One irreducible representation whose square contains every other

← All digests

Family 205 · Algebra
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-21

1. The question

The symmetric group $S_n$ contains every permutation of $n$ objects. Its irreducible complex representations are the basic linear ways in which those permutations can act. There are many different types, and their number grows rapidly with $n$.

Could one irreducible representation be so rich that combining it with itself produces every irreducible type? This is the universal tensor-square problem. The same representation must work for all targets; choosing a different starting representation for each desired constituent would be a much weaker statement.

Saxl’s conjecture makes a particularly elegant choice at triangular values of $n$: use the representation whose Young diagram is a staircase. Family 205 contains two manuscripts, both dated 24 September 2026. M2 claims Saxl’s conjecture, and M1 uses a strengthened form of that result to claim universal squares for every positive degree except $2$, $4$ and $9$.

2. Background and essential definitions

Partitions and representations

A partition $\lambda\vdash n$ is a decreasing sequence of positive integers with sum $n$. Draw $\lambda_i$ boxes in row $i$, aligned on the left, to obtain its Young diagram. For example, $(3,2,1)$ is a staircase containing six boxes.

The irreducible complex representations of $S_n$ are indexed by partitions of $n$. Write $S^\lambda$ for the Specht module with label $\lambda$. Transposing the diagram gives the conjugate partition $\lambda^t$.

The tensor product $S^\lambda\otimes S^\mu$ uses the diagonal action: a permutation acts on both factors simultaneously. Because the coefficients are complex and the group is finite, the tensor product decomposes as a direct sum of irreducibles:

$$S^\lambda\otimes S^\mu \cong\bigoplus_{\nu\vdash n}(S^\nu)^{\oplus g(\lambda,\mu,\nu)}.$$

The nonnegative integers $g(\lambda,\mu,\nu)$ are Kronecker coefficients. Positivity means that the target representation occurs at least once. A universal square requires $g(\lambda,\lambda,\nu)>0$ for every partition $\nu$ of the same integer.

Why the diagram must be self-conjugate

The partition $(n)$ labels the trivial representation, while $(1^n)$ labels the sign representation, which assigns $+1$ to even permutations and $-1$ to odd ones. Tensoring a Specht module with the sign representation transposes its label.

Consequently,

$$g(\lambda,\lambda,(1^n))= \begin{cases} 1,&\lambda=\lambda^t,\\ 0,&\lambda\ne\lambda^t. \end{cases}$$

A universal square must therefore start from a self-conjugate diagram. The staircase

$$\rho_m=(m,m-1,\ldots,1),\qquad N_m=\frac{m(m+1)}2,$$

has exactly that symmetry. The conjecture asks whether its square covers every irreducible of $S_{N_m}$.

“Square” here means the full tensor product. It is not the symmetric square alone, which is only one part of the tensor product and has a separate representation-theoretic meaning.

3. Development history and previous partial results

Tensor-square covering questions have precedents for finite simple groups of Lie type, where Steinberg representations provide important examples. Saxl proposed the staircase conjecture in a UCLA seminar in March 2012. Pak, Panova and Vallejo placed it in the broader symmetric-group tensor-square problem and proved positivity for substantial families of targets.

The difficulty is not a shortage of evidence. It is the possibility of an exceptional constituent that remains absent even when almost every other partition occurs.

Ikenmeyer proved the staircase assertion for partitions comparable with the staircase in dominance order. Dominance compares all initial row sums: $\lambda$ dominates $\mu$ if the first $k$ rows of $\lambda$ contain at least as many boxes as those of $\mu$, for every $k$. His construction with two triangular layers of alternating tensors is an explicit methodological predecessor of M2.

Character-theoretic, spin-character and modular methods established further support families. The papers cite Bessenrodt’s double-hook results, Li’s triple-hook theorem, and Bessenrodt, Bowman and Sutton’s results using modular decomposition numbers. Double and triple hooks here refer to diagrams whose largest square has side two or three.

