ppt_qutrit_steered_effect_is_valid
Each of the three normalized steered operators and its complement is positive semidefinite.
Statement
theorem ppt_qutrit_steered_effect_is_valid : ∀ j : Fin 3, (ppt_qutrit_steered_effect j).PosSemidef ∧ (1 - ppt_qutrit_steered_effect j).PosSemidef- Source
- Helper for OP-693; construction from arXiv:1405.4502, and steering normalization in arXiv:2010.04048, Appendix A.
- Verified
- 12 Nov 2026
- Axioms
Classical.choiceQuot.soundpropext- Built on 6
- ppt_qutrit_steered_effectThe normalized steered effects obtained from the explicit PPT Bell example.
- bipartite_conditional_operatorThe (unnormalized) conditional operator on the second system β of a bipartite matrix ρ produced by the operator M on the first system:…
- ppt_qutrit_bell_settingThe three unit vectors defining the binary projective settings in the Bell example.
- ppt_qutrit_bell_stateThe explicit two-qutrit positive-partial-transpose state in the Bell-violation example.
- ppt_qutrit_bell_vectorFour explicit real vectors in two-qutrit space used in the PPT Bell-violation example.
- ppt_qutrit_steered_normalizerThe diagonal inverse square root of the second-qutrit reduced state.
- Used by 2
- ppt_qutrit_steered_effect_all_proper_subspaces_compatibleEvery proper subspace compression of the three normalized steered effects has a joint parent.
- open_problem_693_three_dim_fewer_than_nineThere exists a natural number k strictly less than 9 such that has_subspace_compatible_incompatible_family 3 k holds, meaning there is a…
The proof, explained
To show that each normalized steered effect E_j = N S(A_j) N and its complement 1 - E_j are positive semidefinite, the proof first establishes that Alice's effect A_j is an orthogonal projector onto a unit vector, so both A_j and 1 - A_j are positive semidefinite. Because the shared Bell state ρ is a conic mixture of pure states with non-negative weights, the linear steering map S preserves positive semidefiniteness. Conjugating the steered operators S(A_j) and S(1 - A_j) by the Hermitian normalizer N therefore yields positive semidefinite operators for both E_j and N S(1 - A_j) N. Finally, the normalization condition N M N = 1 for Bob's marginal state M = S(1) equates N S(1 - A_j) N with 1 - E_j.
For each setting index j ∈ {0, 1, 2}, Alice's measurement vector v_j is a unit vector (v_jᴴ ⬝ᵥ v_j = 1).
The proof records the radical squares (√3)^2 = 3 and (√21)^2 = 21, and exhausts all three values of j ∈ {0, 1, 2}. For each case, it expands the definition of Alice's setting vector into a sum of three terms and simplifies the rational arithmetic to 1.
uses
Real.sq_sqrt,Fin.sum_univ_three,map_ofNathave h3 : (Real.sqrt 3 : ℂ)^2 = 3 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ) ≤ 3 by norm_num) have h21 : (Real.sqrt 21 : ℂ)^2 = 21 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ) ≤ 21 by norm_num) fin_cases j <;> simp [ppt_qutrit_bell_setting, dotProduct, Fin.sum_univ_three] <;> ring_nf <;> simp [h3,h21, map_ofNat] <;> norm_num
For each setting index j, Alice's effect matrix A_j = v_j v_jᴴ and its complement 1 - A_j are positive semidefinite.
The outer product A_j = v_j v_jᴴ is positive semidefinite by Matrix.posSemidef_vecMulVec_self_star. Since v_j is a unit vector, A_j is an idempotent projector (A_j^2 = A_j). The complement 1 - A_j is Hermitian and also idempotent, (1 - A_j)^2 = 1 - 2 A_j + A_j^2 = 1 - A_j, which implies it can be written as (1 - A_j)ᴴ (1 - A_j) and is therefore positive semidefinite.
