---
title: 'Butterfly SHTs: Harmonic Index t Designs'
url: https://www.emergentmind.com/topics/butterfly-shts
type: topic
---

# Butterfly SHTs: Harmonic Index t Designs

A spherical design of harmonic index $t$ (“butterfly SHT”; Editor's term for compact reference) is a finite subset of the unit sphere $S^{n-1}\subset \mathbb{R}^n$ such that the sum of any real homogeneous harmonic polynomial of degree exactly $t$ vanishes over the set. This relaxation, strictly weaker than the classical spherical $t$-design condition, requires only the vanishing of moments for harmonics of degree $t$—not for all degrees up to $t$. The theory, formally introduced and systematically developed by Bannai, Okuda, and Tagami, provides new constructions, asymptotic cardinality bounds, rare “tight” minimal examples, and relates closely to equiangular line sets and symmetry phenomenon in finite reflection groups and root systems [1308.5101].

## 1. Definitions and Fundamental Properties

Let $S^{n-1}\subset \mathbb{R}^n$ denote the unit sphere with rotation-invariant measure $d\sigma$ normalized so that $\int_{S^{n-1}} d\sigma(x)=|S^{n-1}|$. A real polynomial $f(x)$ on $\mathbb{R}^n$ of degree at most $t$ is called homogeneous harmonic of degree $t$ if $\Delta f=0$, where $\Delta$ is the Laplacian.

- **Classical spherical $t$-design:** A finite, nonempty set $X\subset S^{n-1}$ is a $t$-design if for every real polynomial $f$ of degree $\leq t$,
  $$
  \frac{1}{|S^{n-1}|}\int_{S^{n-1}} f(x)\, d\sigma(x) = \frac{1}{|X|}\sum_{x\in X} f(x)
  $$
  An equivalent condition due to Delsarte–Goethals–Seidel: $X$ is a $t$-design if and only if for every harmonic, homogeneous polynomial $p$ of each degree $k=1,\dots,t$, $\sum_{x\in X}p(x)=0$.

- **Spherical design of harmonic index $t$:** $X\subset S^{n-1}$ is a design of harmonic index $t$ if
  $$
  \sum_{x\in X}p(x) = 0\ \text{for every}\ p\in \mathrm{Harm}_t(\mathbb{R}^n)
  $$
  where $\mathrm{Harm}_t(\mathbb{R}^n)$ is the space of homogeneous harmonic polynomials of degree $t$. Every classical $t$-design is a harmonic index $t$ design, but not conversely [1308.5101].

Let $A(n,t)$ denote the minimal cardinality of a harmonic index $t$ design in $S^{n-1}$.

## 2. Construction Principles and Explicit Families

A foundational construction is available via Theorem 1 of Bannai–Okuda–Tagami:

**Lifting Construction:** Given a classical spherical $t$-design $X\subset S^{n-2}$ and a real root $r$ of the Gegenbauer polynomial $Q_{n,t}(s)$ on $[-1,1]$, the set
$$
X' = \{\, (r,\sqrt{1-r^2}\,x): x\in X\,\} \subset S^{n-1}
$$
is a harmonic index $t$ design in $S^{n-1}$ [1308.5101].

In particular, because optimal classical $t$-designs exist on $S^{n-2}$ with $O(t^{n-2})$ points (by Bondarenko–Radchenko–Viazovska), this yields a harmonic index $t$ design in $S^{n-1}$ with $O(t^{n-2})$ points. For fixed $n$ and large $t$, such designs exist, and the construction offers a dimensional induction improvement over the $O(t^{n-1})$ for classical $t$-designs.

## 3. Lower Bounds, Tight Designs, and Equiangularity

For each harmonic index $t$ design $X\subset S^{n-1}$, a Fisher-type lower bound controls minimal size:
$$
|X|\ge 1+\frac{Q_{n,t}(1)}{C_{n,t}} =: b_{n,t}
$$
where $Q_{n,t}(s)$ is the normalized Gegenbauer polynomial of degree $t$ satisfying $Q_{n,t}(1)=\dim\mathrm{Harm}_t(S^{n-1})$, and $C_{n,t}=-\min_{s\in[-1,1]} Q_{n,t}(s)>0$ [1308.5101].

A harmonic-index-$t$ design attaining this lower bound is called “tight.” Tightness is extremely rare—attaining tightness forces that every pairwise inner product $(x,y)$ of distinct points satisfies $Q_{n,t}(x\cdot y) = -C_{n,t}$, so all pairwise lines are equiangular. The absolute upper bound on equiangular lines for $S^{n-1}$ (at most $n(n+1)/2$ lines) and the asymptotic behavior of $b_{n,t}$ for large $t$ mean tight harmonic-index-$t$ designs can only exist for small $t$ or $n$.

## 4. Examples for Small $n,t$ and Sporadic Symmetric Cases

The construction yields exact solutions for several special cases:
- For $n=2$, any even $t$ gives $b_{2,t}=2$; two antipodal points at angle $\pi/t$ form a minimal design: $A(2,t)=2$.
- For $t=2$, any $n\ge2$, $b_{n,2}=n$; the coordinate axis points ${e_1, ..., e_n}$ in a hemisphere yield $A(n,2)=n$.
- For $(n,t)=(3,4)$: $b_{3,4}=10/3$, $A(3,4)\ge4$, but explicit construction with the regular pentagon in $S^1$ gives $A(3,4)=5$ (no 4-point solution exists, and only two 5-point “pentagonal” configurations, up to antipodal symmetry, realize the minimum).
- “Sporadic” high-symmetry cases include: half of the icosahedron’s 12 vertices (size 6, $t=8,14$), half the 600-cell (120 points in $S^3$, $t=58$), and the $E_8$ root-system (120 points in $S^7$, $t=10$). The sizes in these cases are far below the general $b_{n,t}$ bound, indicating that such “exceptional” designs are tied to rare combinatorial symmetry [1308.5101].

## 5. Existence and Nonexistence of Tight Harmonic Index $t$ Designs

Tight designs must satisfy not only the Fisher-type lower bound but stringent matrix-theoretic and combinatorial constraints, notably maximizing the number of equiangular lines. For $t=4$, the only possible $n$ where $b_{n,4}$ is integral and not ruled out by other methods are $n=3p^2-4$ for odd $p\ge5$. Semi-definite programming and combinatorial nonexistence arguments further narrow possible cases. In small dimensions, all but one sequence of tight harmonic-index-4 designs are ruled out [1308.5101].

## 6. Comparison to Classical Spherical Designs and Extensions

While classical spherical $t$-designs require vanishing harmonic moments up to degree $t$, harmonic-index-$t$ designs sharply reduce the necessary conditions, allowing for smaller sets and more flexible configurations. For instance, for large $t$ and fixed $n$, harmonic-index-$t$ designs of size $O(t^{n-2})$ exist compared to $O(t^{n-1})$ for full $t$-designs.

Some open problems and research directions include:
- Obtaining sharp bounds (as opposed to generic $O(t^{n-2})$ rate) on minimal $A(n,t)$.
- Systematic classification and explicit construction of minimal harmonic-index-$t$ designs beyond low dimensions.
- Deeper investigation into the connections with equiangular line systems and root-system symmetries for producing potentially infinite families of such designs.

Minimal examples, uniqueness for given $(n,t)$, and full classification in low dimensions (e.g., the $A(3,4)=5$ pentagon case) remain open and of special interest. The correspondence with combinatorics of equiangular lines and finite group actions invites further exploration of the algebraic underpinnings [1308.5101].

## 7. Summary Table: Core Results and Examples

The following table succinctly summarizes key instances and bounds for harmonic-index-$t$ designs:

| $(n, t)$        | Lower bound $b_{n,t}$           | Minimal $A(n,t)$ and/or Example                  | Remarks                  |
|-----------------|---------------------------------|--------------------------------------------------|--------------------------|
| $(2, \text{even } t)$ | $2$                             | Any two antipodal points at $\pi/t$              | General for $n=2$        |
| $(n,2)$         | $n$                             | $n$ coordinate unit vectors                      | General for $t=2$        |
| $(3,4)$         | $10/3$                          | $5$ (two pentagonal configurations)              | Uniqueness proven        |
| $(n,t)$, sporadic  | varies                          | 6 (icosahedron), 120 ($E_8$), 120 (600-cell, $S^3$) | Exceptional symmetries   |

Tightness (equality in the lower bound) is rare and typically linked to deep combinatorial or algebraic structures.

---

Further details, rigorous proofs, and computational/numerical examples can be found in the foundational article by Bannai, Okuda, and Tagami [1308.5101]. The classification, existence, and potential applications of butterfly SHTs (harmonic-index-$t$ designs) remain an active topic closely tied to extremal problems in discrete geometry and the theory of point measures on spheres.

Source: https://www.emergentmind.com/topics/butterfly-shts