Luo and Sellke proved that almost every partition occurs in staircase squares, for both the uniform and Plancherel probability distributions. They also obtained universal fourth tensor powers in all sufficiently large degrees. Harman and Ryba later proved universal coverage for every staircase tensor cube.

These achievements leave two precise gaps. Probability tending to one does not eliminate every exceptional target. A cube or fourth power has more room than a square, so its universality does not establish the two-factor claim.

There was also a recent staircase-square theorem for unipotent characters of finite general linear groups, due to Letellier and Nam. It is an analogue in another character theory. Ordinary Kronecker positivity transfers to the relevant unipotent positivity, but that implication does not reverse automatically.

M2 claims to close the staircase-square gap through a specific cyclic generator. M1 then addresses all nontriangular degrees by enlarging staircases in a controlled, self-conjugate way.

4. What the new manuscripts claim

M2: every staircase square is universal

For every $m\geq1$ and every $\nu\vdash N_m$, M2 asserts

$$g(\rho_m,\rho_m,\nu)>0.$$

It proves a stronger stated result inside the square. The manuscript specifies one vector $w_m$, built by alternating along rows in one tensor layer and columns in the other. Its orbit span

$$W_m=\operatorname{span}_{\mathbb C}\{\sigma w_m:\sigma\in S_{N_m}\}$$

already contains every irreducible type. Such an orbit-generated submodule is called cyclic. The assertion is about support, not about containing every multiplicity present in the entire square.

M1: universal squares in all eligible degrees

M1 asserts that, for every positive integer $n\notin\{2,4,9\}$, there is one partition $\lambda_n\vdash n$ such that

$$g(\lambda_n,\lambda_n,\nu)>0 \quad\text{for every }\nu\vdash n.$$

The partition is chosen from $n$ before the target $\nu$ is specified. For large degrees, it comes from a parity-compatible staircase with matching additions to its first two rows and columns. Small degrees are handled separately by the manuscript’s exact character calculations. The reported calculation also finds no covering square in degrees $2$, $4$ and $9$.

The dependence is explicit: M1 uses M2’s cyclic theorem, not merely its bare positivity conclusion. The prescribed generator is needed to ensure constituents survive the construction of enlarged diagrams.

M1 also obtains universal coverage of all unipotent irreducible characters of $\operatorname{GL}_n(\mathbb F_q)$ by squares with these same labels, for every prime power $q$. It credits Letellier’s transfer theorem. Universal unipotent coverage itself already had other witnesses, including the Steinberg character, so the novelty of this corollary is the additional family of labels.

5. Why the advance matters

Importance: 5/5, provisional. The family claims resolutions of two central tensor-support conjectures, uniformly over all target partitions and all eligible degrees.

Kronecker coefficients are difficult even to understand qualitatively. A universal square theorem supplies a remarkably simple global support pattern inside this complicated landscape: one irreducible can interact with itself to reach every irreducible type.

The cyclic refinement is also meaningful. Knowing that a constituent exists somewhere in a large tensor product does not guarantee that it appears in the submodule preserved by an inductive construction. A specified generator provides information that can be transported to the next step.

The score concerns these structural advances. The manuscripts do not claim a general positive combinatorial formula for arbitrary Kronecker coefficients, and their results should not be presented as solving that broader problem.

6. The main idea

Preserve a particular generator

Arrange $N_m$ positions in a triangular diagram. In the first layer, alternate basis letters within each row; in the second, alternate within each column. Alternation is the familiar signed sum over permutations, analogous to the construction of a determinant.

The two layers produce $w_m$. Rather than changing vectors whenever a new target is considered, M2 studies the orbit span of this one vector throughout the induction.

The first targets are those dominated by the staircase. A sign twist and a permutation-module quotient supply them. The remaining targets are reached by cutting off a narrow band from the triangular position set.

Separate a smaller triangle from a band

Divide the letters into low and high alphabets. A coordinate projection keeps only words whose two letters at each position are both high or both low. The row and column counts force the high positions to form a smaller triangle in the projected generator.

