OctoInform Beta

OpenAI preprint, September 2026 · §1. Introduction · §1

Theorem 1.1 [Saxl's conjecture]: Saxl’s conjecture for every m≥1m \ge 1: the Kronecker coefficient g(ρm,ρm,λ)g(\rho_m,\rho_m,\lambda) is positive for every partition λ\lambda of m(m+1)/2m(m+1)/2

The staircase Specht module’s tensor square contains every irreducible.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

For m≥1m \ge 1, the staircase ρm\rho_m, with rows of lengths m,m−1,…,1m, m-1, \dots, 1, is the Young diagram ρm={(i,j):i+j<m},\rho_m = \{(i, j) : i + j < m\},(1)✦ its cells (i,j)(i, j) numbered from 00 with ii the row, and Nm=|ρm|N_m = |\rho_m| is its number of cells. Saxl’s conjecture asks whether the tensor square Sρm⊗SρmS^{\rho_m} \otimes S^{\rho_m} of the Specht module SρmS^{\rho_m} of the staircase, with the symmetric group SNmS_{N_m} acting diagonally, contains every irreducible representation of SNmS_{N_m}; equivalently, whether the Kronecker coefficient g(ρm,ρm,λ)g(\rho_m, \rho_m, \lambda) is positive for every partition λ\lambda of NmN_m.

For m=3m = 3 the staircase is (3,2,1)(3, 2, 1), a partition of 66, its Specht module has dimension 1616, and the question is whether each of the eleven irreducible representations of S6S_6 occurs in a representation of dimension 256256.

The library builds Specht modules inside a space of words and states the conjecture there in both forms.

Specht Modules in the Space of Words

Fix nn positions 0,…,n−10, \dots, n-1 and dd letters 0,…,d−10, \dots, d-1; a word is a map ww from the positions to the letters. The word space Wn,dW_{n,d} is the complex vector space of functions from the words to ℂ\mathbb{C}.✦ Write e1,…,ede_1, \dots, e_d for the standard basis of ℂd\mathbb{C}^d. The indicator ewe_w of a word ww, the function on words that is 11 at ww and 00 at every other word, corresponds to the tensor ew(0)+1⊗⋯⊗ew(n−1)+1e_{w(0)+1} \otimes \cdots \otimes e_{w(n-1)+1}, which identifies Wn,dW_{n,d} with (ℂd)⊗n(\mathbb{C}^d)^{\otimes n}. The symmetric group SnS_n acts on Wn,dW_{n,d} by permuting the positions: (gf)(w)=f(w∘g),gew=ew∘g−1.(g f)(w) = f(w \circ g), \qquad g\, e_w = e_{w \circ g^{-1}}.(2)✦

For a representation VV of a group GG and a vector v∈Vv \in V, the cyclic subrepresentation ℂ[G]v\mathbb{C}[G]v is the subspace of VV spanned by the translates gvgv, g∈Gg \in G.✦

Let μ\mu be a Young diagram with nn cells and ℓ(μ)\ell(\mu) rows. A tableau tt of shape μ\mu is a bijection from the positions 0,…,n−10, \dots, n-1 to the cells of μ\mu.✦ The row word wtw_t of a tableau tt of shape μ\mu is the word in the letters 0,…,ℓ(μ)−10, \dots, \ell(\mu)-1 that sends each position ii to the row of the cell t(i)t(i).✦ The column group CtC_t of a tableau tt is the subgroup of SnS_n of the permutations gg such that t(g(i))t(g(i)) lies in the column of t(i)t(i) for every position ii.✦

The polytabloid of a tableau tt of shape μ\mu is the vector of Wn,ℓ(μ)W_{n,\ell(\mu)} et=∑σ∈Ctsgn(σ)σewt,e_t = \sum_{\sigma \in C_t} \mathrm{sgn}(\sigma)\, \sigma\, e_{w_t},(3)✦ and the Specht module StS^t of tt is the cyclic subrepresentation ℂ[Sn]et\mathbb{C}[S_n]\, e_t of Wn,ℓ(μ)W_{n,\ell(\mu)}, with SnS_n acting by (2).✦ For each Young diagram μ\mu, the canonical tableau tμt_\mu of μ\mu is a tableau of shape μ\mu fixed once for all, and SμS^\mu denotes the Specht module of tμt_\mu. Tableaux of the same shape give the same subspace. The staircase tableau is the canonical tableau of ρm\rho_m.✦

The Kronecker coefficient g(α,β,λ)g(\alpha, \beta, \lambda) of three Young diagrams with nn cells is the dimension of a space of intertwiners, g(α,β,λ)=dim⁡ℂHomSn(Sλ,Sα⊗Sβ),g(\alpha, \beta, \lambda) = \dim_{\mathbb{C}} \mathrm{Hom}_{S_n}\bigl(S^\lambda, S^\alpha \otimes S^\beta\bigr),(4)✦ with SnS_n acting diagonally on Sα⊗SβS^\alpha \otimes S^\beta.

The Claim

Theorem 1 (Saxl’s conjecture). Let m≥1m \ge 1 and let λ\lambda be a Young diagram. The Kronecker coefficient of the staircase with itself and λ\lambda is positive, for λ\lambda with as many cells as ρm\rho_m, 0<g(ρm,ρm,λ)for |λ|=|ρm|,0 < g(\rho_m, \rho_m, \lambda) \qquad \text{for } |\lambda| = |\rho_m|,(5)✦ and, for λ\lambda with m(m+1)/2m(m+1)/2 cells, 0<g(ρm,ρm,λ)for |λ|=m(m+1)/2.0 < g(\rho_m, \rho_m, \lambda) \qquad \text{for } |\lambda| = m(m+1)/2.(6)✦ Every finite-dimensional irreducible complex representation VV of SNmS_{N_m} has an injective intertwiner into the tensor square of the staircase Specht module, with SNmS_{N_m} acting diagonally: V↪Sρm⊗Sρm.V \hookrightarrow S^{\rho_m} \otimes S^{\rho_m}.(7)✦

What Is Checked

The claim is stated over the staircase (1) and its canonical tableau, the Specht modules Sμ=ℂ[Sn]etS^\mu = \mathbb{C}[S_n]\, e_t spanned by the translates of the polytabloids (3) in the word space with the action (2), and the Kronecker coefficient (4); (7) ranges over every finite-dimensional irreducible complex representation of SNmS_{N_m}.

The library proves that the staircase (1) is a Young diagram with Nm=m(m+1)/2N_m = m(m+1)/2 cells.✦ It also proves that the column group CtC_t is a subgroup of SnS_n and that a cyclic span ℂ[G]v\mathbb{C}[G]v is stable under GG.