay without cutting or gluing. The paper provides a filter (a set of rules) to tell you which shapes in the bag are actually valid "solid block" shadows.

For 2D shapes (polygons), the rule is simple: The shadow must be a "concave" curve (it must bulge outward, like a hill, not a valley).
For higher dimensions, the rule is more complex but follows the same logic: The shape must obey specific "convexity" laws.

4. The "Matching" Problem (The Multi-Angle Puzzle)

So far, we've talked about looking at the object from one direction. But what if you look at it from two or three different angles?

The Challenge: From Angle A, you see a list of shadow points. From Angle B, you see a different list. The paper points out that just knowing the lists isn't enough. You have to match them up.
The Analogy: Imagine you have a photo of a person's face from the front and a photo from the side. You know the nose is at a certain spot in the front photo and a certain spot in the side photo. But you have to figure out: Is the nose in the front photo the same nose as the one in the side photo?
The Finding: The paper shows that if you have enough angles, the math forces these lists to match up in very specific ways. If the lists don't match the rules of "compatibility," then no single solid object could have created those shadows. It's like a puzzle where the pieces from different angles must fit together perfectly to form one 3D picture.

Summary

This paper takes a complex problem—figuring out the shape of a solid object from its 1D shadows—and breaks it down into three steps:

Identify the Pattern: Realize the shadow is a specific type of smooth curve (a spline).
Decode the Signal: Use a new, upgraded version of an old math recipe (Prony system) to find the key points of the curve.
Verify the Shape: Check if the curve is "convex" enough to be a real solid object, and if you have multiple angles, ensure the shadows from all angles tell a consistent story about the same object.

The author isn't just solving a puzzle; they are providing the rulebook for how these puzzles fit together, distinguishing between "mathematically possible" curves and "physically real" solid shapes.

Technical Summary: Projections of Convex Polytopes to a Line and Higher Univariate Prony Systems

Problem Statement
The paper investigates the inverse moment problem for convex polytopes, specifically focusing on the pushforward of the Lebesgue measure restricted to a convex $d$ -dimensional polytope $P$ onto a line. While the classical Prony system (introduced by de Prony in 1795) addresses the recovery of a finite configuration of point masses from their moments, this work generalizes the problem to the case where the measure is continuous, represented by a spline density of degree $d-1$ . The central question is how to reconstruct the underlying geometric structure (nodes and amplitudes) from the moment sequence of such projected measures and to understand the algebraic relations governing these moments.

Methodology
The author employs a synthesis of algebraic geometry, spline theory, and moment problems. The methodology proceeds through the following steps:

Geometric Characterization: The pushforward measure is identified as a nonnegative spline density supported on the interval defined by the projected vertices of the polytope. The $d$ -th distributional derivative of this density is shown to be a discrete signed measure supported on the projected vertex set.
Spline Space Formulation: The problem is framed within a fixed-knot univariate spline space $L_{d,X}$ , generated by consecutive normalized simplex splines (B-splines) associated with the projected vertices.
Algebraic Reconstruction: The moment sequence is analyzed using rational generating functions. The paper establishes that normalized moments satisfy a linear recurrence relation determined by the projected vertex coordinates.
Hankel Matrix Analysis: The reconstruction of nodes is achieved via the kernel of a rectangular Hankel matrix constructed from the normalized moments. The recovery of amplitudes is reduced to solving a linear system involving a specific coefficient matrix derived from complete homogeneous symmetric polynomials.
Positivity and Realization: The paper distinguishes between the "spline relaxation" (measures in the positive simplicial cone generated by B-splines) and the "literal single-polytope cone" (measures arising from a single convex polytope). It utilizes the Brunn–Minkowski inequality to derive necessary and sufficient conditions for the latter in low dimensions.
Multidirectional Compatibility: The study extends to multiple projection directions, analyzing the compatibility constraints required for a set of univariate moment sequences to originate from a single common measure.

Key Contributions and Results

Higher Prony Systems: The paper defines a "higher" univariate Prony system (Definition 2.10) for $d \geq 0$ . For $d=0$ , this recovers the classical Prony system. For $d > 0$ , the system relates moments to a sum of B-spline measures with knots at the projected vertices.
Rational Generating Functions and Recurrence: A rational generating function for the scaled moments is derived (Theorem 2.8). It is shown that the normalized moments satisfy a linear recurrence of order $n$ (where $n$ is the number of vertices), with the characteristic polynomial having roots at the projected vertex coordinates.
Explicit Reconstruction Algorithm:
- Node Recovery: For generic data, the projected vertices are recovered as the roots of the kernel polynomial of a Hankel matrix formed from normalized moments (Proposition 2.11).
- Amplitude Recovery: Once nodes are known, amplitudes are uniquely determined by inverting a matrix $A_{d,n}(X)$ , whose determinant admits a closed factorization involving differences of the nodes (Proposition 2.12).
Positivity Criteria: The paper provides explicit linear inequalities that characterize the positive simplicial spline cone in terms of the first $n-d$ normalized moments (Theorem 2.13).
Directional Moment Varieties: The directional moment variety for $d$ -polytopes with $n$ vertices is identified with a Hankel determinantal variety (Theorem 2.17). The prime ideal defining this variety is generated by the maximal minors of a specific Hankel matrix.
Single-Polytope Realization:
- For $d=2$ , the paper provides a complete characterization: a piecewise-linear density corresponds to a convex polygon if and only if it is nonnegative, concave, and vanishes at the endpoints (Theorem 4.1).
- For $d \geq 2$ , a necessary condition is derived via the Brunn–Minkowski theorem: the $(d-1)$ -th root of the slice-volume density must be concave (Proposition 4.3).
Multidirectional Compatibility: The paper demonstrates that independent KSS (Kohn–Sturmfels–Shapiro) conditions in multiple directions are insufficient for the existence of a common measure. It establishes that directional moments of order $r$ form a homogeneous polynomial of degree $r$ in the direction vector. Consequently, compatibility imposes linear constraints (Proposition 6.4) and nontrivial matching problems between projected vertices in different directions (Proposition 5.3).

Significance and Claims
The paper positions itself not as a reproof of the Kohn–Sturmfels–Shapiro (KSS) theory but as a reinterpretation of it through the lens of Prony reconstruction and inverse problems. The author claims the following specific contributions:

Explicit Prony Interpretation: Providing a clear reconstruction procedure for the higher-order system, including explicit formulas for node and amplitude recovery.
Positivity and Spline Cones: Clarifying the distinction between the cone of all positive spline combinations and the more restrictive cone of measures realizable by a single convex polytope.
Multidirectional Constraints: Identifying that the multidirectional inverse problem introduces fundamentally new compatibility constraints (beyond the single-direction Hankel relations) related to the matching of projected data and the algebraic structure of directional moments.

The work concludes that while the one-directional geometry is well-understood via KSS theory, the multidirectional setting presents a more rigid structure involving phase-retrieval-like matching problems and additional linear and nonlinear constraints, suggesting a rich area for further algebraic and geometric investigation.