uses
Matrix.posSemidef_vecMulVec_self_star,Matrix.vecMulVec_mul_vecMulVec,in767_effectvalidproof_alice_setting_unit,one_smul,Matrix.posSemidef_conjTranspose_mul_selflet v := ppt_qutrit_bell_setting j let A : Matrix (Fin 3) (Fin 3) ℂ := vecMulVec v (star v) have hA : A.PosSemidef := Matrix.posSemidef_vecMulVec_self_star v have hAsq : A*A=A := by dsimp [A] rw [Matrix.vecMulVec_mul_vecMulVec, in767_effectvalidproof_alice_setting_unit j, one_smul] have hQherm : (1-A).IsHermitian := Matrix.isHermitian_one.sub hA.isHermitian have hQsq : (1-A)*(1-A)=1-A := by calc (1-A)*(1-A) = 1-A-A+A*A := by noncomm_ring _ = 1-A := by rw [hAsq]; abel have hQ := Matrix.posSemidef_conjTranspose_mul_self (1-A) rw [hQherm.eq, hQsq] at hQ exact ⟨hA,hQ⟩
Bob's unnormalized marginal matrix M(x, y) = ∑_i ρ_{(i,x), (i,y)} is diagonal with diagonal entries 3437/6884, 3437/6884, and 5/3442. automation
The proof evaluates the matrix entrywise over all nine index pairs (x, y) ∈ {0, 1, 2}². For each pair, it expands the definitions of the shared state and the four pure state vectors, clears denominators using the non-zeroness of √2 and √3, substitutes the radical squares (√2)^2 = 2, (√3)^2 = 3, and (√131)^2 = 131, and simplifies the rational arithmetic.
uses
Real.sq_sqrt,ne_of_gt,Matrix.sum_apply,Matrix.smul_apply,Matrix.vecMulVec_apply,Fin.sum_univ_three,Fin.sum_univ_four,Matrix.diagonal_apply,Complex.real_smul,map_ofNathave h2 : (Real.sqrt 2 : ℂ)^2=2 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤2 by norm_num) have h3 : (Real.sqrt 3 : ℂ)^2=3 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤3 by norm_num) have h131 : (Real.sqrt (131 : ℝ) : ℂ)^2=131 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤131 by norm_num) have hn2 : (Real.sqrt 2 : ℂ) ≠ 0 := by exact_mod_cast (ne_of_gt (Real.sqrt_pos.2 (by norm_num : (0:ℝ)<2))) have hn3 : (Real.sqrt 3 : ℂ) ≠ 0 := by exact_mod_cast (ne_of_gt (Real.sqrt_pos.2 (by norm_num : (0:ℝ)<3))) ext x y fin_cases x <;> fin_cases y <;> simp [ppt_qutrit_bell_state, ppt_qutrit_bell_vector, Matrix.sum_apply, Matrix.smul_apply, Matrix.vecMulVec_apply, Fin.sum_univ_three, Fin.sum_univ_four, Matrix.diagonal_apply, Complex.real_smul, map_ofNat] <;> field_simp [hn2,hn3] <;> ring_nf <;> simp [h2,h3,h131] <;> norm_num
Bob's normalizer matrix N satisfies N * M * N = 1, where M is Bob's marginal matrix. automation
Rewriting M with its diagonal value from the previous step reduces N * M * N to a diagonal matrix. The proof verifies by linear combination that the diagonal squares satisfy (√6884/√3437)^2 * (3437/6884) = 1 and (√3442/√5)^2 * (5/3442) = 1, and then checks entrywise across all index pairs that the product equals the identity matrix.
uses
in767_effectvalidproof_source_second_marginal,Real.sq_sqrt,ne_of_gt,Fin.sum_univ_three,mul_comm,mul_left_comm,mul_assocdsimp only rw [in767_effectvalidproof_source_second_marginal] have h6884 : (Real.sqrt 6884 : ℂ)^2=6884 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤6884 by norm_num) have h3437 : (Real.sqrt 3437 : ℂ)^2=3437 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤3437 by norm_num) have h3442 : (Real.sqrt 3442 : ℂ)^2=3442 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤3442 by norm_num) have h5 : (Real.sqrt 5 : ℂ)^2=5 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤5 by norm_num) have hn3437 : (Real.sqrt 3437 : ℂ) ≠ 0 := by exact_mod_cast (ne_of_gt (Real.sqrt_pos.2 (by norm_num : (0:ℝ)<3437))) have hn5 : (Real.sqrt 5 : ℂ) ≠ 0 := by exact_mod_cast (ne_of_gt (Real.sqrt_pos.2 (by norm_num : (0:ℝ)<5))) have hs0 : (Real.sqrt 6884 : ℂ) / (Real.sqrt 3437 : ℂ) * (3437/6884) * ((Real.sqrt 6884 : ℂ) / (Real.sqrt 3437 : ℂ)) = 1 := by field_simp [hn3437] linear_combination 3437*h6884-6884*h3437 have hs2 : (Real.sqrt 3442 : ℂ) / (Real.sqrt 5 : ℂ) * (5/3442) * ((Real.sqrt 3442 : ℂ) / (Real.sqrt 5 : ℂ)) = 1 := by field_simp [hn5] linear_combination 5*h3442-3442*h5 ext i j fin_cases i <;> fin_cases j <;> simp only [ppt_qutrit_steered_normalizer, Matrix.mul_apply, Fin.sum_univ_three, Matrix.diagonal_apply, Matrix.cons_val_zero, Matrix.cons_val_one, Matrix.cons_val_two] <;> norm_num <;> first | exact hs0 | simpa [mul_comm,mul_left_comm,mul_assoc] using hs0 | exact hs2
Bob's normalizer matrix N is Hermitian: Nᴴ = N.
