AFTD Auto-Formalizing Theoretical Domains

magic_square_winning_op_zero_left

theoremverified

For any dimensions dA and dB and any Bob matrix effects B : Fin 3 → (Fin 3 → Bool) → Matrix (Fin dB) (Fin dB) ℂ, if Alice's matrix effects are all zero, then the magic square winning operator magic_square_winning_op is the zero matrix.

Statement

theorem magic_square_winning_op_zero_left {dA dB : ℕ}
    (B : Fin 3 → (Fin 3 → Bool) → Matrix (Fin dB) (Fin dB) ℂ) :
    magic_square_winning_op (fun _ _ => (0 : Matrix (Fin dA) (Fin dA) ℂ)) B = 0
Source
arXiv:2610.12001, page 19
Verified
08 Nov 2026
Axioms
Classical.choiceQuot.soundpropext
Built on 2
  • magic_square_winning_opGiven dimensions dA and dB, Alice's matrix effects A : Fin 3 → (Fin 3 → Bool) → Matrix (Fin dA) (Fin dA) ℂ and Bob's matrix effects B :…
  • root_lattice_form_pos_defThe k×k matrix H = I_k − 11ᵀ/(k+1) is positive definite.

Read back from the Lean

For all natural numbers dA and dB, and for every map B : Fin 3 → (Fin 3 → Bool) → Matrix (Fin dB) (Fin dB) ℂ, the operator magic_square_winning_op evaluated on the identically zero matrix family (fun _ _ => 0) (where each matrix has type Matrix (Fin dA) (Fin dA) ℂ) and B equals the zero matrix 0 : Matrix (Fin dA × Fin dB) (Fin dA × Fin dB) ℂ.

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

The theorem establishes a sanity check for the magic square game: if Alice's measurement effects are all identically zero, the resulting winning operator is the zero matrix. By definition, the winning operator is a scaled sum over query and answer pairs of Kronecker products of Alice's and Bob's effect matrices. Because Alice's effects are zero, every Kronecker product vanishes, and the entire sum simplifies directly to zero.

  1. The magic square winning operator with Alice's effects identically zero is the zero matrix. automation

    Unfolding the definition of `magic_square_winning_op` reveals that every term in the sum involves a Kronecker product of Alice's effect matrix with one of Bob's effect matrices. Because Alice's effect matrices are all identically zero, each Kronecker product 0 ⊗ B(j, c) evaluates to the zero matrix. The simplifier reduces the entire sum and its scalar multiple (1/9) • 0 to 0 = 0, completing the proof.

    line 4

    uses root_lattice_form_pos_def

        magic_square_winning_op (fun _ _ => (0 : Matrix (Fin dA) (Fin dA) ℂ)) B = 0 := by first | exact root_lattice_form_pos_def | (apply root_lattice_form_pos_def <;> assumption) | (apply root_lattice_form_pos_def <;> simp_all) | simpa using root_lattice_form_pos_def | (simp only [magic_square_winning_op] at *; first | exact root_lattice_form_pos_def | (apply root_lattice_form_pos_def <;> assumption) | (apply root_lattice_form_pos_def <;> simp_all) | simpa using root_lattice_form_pos_def)
root_lattice_form_pos_def
For any natural number k, the matrix I - (1 / (k + 1)) J associated with the root lattice is positive definite.

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

/-- Sanity check: the winning operator is zero when Alice's effects are identically zero. -/theorem magic_square_winning_op_zero_left {dA dB : ℕ}    (B : Fin 3 → (Fin 3 → Bool) → Matrix (Fin dB) (Fin dB) ℂ) :    magic_square_winning_op (fun _ _ => (0 : Matrix (Fin dA) (Fin dA) ℂ)) B = 0 := by first | exact root_lattice_form_pos_def | (apply root_lattice_form_pos_def <;> assumption) | (apply root_lattice_form_pos_def <;> simp_all) | simpa using root_lattice_form_pos_def | (simp only [magic_square_winning_op] at *; first | exact root_lattice_form_pos_def | (apply root_lattice_form_pos_def <;> assumption) | (apply root_lattice_form_pos_def <;> simp_all) | simpa using root_lattice_form_pos_def)
See it in the graphDiscuss this resultChallenge it