AFTD Auto-Formalizing Theoretical Domains

linear_flattening_potency_gt_antitone

theoremverified

If a linear flattening L on (ℂⁿ)⊗³ has potency strictly greater than r, then for any r' ≤ r it also has potency strictly greater than r'.

Statement

theorem linear_flattening_potency_gt_antitone {n u v : ℕ}
    (L : (Fin n → Fin n → Fin n → ℂ) →ₗ[ℂ] Matrix (Fin u) (Fin v) ℂ)
    {r r' : ℕ} (hr : r' ≤ r) (h : linear_flattening_potency_gt L r) :
    linear_flattening_potency_gt L r'
Source
arXiv:2610.08504 Sec. 1
Verified
20 Oct 2026
Axioms
Classical.choiceQuot.soundpropext
Built on 2
  • linear_flattening_potency_gtA linear flattening L : ℂⁿ ⊗ ℂⁿ ⊗ ℂⁿ → Mat_{u×v}(ℂ) has potency greater than r: some tensor T has rank L(T) > r · ρ(L), where ρ(L) =…
  • tensor3_pureThe pure tensor x ⊗ y ⊗ z in ℂⁿ ⊗ ℂⁿ ⊗ ℂⁿ, with entries x_i y_j z_k.
Used by 1
  • linear_flattening_testable_antitoneIf the secant variety σ_r((ℂⁿ)⊗³) is testable by the linear flattening method, then so is σ_{r'}((ℂⁿ)⊗³) for every r' ≤ r.

Read back from the Lean

For any natural numbers $n, u, v$, any $\mathbb{C}$-linear map $L$ from 3-tensors $(\mathrm{Fin}\ n \to \mathrm{Fin}\ n \to \mathrm{Fin}\ n \to \mathbb{C})$ to matrices $\mathrm{Matrix}\ (\mathrm{Fin}\ u)\ (\mathrm{Fin}\ v)\ \mathbb{C}$, and any natural numbers $r, r'$: if $r' \le r$ and linear_flattening_potency_gt L r holds, then linear_flattening_potency_gt L r' holds.

Written by a model that saw only the Lean, never the English above. If the two disagree, that is worth a challenge.

The proof, explained

By definition, a linear flattening L has potency strictly greater than r if there exists a tensor T whose flattened rank exceeds r times the flattened rank of any pure tensor. The proof takes this same tensor T as the witness for r'. Since r' ≤ r, multiplying by the rank of any pure tensor yields r' * rank(L(x ⊗ y ⊗ z)) ≤ r * rank(L(x ⊗ y ⊗ z)) < rank(L(T)), establishing that L has potency strictly greater than r'.

  1. To show L has potency strictly greater than r', it suffices to verify that the witness tensor T for r also works for r'.

    By definition, the hypothesis that L has potency strictly greater than r provides a tensor T such that r * rank(L(x ⊗ y ⊗ z)) < rank(L(T)) for all vectors x, y, z. We supply this same tensor T as the witness for r', so it remains to check that r' * rank(L(x ⊗ y ⊗ z)) < rank(L(T)) for arbitrary x, y, z.

    lines 6–7

      rcases h with ⟨T, hT⟩
      refine ⟨T, fun x y z => ?_⟩
  2. r' * rank(L(x ⊗ y ⊗ z)) ≤ r * rank(L(x ⊗ y ⊗ z)) and r * rank(L(x ⊗ y ⊗ z)) < rank(L(T)).

    Specializing the property of T to x, y, z gives the strict inequality r * rank(L(x ⊗ y ⊗ z)) < rank(L(T)). Multiplying the assumption r' ≤ r on the right by rank(L(x ⊗ y ⊗ z)) yields the weak inequality r' * rank(L(x ⊗ y ⊗ z)) ≤ r * rank(L(x ⊗ y ⊗ z)).

    lines 8–10

    uses Nat.mul_le_mul_right

      have h1 := hT x y z
      have h2 : r' * (L (tensor3_pure x y z)).rank ≤ r * (L (tensor3_pure x y z)).rank :=
        Nat.mul_le_mul_right _ hr
  3. r' * rank(L(x ⊗ y ⊗ z)) < rank(L(T)).

    Chaining the weak inequality and the strict inequality via transitivity (lt_of_le_of_lt) proves the goal.

    line 11

    uses lt_of_le_of_lt

      exact lt_of_le_of_lt h2 h1
Nat.mul_le_mul_right
Multiplying both sides of an inequality of natural numbers on the right by a constant preserves the inequality: if n ≤ m, then n * k ≤ m * k.
lt_of_le_of_lt
Transitivity of order: if a ≤ b and b < c, then a < c.

Written by a model (gemini-3.8-flash) from the Lean proof and the goal states Lean reported at each step. Lean verified the proof; this explanation is a reading aid and is not itself machine-checked.

Lean source view module on GitHub

/-- If a linear flattening has potency greater than r, it also has potency greater than any r' ≤ r. -/theorem linear_flattening_potency_gt_antitone {n u v : ℕ}    (L : (Fin n → Fin n → Fin n → ℂ) →ₗ[ℂ] Matrix (Fin u) (Fin v) ℂ)    {r r' : ℕ} (hr : r' ≤ r) (h : linear_flattening_potency_gt L r) :    linear_flattening_potency_gt L r' := by  rcases h with ⟨T, hT⟩  refine ⟨T, fun x y z => ?_⟩  have h1 := hT x y z  have h2 : r' * (L (tensor3_pure x y z)).rank ≤ r * (L (tensor3_pure x y z)).rank :=    Nat.mul_le_mul_right _ hr  exact lt_of_le_of_lt h2 h1
See it in the graphDiscuss this resultChallenge it