Entrywise, if i = j, the diagonal entries of N are real square roots, so their complex conjugate equals themselves. If i ≠ j, the off-diagonal entries are 0, so both N_{i,j} and N_{j,i} vanish.
uses
Matrix.conjTranspose_apply,Matrix.diagonal_apply,Ne.symmext i j by_cases h : i = j · subst j by_cases hi : i = 2 <;> simp [ppt_qutrit_steered_normalizer, Matrix.conjTranspose_apply, Matrix.diagonal_apply, hi] · simp [ppt_qutrit_steered_normalizer, Matrix.conjTranspose_apply, Matrix.diagonal_apply, h, Ne.symm h]
Steering any four-vector conic mixture by a positive semidefinite matrix produces a positive semidefinite operator.
For each pure vector v_t, define the operator (B_t)_{x, i} = v_t(i, x). Since A is positive semidefinite, its transpose A^T is positive semidefinite, hence B_t A^T B_tᴴ is positive semidefinite. By finite set induction, the conic combination ∑_t λ_t (B_t A^T B_tᴴ) is positive semidefinite because each weight λ_t is non-negative. Expanding entrywise shows this conic sum equals the steered operator formula.
uses
Matrix.posSemidef_transpose_iff.mpr,Finset.induction_on,Matrix.PosSemidef.zero,Finset.sum_insert,Finset.univ,Matrix.sum_apply,Matrix.smul_apply,Complex.real_smul,Fin.sum_univ_three,Fin.sum_univ_fourlet B : Fin 4 → Matrix (Fin 3) (Fin 3) ℂ := fun t x i => v t (i,x) have hterm (t : Fin 4) : (B t * A.transpose * (B t)ᴴ).PosSemidef := (Matrix.posSemidef_transpose_iff.mpr hA).mul_mul_conjTranspose_same (B t) have hsum : (∑ t, lam t • (B t * A.transpose * (B t)ᴴ)).PosSemidef := by have H (T : Finset (Fin 4)) : (∑ t ∈ T, lam t • (B t * A.transpose * (B t)ᴴ)).PosSemidef := by induction T using Finset.induction_on with | empty => simpa using (Matrix.PosSemidef.zero : (0 : Matrix (Fin 3) (Fin 3) ℂ).PosSemidef) | @insert t T ht ih => rw [Finset.sum_insert ht] exact ((hterm t).smul (hlam t)).add ih exact H Finset.univ have heq : (fun x y : Fin 3 => ∑ i : Fin 3, ∑ j : Fin 3, (∑ t : Fin 4, (lam t : ℂ) * v t (i,x) * star (v t (j,y))) * A j i) = ∑ t, lam t • (B t * A.transpose * (B t)ᴴ) := by ext x y simp [B, Matrix.sum_apply, Matrix.smul_apply, Complex.real_smul, Matrix.mul_apply, Matrix.transpose_apply, Matrix.conjTranspose_apply, Fin.sum_univ_three, Fin.sum_univ_four] ring rw [heq] exact hsum
For each setting j, the normalized steered effect E_j = N * S(A_j) * N and its complement 1 - E_j are positive semidefinite.
