Source-linked AI summary
Optimal asymptotic bounds for spherical designs
Andriy Bondarenko, Danylo Radchenko, Maryna Viazovska
TL;DR
The paper addresses how small spherical t-designs can be while retaining existence for arbitrary degree and dimension. It uses Brouwer degree theory together with area-regular partitions to prove the Korevaar–Meyers conjecture: every N at least a dimension-dependent multiple of t^d is achievable.
Problem
The paper asks for asymptotic upper bounds on N(d, t), since earlier existence results did not indicate how large the minimum design size is.
Method
The proof applies Brouwer degree theory and constructs the required point map using area-regular partitions of S^d.
Results
For each N ≥ C_d t^d, there exists a spherical t-design in S^d consisting of N points.
Takeaways & Limitations
The theorem guarantees spherical t-design existence for every number of points above the dimension-dependent t^d threshold.
Abstract
from arXiv · showhide
In this paper we prove the conjecture of Korevaar and Meyers: for each $N\ge c_dt^d$ there exists a spherical $t$-design in the sphere $S^d$ consisting of $N$ points, where $c_d$ is a constant depending only on $d$.
1 Introduction
The paper studies asymptotic upper bounds for spherical t-designs and proves the Korevaar–Meyers conjecture, guaranteeing designs with any N at least a dimension-dependent multiple of t^d.
- Definitions: Spherical t-designs are point sets reproducing the relevant polynomial condition for all algebraic polynomials of total degree at most t.N(d, t) denotes the minimum number of points in such a design.
- Background: Tight designs attain the standard lower bound, but they rarely exist; for d ≥2 and t ≥4, possible tight-design degrees are restricted to 4, 5, 7, or 11.The introduction also notes that exactly eight tight spherical designs are known for d ≥2 and t ≥4.
- Motivation: Seymour and Zaslavsky established existence for all d and t, but their nonconstructive proof gives no estimate for N(d, t).Determining exact values can remain difficult even for small parameters.
- Motivation: For fixed d ≥2 and t →∞, the paper focuses on asymptotic upper bounds for the minimum size N(d, t).Earlier bounds improved from powers involving d^4 and d^3 to t^(d²+d)/2.
- Main result: For each N ≥ C_d t^d, the paper proves that S^d contains a spherical t-design consisting of N points.The constants C_d and c_d depend only on d.
- Main result: The result is slightly stronger than the original conjecture because it guarantees existence for every N above the threshold, rather than only for some admissible size.The paper uses Brouwer degree theory and organizes the proof through auxiliary results before proving the theorem.
2 Preliminaries and the main idea
The proof translates the spherical-design condition into a zero of a continuous map on a polynomial space, then constructs that map by moving an initially well-distributed point set along spherical gradient fields.
- Design criterion: A Riesz-represented polynomial G_x is associated with each sphere point, and x_1,…,x_N form a spherical t-design exactly when G_x1 + ··· + G_xN = 0.This reformulates the design condition as a vector equation in the polynomial space P_t.
- Construction of F: The spherical gradient is defined for polynomials in P_t using the Euclidean norm in R^(d+1).The construction relies on this sphere-tangent gradient field to keep the moving points on S^d.
- Topological setup: The construction seeks a continuous map F from P_t into (S^d)^N whose boundary behavior allows the Brouwer degree theorem to produce a zero.The map L from point configurations to P_t is composed with F to form f = L ◦ F.
- Topological setup: A zero f(Q) = 0 yields a configuration F(Q) whose N components form a spherical t-design.This follows from the reformulated criterion and the composition f = L ◦ F.
- Construction of F: The explicit map F starts from a well-distributed collection of points and moves each point along the spherical gradient vector field of a polynomial P.The motion is chosen to increase the sum of point evaluations on the relevant boundary.
3 Auxiliary results
The proof uses area-regular partitions of the sphere with controlled diameter, together with a Marcinkiewicz–Zygmund inequality for polynomial sampling. These tools provide the geometric and analytic estimates needed for the construction.
- Area-regular partitions divide S^d into N closed sets of equal measure 1/N with pairwise intersections of measure zero.
- Every N admits an area-regular partition whose norm is at most B_dN^-1/d.The partition norm is governed by the maximum geodesic diameter of its cells.
- Theorem C gives a spherical Marcinkiewicz–Zygmund inequality for degree-m polynomials sampled at arbitrary points x_i ∈ R_i when the partition norm is sufficiently small.The inequality applies to area-regular partitions satisfying a threshold involving r_d and m.
- A corresponding corollary applies the same sampling framework under a threshold involving m+1.
4 Proof of Theorem 1
The proof constructs a continuous map from polynomial space to N-tuples of points on S^d using an area-regular partition and differential equations. It then establishes the boundary condition required to finish Theorem 1.
- For N ≥ C_dt^d, the construction begins with an area-regular partition and one selected point x_i from each cell.The constant is chosen using the partition and sampling constants B_d and r_d.
- The map F is built from trajectories y_i solving differential equations on S^d with prescribed initial conditions.The trajectories remain on the sphere by the definition of the spherical gradient.
- Lipschitz continuity of the governing map ensures that each trajectory y_i is well defined and continuous in both the polynomial parameter and time.
- The resulting mapping F is continuous on polynomial space, reducing the remaining task to proving the stated boundary condition.
- The boundary estimate is obtained through Newton–Leibniz calculations, distance bounds, modified partitions, and repeated applications of the sampling inequality.The argument introduces enlarged partitions and partitions depending on a parameter s before combining the resulting inequalities.
- The final inequalities imply the boundary claim, completing Lemma 1 and the proof.