AFTD Auto-Formalizing Theoretical Domains

ghz_network_lambda_at_pi_div_four_le

theoremverifiedformalization

For the maximally entangled state parameter alpha = pi/4, all orthonormal frames e, f in O(3), and all setting indices mu, nu, mu', nu' in {0, 1, 2, 3}, the two-copy closed-network quantity Lambda(tau_{mu nu}, tau_{mu' nu'}) is at most 1 + sin^2(2 * (pi / 4)).

Statement

theorem ghz_network_lambda_at_pi_div_four_le (e f : Matrix (Fin 3) (Fin 3) ℝ) (he : e ∈ Matrix.orthogonalGroup (Fin 3) ℝ) (hf : f ∈ Matrix.orthogonalGroup (Fin 3) ℝ) (μ ν μ' ν' : Fin 4) :
    bell_network_lambda (ghz_correlation_vector (Real.pi / 4) e f μ ν) (ghz_correlation_vector (Real.pi / 4) e f μ' ν') ≤ 1 + Real.sin (2 * (Real.pi / 4)) ^ 2
Formalizes
arXiv:2610.35838 Eq. 14
Verified
02 Nov 2026
Axioms
Classical.choiceQuot.soundpropext
Built on 3
  • bell_network_lambdaFor vectors u, v in R^3, Lambda(u, v) is defined as (1/2) * (s + sqrt(s^2 - 4 * (u . v)^2)) where s = |u|^2 + |v|^2.
  • ghz_correlation_vectorFor parameter alpha in R, orthonormal frames e, f of R^3, and indices mu, nu in {0, 1, 2, 3}, the correlation vector tau_{mu nu} has…
  • ghz_pauli_coeffFor an orthonormal frame e of R^3 and index mu in {0, 1, 2, 3}, the Pauli coefficients of A_mu are given by A_0 = I (coefficient 1 at…

Read back from the Lean

For all 3 × 3 real matrices $e$ and $f$, if $e$ and $f$ both belong to the orthogonal group $O(3, \mathbb{R})$, then for all indices $\mu, \nu, \mu', \nu' \in \{0, 1, 2, 3\}$ (i.e. Fin 4), the function bell_network_lambda evaluated on the two vectors $u = \text{ghz\_correlation\_vector}(\pi / 4, e, f, \mu, \nu)$ and $v = \text{ghz\_correlation\_vector}(\pi / 4, e, f, \mu', \nu')$ is less than or equal to $1 + \sin^2(2 \cdot (\pi / 4))$.

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

To prove that the Bell network quantity Lambda(u, v) is at most 1 + sin^2(2 * (pi / 4)) = 2 for maximally entangled states (alpha = pi / 4), we first establish that every GHZ correlation vector has squared norm at most 1. This follows because each row of an orthogonal matrix in O(3) is a unit vector, making the 4-component Pauli coefficient vectors unit vectors, which bounds the correlation vector's squared norm by 1. Finally, with s = ||u||^2 + ||v||^2 at most 2, the square root term sqrt(s^2 - 4(u · v)^2) is bounded above by s, showing that Lambda(u, v) = (1/2)(s + sqrt(s^2 - 4(u · v)^2)) is at most s <= 2.

  1. Each row of an orthogonal 3×3 matrix e has squared norm equal to 1: e i 0 ^ 2 + e i 1 ^ 2 + e i 2 ^ 2 = 1.

    By definition of the orthogonal group, e * eᵀ = 1. Specializing this matrix equality to the diagonal entry (i, i), expanding the matrix product into a sum over the three coordinates, and applying the definition of the transpose yields e i 0 ^ 2 + e i 1 ^ 2 + e i 2 ^ 2 = 1.

    lines 6–11

    uses Matrix.mem_orthogonalGroup_iff, congr_fun, Matrix.mul_apply, Matrix.one_apply_eq, Fin.sum_univ_three, Matrix.transpose_apply

      have h : e * eᵀ = 1 := (Matrix.mem_orthogonalGroup_iff (Fin 3) ℝ).mp he
      have hii := congr_fun (congr_fun h i) i
      rw [Matrix.mul_apply, Matrix.one_apply_eq] at hii
      rw [Fin.sum_univ_three] at hii
      simp only [Matrix.transpose_apply] at hii
      nlinarith [hii]
  2. For each setting index μ ∈ {0, 1, 2, 3}, the Pauli coefficient vector of e has squared norm 1.