The shared Bell state is expressed as a conic combination of four pure vectors with non-negative weights λ_t ≥ 0. By the general steering positivity lemma, the linear steering map S preserves positive semidefiniteness. Since Alice's effect A_j and its complement 1 - A_j are positive semidefinite, S(A_j) and S(1 - A_j) are positive semidefinite. Conjugation by the Hermitian normalizer N preserves positivity, yielding (N * S(A_j) * N).PosSemidef and (N * S(1 - A_j) * N).PosSemidef. Finally, linearity of S together with the normalization N * M * N = 1 yields N * S(1 - A_j) * N = 1 - N * S(A_j) * N.
uses
in767_effectvalidproof_alice_settings_valid,in767_effectvalidproof_positive_steering_from_four_vectors,in767_effectvalidproof_normalizer_hermitian,Matrix.mul_sub,Matrix.sub_mul,in767_effectvalidproof_source_normalizer_marginallet lam : Fin 4 → ℝ := fun t => if t = 0 then 3257/6884 else if t = 3 then 27/6884 else 450/1721 let ρ : Matrix (Fin 3 × Fin 3) (Fin 3 × Fin 3) ℂ := fun ax cy => ∑ t : Fin 4, (lam t : ℂ) * ppt_qutrit_bell_vector t ax * star (ppt_qutrit_bell_vector t cy) have hlam : ∀ t, 0 ≤ lam t := by intro t; dsimp [lam]; split_ifs <;> norm_num have hρ : ρ = ppt_qutrit_bell_state := by ext ax cy simp [ρ, ppt_qutrit_bell_state, lam, Fin.sum_univ_four, Matrix.sum_apply, Matrix.smul_apply, Matrix.vecMulVec_apply, Pi.star_apply, Complex.real_smul, mul_assoc] let S (B : Matrix (Fin 3) (Fin 3) ℂ) : Matrix (Fin 3) (Fin 3) ℂ := fun x y => ∑ i : Fin 3, ∑ k : Fin 3, ppt_qutrit_bell_state (i,x) (k,y) * B k i let M : Matrix (Fin 3) (Fin 3) ℂ := fun x y => ∑ i : Fin 3, ppt_qutrit_bell_state (i,x) (i,y) let A : Matrix (Fin 3) (Fin 3) ℂ := vecMulVec (ppt_qutrit_bell_setting j) (star (ppt_qutrit_bell_setting j)) let N := ppt_qutrit_steered_normalizer have hA : A.PosSemidef ∧ (1-A).PosSemidef := in767_effectvalidproof_alice_settings_valid j have hS (B : Matrix (Fin 3) (Fin 3) ℂ) (hB : B.PosSemidef) : (S B).PosSemidef := by have hh := in767_effectvalidproof_positive_steering_from_four_vectors ppt_qutrit_bell_vector lam hlam B hB change Matrix.PosSemidef (show Matrix (Fin 3) (Fin 3) ℂ from fun x y => ∑ i : Fin 3, ∑ k : Fin 3, ρ (i,x) (k,y) * B k i) at hh rw [hρ] at hh exact hh have hlin : S (1-A)=M-S A := by ext x y simp [S, M, Matrix.sub_apply, Matrix.one_apply, Fin.sum_univ_three] ring have hpos : (N * S A * N).PosSemidef := by have hh := (hS A hA.1).mul_mul_conjTranspose_same N rw [in767_effectvalidproof_normalizer_hermitian] at hh exact hh have hneg : (N * S (1-A) * N).PosSemidef := by have hh := (hS (1-A) hA.2).mul_mul_conjTranspose_same N rw [in767_effectvalidproof_normalizer_hermitian] at hh exact hh have heq : N*S (1-A)*N = 1-N*S A*N := by rw [hlin, Matrix.mul_sub, Matrix.sub_mul, in767_effectvalidproof_source_normalizer_marginal] rw [heq] at hneg exact ⟨hpos,hneg⟩
For every measurement setting j ∈ {0, 1, 2}, the steered effect is valid: E_j and 1 - E_j are positive semidefinite.
Introducing the setting index j, the theorem directly applies the validity lemma established for arbitrary settings.
uses
in767_effectvalidproof_actual_steered_effects_validintro j exact in767_effectvalidproof_actual_steered_effects_valid j
Real.sq_sqrt- The square of the square root of a non-negative real number equals the number itself: √x ^ 2 = x for x ≥ 0.