Its projected vector factors into the smaller cyclic staircase vector and a band vector. Different placements of the smaller triangle occupy disjoint coordinate sectors. This separation makes the resulting quotient the full induced representation from the two factors, rather than an uncontrolled part of it.

Here induction means enlarging a representation from a product of symmetric groups on disjoint position sets to the symmetric group on all positions. It is distinct from the diagonal tensor product whose constituents the theorem studies.

Make the width-two band cover four-row targets

A width-two band forms an odd path of alternating pairs. Pairing it with suitable dual tensors becomes a product of two-by-two matrices and their adjugates. A nonzero pairing detects the desired irreducible constituent.

The paper chooses a matrix basis so that alternated path segments of length at most four give invertible matrices. A generic change of basis then makes the required endpoint coefficient nonzero. This proves coverage of every target with at most four rows inside the specific cyclic band module.

The ambient tensor-square coverage of these four-row shapes had already been established by Brown, van Willigenburg and Zabrocki. The generator-specific placement is the extra statement needed here.

A diagram-removal argument completes the induction. A target outside the starting dominance range either allows one horizontal strip to be removed, or allows exactly $2m-1$ boxes to be removed in at most four horizontal strips. A horizontal strip has at most one box in each column. The Pieri rule translates these removals into induced-representation constituents, joining the smaller triangle to the band.

Enlarge the staircase without losing targets

M1 adds transposed pairs of cells, preserving self-conjugacy. For example, its degree-49 construction has row and column lengths

$$(10,9,7,6,5,4,3,2,2,1).$$

Two complementary tests cover possible target diagrams. One detects a supported band with attachments inside the target; the other handles targets whose first few rows or columns are sufficiently small. If both fail, simultaneous row and column capacity bounds eventually force the target to have fewer boxes than its prescribed size.

The remaining bounded parameters and eligible degrees through $64$ are handled by exact computations supplied in the manuscript. These are stated components of the source argument; the digest does not add a new execution or verification of those programs.

7. Scope, limitations, and open questions

All representations here are finite-dimensional complex representations. The claims do not automatically transfer to modular representations in characteristic dividing $n!$, where semisimplicity fails.

The theorem asserts positive multiplicity, not a prescribed multiplicity or an equality with a regular representation. A universal square may contain some constituents many times and others only once.

The existential theorem for arbitrary degree uses particular self-conjugate candidates. It does not say every self-conjugate partition outside the exceptional degrees has a universal square. Stronger conjectures concerning the symmetric and alternating parts of tensor squares also remain distinct.

Finally, the argument’s finite computations and companion dependency are part of its stated scope. M1 should not be summarized as following from the staircase positivity theorem by simply adding boxes: preserving constituents through the actual generator and controlling cancellation are substantive additional tasks.

8. Sources and further reading

Pak, Panova and Vallejo, Kronecker products, characters, partitions, and the tensor square conjectures, Advances in Mathematics 288 (2016), 702–731, gives the conjectures and early support results.

Ikenmeyer, The Saxl conjecture and the dominance order, Discrete Mathematics 338 (2015), 1970–1975, supplies the triangular alternating-tensor predecessor. Luo and Sellke, The Saxl conjecture for fourth powers via the semigroup property, Journal of Algebraic Combinatorics 45 (2017), 33–80, covers asymptotic support and fourth powers.

Harman and Ryba, A tensor-cube version of the Saxl conjecture, Algebraic Combinatorics 6 (2023), 507–511, proves the cube theorem. Brown, van Willigenburg and Zabrocki, Expressions for Catalan Kronecker products, Pacific Journal of Mathematics 248 (2010), 31–48, provides the ambient four-row band result.

Source manuscripts

• M1. OpenAI, Universal Tensor Squares for Symmetric Groups (September 24, 2026).
• M2. OpenAI, A Cyclic Polytabloid Proof of Saxl’s Conjecture (September 24, 2026).

Sources consulted on 21 October 2026 at repository revision 1581d4c. Download the manuscript citations.