    We analyze the four cases for μ. For μ = 0, the coefficients are (1, 0, 0, 0), which have squared norm 1. For μ ∈ {1, 2, 3}, the coefficients are (0, e (μ-1) 0, e (μ-1) 1, e (μ-1) 2), whose sum of squares equals the row norm of row μ - 1 of e, which is 1 by the previous row-norm lemma.

    lines 19–29

    uses ghz_network_lambda_at_pi_div_four_le_row_norm

      fin_cases μ
      · simp [ghz_pauli_coeff]
      · have h := ghz_network_lambda_at_pi_div_four_le_row_norm e he 0
        simp (config := {decide := true}) [ghz_pauli_coeff]
        linarith [h]
      · have h := ghz_network_lambda_at_pi_div_four_le_row_norm e he 1
        simp (config := {decide := true}) [ghz_pauli_coeff]
        linarith [h]
      · have h := ghz_network_lambda_at_pi_div_four_le_row_norm e he 2
        simp (config := {decide := true}) [ghz_pauli_coeff]
        linarith [h]
  3. At alpha = pi / 4, every GHZ correlation vector has squared norm at most 1. automation

    At α = π / 4, we have sin(2α) = sin(π / 2) = 1 and cos(2α) = cos(π / 2) = 0. Expanding the dot product of the correlation vector with itself into its three components and substituting these values, the sum of squares is bounded by 1 using the unit-norm property of the Pauli coefficient vectors and the non-negativity of squares.

    lines 37–50

    uses Real.sin_pi_div_two, Real.cos_pi_div_two, ghz_network_lambda_at_pi_div_four_le_coeff_unit, Fin.sum_univ_three, sq_nonneg

      have hsin : Real.sin (2 * (Real.pi / 4)) = 1 := by
        rw [show 2 * (Real.pi / 4) = Real.pi / 2 by ring, Real.sin_pi_div_two]
      have hcos : Real.cos (2 * (Real.pi / 4)) = 0 := by
        rw [show 2 * (Real.pi / 4) = Real.pi / 2 by ring, Real.cos_pi_div_two]
      have ha := ghz_network_lambda_at_pi_div_four_le_coeff_unit e he μ
      have hc := ghz_network_lambda_at_pi_div_four_le_coeff_unit f hf ν
      rw [dotProduct, Fin.sum_univ_three]
      simp only [ghz_correlation_vector, hsin, hcos, one_mul, zero_mul, add_zero, zero_add]
      simp only [show ((1 : Fin 3)) ≠ 0 by decide, show ((2 : Fin 3)) ≠ 0 by decide,
        show ((2 : Fin 3)) ≠ 1 by decide, if_false, if_true]
      nlinarith [sq_nonneg (ghz_pauli_coeff e μ 0 * ghz_pauli_coeff f ν 0
          - ghz_pauli_coeff e μ 3 * ghz_pauli_coeff f ν 3),
        sq_nonneg (ghz_pauli_coeff e μ 1), sq_nonneg (ghz_pauli_coeff e μ 2),
        sq_nonneg (ghz_pauli_coeff f ν 1), sq_nonneg (ghz_pauli_coeff f ν 2)]
  4. The target bound 1 + sin²(2 * (pi / 4)) simplifies to 2, and the correlation vectors u and v have squared norm at most 1.