Fin.sum_univ_three- A sum over the three elements of Fin 3 evaluates to the sum of the values at 0, 1, and 2.
map_ofNat- Ring homomorphisms preserve natural number literals.
Matrix.posSemidef_vecMulVec_self_star- The outer product v vᴴ of any vector v with its conjugate transpose is positive semidefinite.
Matrix.vecMulVec_mul_vecMulVec- The product of two rank-one outer products satisfies (u vᴴ)(w xᴴ) = (v ⬝ᵥ w) • (u xᴴ).
in767_effectvalidproof_alice_setting_unit- Each of Alice's three measurement vectors is normalized: v_jᴴ ⬝ᵥ v_j = 1.
one_smul- Multiplying by scalar 1 acts as the identity: 1 • b = b.
Matrix.posSemidef_conjTranspose_mul_self- For any matrix A, the product Aᴴ A is positive semidefinite.
ne_of_gt- If b < a in a preorder, then a ≠ b.
Matrix.sum_apply- Entrywise evaluation of a finite sum of matrices equals the sum of the entries.
Matrix.smul_apply- The (i, j) entry of a scalar-multiplied matrix r • A is r • A i j.
Matrix.vecMulVec_apply- The (i, j) entry of the outer product of two vectors u and v is u i * v j.
Fin.sum_univ_four- A sum over the four elements of Fin 4 evaluates to the sum of the values at 0, 1, 2, and 3.
Matrix.diagonal_apply- The (i, j) entry of a diagonal matrix is d i when i = j, and 0 when i ≠ j.
Complex.real_smul- Real scalar multiplication on complex numbers coincides with multiplication by the embedded real number.
in767_effectvalidproof_source_second_marginal- The partial trace marginal state of the two-qutrit Bell state is the diagonal matrix diag(3437/6884, 3437/6884, 5/3442).
mul_comm- Multiplication is commutative: a * b = b * a.
mul_left_comm- Left commutativity: a * (b * c) = b * (a * c).
mul_assoc- Multiplication is associative: a * b * c = a * (b * c).
Matrix.conjTranspose_apply- The (j, i) entry of the conjugate transpose Mᴴ is star (M i j).
Ne.symm- Inequality is symmetric: a ≠ b implies b ≠ a.
Matrix.posSemidef_transpose_iff.mpr- If a matrix is positive semidefinite, then its transpose is also positive semidefinite.
Finset.induction_on- Induction principle on finite sets starting from the empty set.
Matrix.PosSemidef.zero- The zero matrix is positive semidefinite.
Finset.sum_insert- Inserting a new element into a finite set splits the sum into the new term and the sum over the rest.
Finset.univ- The universal finset containing all elements of a finite type.
in767_effectvalidproof_alice_settings_valid- For each setting j, Alice's projector A and its complement 1 - A are positive semidefinite.
in767_effectvalidproof_positive_steering_from_four_vectors- Steering a four-vector conic mixture with any positive semidefinite matrix yields a positive semidefinite matrix.
in767_effectvalidproof_normalizer_hermitian- Bob's normalizer matrix N is Hermitian: Nᴴ = N.
Matrix.mul_sub- Matrix multiplication distributes over subtraction from the left: M * (N - N') = M * N - M * N'.
Matrix.sub_mul- Matrix multiplication distributes over subtraction from the right: (M - M') * N = M * N - M' * N.
in767_effectvalidproof_source_normalizer_marginal- Bob's normalizer matrix N inverts the marginal state: N * M * N = 1.
in767_effectvalidproof_actual_steered_effects_valid- For each setting j, the normalized steered effect and its complement are both positive semidefinite.
- The verification of entrywise matrix identities for the marginal state and normalizer in Steps 3 and 4 relies heavily on automated case analysis (fin_cases) and rational arithmetic simplification with square roots (norm_num, ring_nf, field_simp).
- In Step 2, the complement 1 - A is proved positive semidefinite by establishing that it is both Hermitian and idempotent, so that 1 - A = (1 - A)ᴴ (1 - A).
