OctoInformBeta

OpenAI preprint, September 2026 · §5. Representation comparison and scope

Corollary 4: Embeddings between iterated symmetric powers, including the sixth power for natural degrees b≥30b\ge 30

Equivariant comparison under a quadratic bound.

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

Let VV be a complex vector space and let a,ba,b be natural numbers. The comparison concerns Sym⁡a(Sym⁡bV)\operatorname{Sym}^{a}(\operatorname{Sym}^{b}V) and Sym⁡b(Sym⁡aV)\operatorname{Sym}^{b}(\operatorname{Sym}^{a}V) under linear changes of coordinates in VV. Write GL(V)\mathrm{GL}(V) for the group of linear automorphisms of VV.

For a natural number nn, write Finn\mathrm{Fin}\,n for the set of nonnegative integers less than nn; for a family v:Finn→Vv:\mathrm{Fin}\,n\to V, the ambient symmetric monomial pnV(v)p_n^V(v) is the product in (1), where ηV:V→SymmetricAlgebra⁡(ℂ,V)\eta_V:V\to\operatorname{SymmetricAlgebra}(\mathbb C,V) is the canonical inclusion. pnV(v)=∏i∈FinnηV(vi).p_n^V(v)=\prod_{i\in\mathrm{Fin}\,n}\eta_V(v_i).(1) The homogeneous component of degree nn in the symmetric algebra of VV is the linear span of the ambient symmetric monomials pnV(v)p_n^V(v) for families v:Finn→Vv:\mathrm{Fin}\,n\to V; its underlying vector space is the symmetric power Sym⁡nV\operatorname{Sym}^{n}V. The symmetric monomial mnV(v)m_n^V(v) is the element of the degree nn symmetric power of VV represented by the product of the vectors of the family vv in the symmetric algebra, as in (2). Sym⁡nV=span⁡ℂ{pnV(v):v:Finn→V},mnV(v)=pnV(v)∈Sym⁡nV.\operatorname{Sym}^{n}V =\operatorname{span}_{\mathbb C} \{p_n^V(v):v:\mathrm{Fin}\,n\to V\}, \qquad m_n^V(v)=p_n^V(v)\in\operatorname{Sym}^{n}V.(2)

Foulkes Maps and Equivariance

Let v:Finb→Fina→Vv:\mathrm{Fin}\,b\to\mathrm{Fin}\,a\to V be an array, and let 𝔖a\mathfrak S_a be the permutation group of Fina\mathrm{Fin}\,a. The Foulkes formula Fa,b,V(v)∈Sym⁡a(Sym⁡bV)F_{a,b,V}(v)\in\operatorname{Sym}^{a}(\operatorname{Sym}^{b}V) is the average (3) over independent permutations of the entries in each row. Fa,b,V(v)=1(a!)b∑σ∈𝔖aFinbmaSym⁡bV(k↦mbV(j↦vj,σj(k))).F_{a,b,V}(v)=\frac{1}{(a!)^b} \sum_{\sigma\in\mathfrak S_a^{\mathrm{Fin}\,b}} m_a^{\operatorname{Sym}^{b}V} \bigl(k\mapsto m_b^V(j\mapsto v_{j,\sigma_j(k)})\bigr).(3)✦ A Foulkes map μa,b,V:Sym⁡b(Sym⁡aV)→Sym⁡a(Sym⁡bV)\mu_{a,b,V}:\operatorname{Sym}^{b}(\operatorname{Sym}^{a}V) \to\operatorname{Sym}^{a}(\operatorname{Sym}^{b}V) is a complex-linear map satisfying (4) for every such array. μa,b,V(mbSym⁡aV(j↦maV(vj)))=Fa,b,V(v).\mu_{a,b,V}\bigl( m_b^{\operatorname{Sym}^{a}V}(j\mapsto m_a^V(v_j))\bigr) =F_{a,b,V}(v).(4)

For natural numbers m,nm,n and g∈GL(V)g\in\mathrm{GL}(V), a plethysm action Lm,n(g)L_{m,n}(g) is a complex-linear endomorphism of Sym⁡m(Sym⁡nV)\operatorname{Sym}^{m}(\operatorname{Sym}^{n}V) satisfying (5) for every array v:Finm→Finn→Vv:\mathrm{Fin}\,m\to\mathrm{Fin}\,n\to V. Lm,n(g)(mmSym⁡nV(i↦mnV(vi)))=mmSym⁡nV(i↦mnV(j↦g(vij))).L_{m,n}(g)\bigl( m_m^{\operatorname{Sym}^{n}V}(i\mapsto m_n^V(v_i))\bigr) =m_m^{\operatorname{Sym}^{n}V} \bigl(i\mapsto m_n^V(j\mapsto g(v_{ij}))\bigr).(5)✦ A GL(V)\mathrm{GL}(V)-equivariant embedding ι:Sym⁡a(Sym⁡bV)→Sym⁡b(Sym⁡aV)\iota:\operatorname{Sym}^{a}(\operatorname{Sym}^{b}V) \to\operatorname{Sym}^{b}(\operatorname{Sym}^{a}V) is an injective complex-linear map such that every g∈GL(V)g\in\mathrm{GL}(V) admits bijective plethysm actions Lb,a(g)L_{b,a}(g) and La,b(g)L_{a,b}(g) satisfying (6). Lb,a(g)∘ι=ι∘La,b(g).L_{b,a}(g)\circ\iota=\iota\circ L_{a,b}(g).(6)

The Claim

Theorem 1. Let VV be a finite-dimensional complex vector space, with complex dual V*=Hom⁡ℂ(V,ℂ)V^*=\operatorname{Hom}_{\mathbb C}(V,\mathbb C).

For any natural numbers a,ba,b for which a surjective Foulkes map μa,b,V*\mu_{a,b,V^*} exists, there is a GL(V)\mathrm{GL}(V)-equivariant embedding ι\iota as in (7). ι:Sym⁡a(Sym⁡bV)↪Sym⁡b(Sym⁡aV).\iota:\operatorname{Sym}^{a}(\operatorname{Sym}^{b}V) \hookrightarrow\operatorname{Sym}^{b}(\operatorname{Sym}^{a}V).(7)✦

Let a,ba,b be natural numbers with 2≤a2\le a and a(a−1)≤ba(a-1)\le b. A GL(V)\mathrm{GL}(V)-equivariant embedding ι\iota exists as in (8). ι:Sym⁡a(Sym⁡bV)↪Sym⁡b(Sym⁡aV).\iota:\operatorname{Sym}^{a}(\operatorname{Sym}^{b}V) \hookrightarrow\operatorname{Sym}^{b}(\operatorname{Sym}^{a}V).(8)✦

For every natural number b≥30b\ge30, there is a GL(V)\mathrm{GL}(V)-equivariant embedding ι\iota with source and target as in (9). ι:Sym⁡6(Sym⁡bV)↪Sym⁡b(Sym⁡6V).\iota:\operatorname{Sym}^{6}(\operatorname{Sym}^{b}V) \hookrightarrow\operatorname{Sym}^{b}(\operatorname{Sym}^{6}V).(9)✦

What Is Checked

The claims are stated over finite-dimensional complex vector spaces and natural degrees, using the symmetric powers in (2), the Foulkes map condition (4), and the equivariant embedding condition (6) with bijective plethysm actions (5).