    Since sin(2 * (π / 4)) = sin(π / 2) = 1, the right-hand side 1 + sin²(2α) evaluates to 1 + 1² = 2. Applying the norm bound lemma to the setting pairs (μ, ν) and (μ', ν') establishes that both correlation vectors u and v have squared norm at most 1.

    lines 55–60

    uses Real.sin_pi_div_two, ghz_network_lambda_at_pi_div_four_le_norm_sq

      have hsin : Real.sin (2 * (Real.pi / 4)) = 1 := by
        rw [show 2 * (Real.pi / 4) = Real.pi / 2 by ring, Real.sin_pi_div_two]
      have hu := ghz_network_lambda_at_pi_div_four_le_norm_sq e f he hf μ ν
      have hv := ghz_network_lambda_at_pi_div_four_le_norm_sq e f he hf μ' ν'
      have hval : 1 + Real.sin (2 * (Real.pi / 4)) ^ 2 = 2 := by rw [hsin]; norm_num
      rw [hval]
  5. Unfolding the Bell network quantity, the sum of the squared norms s = (u ⬝ᵥ u) + (v ⬝ᵥ v) satisfies 0 ≤ s ≤ 2.

    We unfold bell_network_lambda and define s as the sum of the squared norms of u and v. Since each squared norm is a sum of squares of real components, both u ⬝ᵥ u ≥ 0 and v ⬝ᵥ v ≥ 0, which gives s ≥ 0. Moreover, because each squared norm is at most 1, their sum satisfies s ≤ 2.

    lines 61–69

    uses Finset.sum_nonneg, mul_self_nonneg

      unfold bell_network_lambda
      set s := ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ ν
        + ghz_correlation_vector (Real.pi / 4) e f μ' ν' ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ' ν' with hs
      have hunn : 0 ≤ ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ ν := by
        rw [dotProduct]; exact Finset.sum_nonneg (fun i _ => mul_self_nonneg _)
      have hvnn : 0 ≤ ghz_correlation_vector (Real.pi / 4) e f μ' ν' ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ' ν' := by
        rw [dotProduct]; exact Finset.sum_nonneg (fun i _ => mul_self_nonneg _)
      have hsnn : 0 ≤ s := by rw [hs]; linarith
      have hle : s ≤ 2 := by rw [hs]; linarith
  6. The radical √(s² - 4(u ⬝ᵥ v)²) is at most s, which implies bell_network_lambda u v ≤ 2.

    Because 4(u ⬝ᵥ v)² ≥ 0, we have s² - 4(u ⬝ᵥ v)² ≤ s². Monotonicity of the square root implies √(s² - 4(u ⬝ᵥ v)²) ≤ √(s²), which simplifies to s since s ≥ 0. Hence bell_network_lambda u v = (1 / 2) * (s + √(s² - 4(u ⬝ᵥ v)²)) ≤ (1 / 2) * (s + s) = s ≤ 2.