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
Choose a tactic — a marked line — to see the goals Lean had before and after it, as in Lean’s infoview. Step with ← → or j k.
open Matrixopen scoped BigOperators Matrix ComplexOrderset_option backward.isDefEq.respectTransparency falseset_option maxRecDepth 8192set_option maxHeartbeats 4000000set_option linter.unusedSimpArgs falseset_option linter.unnecessarySeqFocus falseprivate lemma in767_effectvalidproof_alice_setting_unit (j : Fin 3) : star (ppt_qutrit_bell_setting j) ⬝ᵥ ppt_qutrit_bell_setting j = 1 := by have h3 : (Real.sqrt 3 : ℂ)^2 = 3 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ) ≤ 3 by norm_num) have h21 : (Real.sqrt 21 : ℂ)^2 = 21 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ) ≤ 21 by norm_num) fin_cases j <;> simp [ppt_qutrit_bell_setting, dotProduct, Fin.sum_univ_three] <;> ring_nf <;> simp [h3,h21, map_ofNat] <;> norm_numprivate lemma in767_effectvalidproof_alice_settings_valid (j : Fin 3) : let A : Matrix (Fin 3) (Fin 3) ℂ := vecMulVec (ppt_qutrit_bell_setting j) (star (ppt_qutrit_bell_setting j)) A.PosSemidef ∧ (1-A).PosSemidef := by let v := ppt_qutrit_bell_setting j let A : Matrix (Fin 3) (Fin 3) ℂ := vecMulVec v (star v) have hA : A.PosSemidef := Matrix.posSemidef_vecMulVec_self_star v have hAsq : A*A=A := by dsimp [A] rw [Matrix.vecMulVec_mul_vecMulVec, in767_effectvalidproof_alice_setting_unit j, one_smul] have hQherm : (1-A).IsHermitian := Matrix.isHermitian_one.sub hA.isHermitian have hQsq : (1-A)*(1-A)=1-A := by calc (1-A)*(1-A) = 1-A-A+A*A := by noncomm_ring _ = 1-A := by rw [hAsq]; abel have hQ := Matrix.posSemidef_conjTranspose_mul_self (1-A) rw [hQherm.eq, hQsq] at hQ exact ⟨hA,hQ⟩private lemma in767_effectvalidproof_source_second_marginal : (fun x y : Fin 3 => ∑ i : Fin 3, ppt_qutrit_bell_state (i,x) (i,y)) = Matrix.diagonal ![(3437/6884 : ℂ), 3437/6884, 5/3442] := by have h2 : (Real.sqrt 2 : ℂ)^2=2 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤2 by norm_num) have h3 : (Real.sqrt 3 : ℂ)^2=3 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤3 by norm_num) have h131 : (Real.sqrt (131 : ℝ) : ℂ)^2=131 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤131 by norm_num) have hn2 : (Real.sqrt 2 : ℂ) ≠ 0 := by exact_mod_cast (ne_of_gt (Real.sqrt_pos.2 (by norm_num : (0:ℝ)<2))) have hn3 : (Real.sqrt 3 : ℂ) ≠ 0 := by exact_mod_cast (ne_of_gt (Real.sqrt_pos.2 (by norm_num : (0:ℝ)<3))) ext x y fin_cases x <;> fin_cases y <;> simp [ppt_qutrit_bell_state, ppt_qutrit_bell_vector, Matrix.sum_apply, Matrix.smul_apply, Matrix.vecMulVec_apply, Fin.sum_univ_three, Fin.sum_univ_four, Matrix.diagonal_apply, Complex.real_smul, map_ofNat] <;> field_simp [hn2,hn3] <;> ring_nf <;> simp [h2,h3,h131] <;> norm_numprivate lemma in767_effectvalidproof_source_normalizer_marginal : let M : Matrix (Fin 3) (Fin 3) ℂ := fun x y => ∑ i : Fin 3, ppt_qutrit_bell_state (i,x) (i,y) ppt_qutrit_steered_normalizer * M * ppt_qutrit_steered_normalizer = 1 := by dsimp only rw [in767_effectvalidproof_source_second_marginal] have h6884 : (Real.sqrt 6884 : ℂ)^2=6884 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤6884 by norm_num) have h3437 : (Real.sqrt 3437 : ℂ)^2=3437 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤3437 by norm_num) have h3442 : (Real.sqrt 3442 : ℂ)^2=3442 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤3442 by norm_num) have h5 : (Real.sqrt 5 : ℂ)^2=5 := by exact_mod_cast Real.sq_sqrt (show (0:ℝ)≤5 by norm_num) have hn3437 : (Real.sqrt 3437 : ