← All problems

Propeller conjecture

7 years old · traced to

Let m>3m>3, let a1,…,am>0a_1,\dots,a_m>0 satisfy ∑i=1mai=1\sum_{i=1}^m a_i=1, and let w∈Rn+1w\in\mathbb{R}^{n+1} be the vector defined in the paper. Set w‾(i)=w/ai\overline{w}^{(i)}=w/a_i. Consider measurable sets Ω1,…,Ωm⊆Rn+1\Omega_1,\dots,\Omega_m\subseteq\mathbb{R}^{n+1} covering Rn+1\mathbb{R}^{n+1}, with γn+1(Ωi)=ai\gamma_{n+1}(\Omega_i)=a_i, that maximize

π2∑i=1m∥∫Ωi(x−w‾(i))γn+1(x) dx∥2.\sqrt{\frac{\pi}{2}}\sum_{i=1}^m\left\|\int_{\Omega_i}(x-\overline{w}^{(i)})\gamma_{n+1}(x)\,dx\right\|^2.

Assume m−1≤n+1m-1\le n+1. Propeller Conjecture. The maximizing sets Ω1,…,Ωm\Omega_1,\dots,\Omega_m are simplicial cones over a regular simplex. This conjecture is a variant of problems motivated by kernel clustering and generalized Grothendieck inequalities. The source relates it to the small-correlation limit of the standard simplex conjecture, but does not establish the stated general case.

References
  1. Steven Heilman, “Stable Gaussian Minimal Bubbles”, arXiv:1901.03934 (2019).

Progress summary

Refreshed
Open · progress reported

A claimed 2026 proof settles a related equal-weight problem, but the weighted conjecture here remains open.

The conjecture predicts regular-simplex conical maximizers for the stated weighted Gaussian partition problem; the 2019 formulation records the case m>3m>3 as open.

Known results

  • Khot and Naor proved the m=3m=3 case in two dimensions.
  • Heilman, Jagannath, and Naor proved the three-dimensional unweighted propeller result in 2013, using rigorously verified computer-assisted inequalities.
  • A 2008 formulation conjectured the three-sector propeller in every dimension and described the problem as open.
  • A 2022 result gives only a conditional proof for 3≤m≤43\leq m\leq 4.

October 23, 2026 claimed related resolution

Steven Heilman’s article On the Propeller Conjecture reports a general-dimensional unweighted bound, but the argument still uses numerical checks without rigorous interval error bounds. It therefore does not establish the weighted, prescribed-measure conjecture here, and the claim remains unverified.

Current status (as of November 2026): The stated weighted conjecture remains open; low-dimensional and conditional cases are known, while the 2026 claim concerns a different problem and is unverified.

Sources

Solutions 0

No solutions have been posted yet.