    lines 70–76

    uses sq_nonneg, Real.sqrt_le_sqrt, Real.sqrt_sq

      have hsq : Real.sqrt (s ^ 2 - 4 * (ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ' ν') ^ 2) ≤ s := by
        have h1 : s ^ 2 - 4 * (ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ' ν') ^ 2 ≤ s ^ 2 := by
          nlinarith [sq_nonneg (ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ' ν')]
        have h2 : Real.sqrt (s ^ 2 - 4 * (ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ' ν') ^ 2) ≤ Real.sqrt (s ^ 2) :=
          Real.sqrt_le_sqrt h1
        rwa [Real.sqrt_sq hsnn] at h2
      linarith
Matrix.mem_orthogonalGroup_iff
A square matrix A belongs to the orthogonal group if and only if A * Aᵀ = 1.
congr_fun
Equal functions take equal values at any argument.
Matrix.mul_apply
The (i, k) entry of the matrix product M * N is the sum over j of M i j * N j k.
Matrix.one_apply_eq
The diagonal entries of the identity matrix are equal to 1.
Fin.sum_univ_three
A sum over the three-element type Fin 3 equals the sum of the terms at indices 0, 1, and 2.
Matrix.transpose_apply
The (i, j) entry of the transpose matrix Mᵀ is M j i.
ghz_network_lambda_at_pi_div_four_le_row_norm
Each row of an orthogonal 3x3 real matrix has squared Euclidean norm equal to 1.
Real.sin_pi_div_two
The sine of pi / 2 equals 1.
Real.cos_pi_div_two
The cosine of pi / 2 equals 0.
ghz_network_lambda_at_pi_div_four_le_coeff_unit
For any orthogonal frame, the 4-component Pauli coefficient vector at any setting index is a unit vector.
sq_nonneg
The square of any real number is non-negative.
ghz_network_lambda_at_pi_div_four_le_norm_sq
At alpha = pi / 4, the GHZ correlation vector has squared Euclidean norm at most 1.
Finset.sum_nonneg
A finite sum of non-negative real numbers is non-negative.
mul_self_nonneg
The product of any real number with itself is non-negative.
Real.sqrt_le_sqrt
The square root function on real numbers preserves non-strict inequality: x ≤ y implies √x ≤ √y.
Real.sqrt_sq
For any non-negative real number x, √(x²) = x.

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

open Matrix in/-- Each row of an orthogonal 3x3 matrix is a unit vector. -/private lemma ghz_network_lambda_at_pi_div_four_le_row_norm (e : Matrix (Fin 3) (Fin 3) ℝ)    (he : e ∈ Matrix.orthogonalGroup (Fin 3) ℝ) (i : Fin 3) :    e i 0 ^ 2 + e i 1 ^ 2 + e i 2 ^ 2 = 1 := by  have h : e * eᵀ = 1 := (Matrix.mem_orthogonalGroup_iff (Fin 3) ℝ).mp he  have hii := congr_fun (congr_fun h i) i  rw [Matrix.mul_apply, Matrix.one_apply_eq] at hii  rw [Fin.sum_univ_three] at hii  simp only [Matrix.transpose_apply] at hii  nlinarith [hii]open Matrix in/-- The Pauli coefficient vector of an orthogonal frame is a unit vector. -/private lemma ghz_network_lambda_at_pi_div_four_le_coeff_unit (e : Matrix (Fin 3) (Fin 3) ℝ)    (he : e ∈ Matrix.orthogonalGroup (Fin 3) ℝ) (μ : Fin 4) :    (ghz_pauli_coeff e μ 0) ^ 2 + (ghz_pauli_coeff e μ 1) ^ 2 + (ghz_pauli_coeff e μ 2) ^ 2      + (ghz_pauli_coeff e μ 3) ^ 2 = 1 := by  fin_cases μ  · simp [ghz_pauli_coeff]  · have h := ghz_network_lambda_at_pi_div_four_le_row_norm e he 0    simp (config := {decide := true}) [ghz_pauli_coeff]    linarith [h]  · have h := ghz_network_lambda_at_pi_div_four_le_row_norm e he 1    simp (config := {decide := true}) [ghz_pauli_coeff]    linarith [h]  · have h := ghz_network_lambda_at_pi_div_four_le_row_norm e he 2    simp (config := {decide := true}) [ghz_pauli_coeff]    linarith [h]open Matrix in/-- At `alpha = pi/4` every GHZ correlation vector has squared norm at most one. -/private lemma ghz_network_lambda_at_pi_div_four_le_norm_sq (e f : Matrix (Fin 3) (Fin 3) ℝ)    (he : e ∈ Matrix.orthogonalGroup (Fin 3) ℝ) (hf : f ∈ Matrix.orthogonalGroup (Fin 3) ℝ)    (μ ν : Fin 4) :    ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ ν ≤ 1 := by  have hsin : Real.sin (2 * (Real.pi / 4)) = 1 := by    rw [show 2 * (Real.pi / 4) = Real.pi / 2 by ring, Real.sin_pi_div_two]  have hcos : Real.cos (2 * (Real.pi / 4)) = 0 := by    rw [show 2 * (Real.pi / 4) = Real.pi / 2 by ring, Real.cos_pi_div_two]  have ha := ghz_network_lambda_at_pi_div_four_le_coeff_unit e he μ  have hc := ghz_network_lambda_at_pi_div_four_le_coeff_unit f hf ν  rw [dotProduct, Fin.sum_univ_three]  simp only [ghz_correlation_vector, hsin, hcos, one_mul, zero_mul, add_zero, zero_add]  simp only [show ((1 : Fin 3)) ≠ 0 by decide, show ((2 : Fin 3)) ≠ 0 by decide,    show ((2 : Fin 3)) ≠ 1 by decide, if_false, if_true]  nlinarith [sq_nonneg (ghz_pauli_coeff e μ 0 * ghz_pauli_coeff f ν 0      - ghz_pauli_coeff e μ 3 * ghz_pauli_coeff f ν 3),    sq_nonneg (ghz_pauli_coeff e μ 1), sq_nonneg (ghz_pauli_coeff e μ 2),    sq_nonneg (ghz_pauli_coeff f ν 1), sq_nonneg (ghz_pauli_coeff f ν 2)]/-- For the maximally entangled state alpha = pi/4, Lambda is at most 1 + sin^2(2*(pi/4)) = 2 for all orthonormal frames and setting quadruples. -/theorem ghz_network_lambda_at_pi_div_four_le (e f : Matrix (Fin 3) (Fin 3) ℝ) (he : e ∈ Matrix.orthogonalGroup (Fin 3) ℝ) (hf : f ∈ Matrix.orthogonalGroup (Fin 3) ℝ) (μ ν μ' ν' : Fin 4) :    bell_network_lambda (ghz_correlation_vector (Real.pi / 4) e f μ ν) (ghz_correlation_vector (Real.pi / 4) e f μ' ν') ≤ 1 + Real.sin (2 * (Real.pi / 4)) ^ 2 := by  have hsin : Real.sin (2 * (Real.pi / 4)) = 1 := by    rw [show 2 * (Real.pi / 4) = Real.pi / 2 by ring, Real.sin_pi_div_two]  have hu := ghz_network_lambda_at_pi_div_four_le_norm_sq e f he hf μ ν  have hv := ghz_network_lambda_at_pi_div_four_le_norm_sq e f he hf μ' ν'  have hval : 1 + Real.sin (2 * (Real.pi / 4)) ^ 2 = 2 := by rw [hsin]; norm_num  rw [hval]  unfold bell_network_lambda  set s := ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ ν    + ghz_correlation_vector (Real.pi / 4) e f μ' ν' ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ' ν' with hs  have hunn : 0 ≤ ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ ν := by    rw [dotProduct]; exact Finset.sum_nonneg (fun i _ => mul_self_nonneg _)  have hvnn : 0 ≤ ghz_correlation_vector (Real.pi / 4) e f μ' ν' ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ' ν' := by    rw [dotProduct]; exact Finset.sum_nonneg (fun i _ => mul_self_nonneg _)  have hsnn : 0 ≤ s := by rw [hs]; linarith  have hle : s ≤ 2 := by rw [hs]; linarith  have hsq : Real.sqrt (s ^ 2 - 4 * (ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ' ν') ^ 2) ≤ s := by    have h1 : s ^ 2 - 4 * (ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ' ν') ^ 2 ≤ s ^ 2 := by      nlinarith [sq_nonneg (ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ' ν')]    have h2 : Real.sqrt (s ^ 2 - 4 * (ghz_correlation_vector (Real.pi / 4) e f μ ν ⬝ᵥ ghz_correlation_vector (Real.pi / 4) e f μ' ν') ^ 2) ≤ Real.sqrt (s ^ 2) :=      Real.sqrt_le_sqrt h1    rwa [Real.sqrt_sq hsnn] at h2  linarith
See it in the graphDiscuss this resultChallenge it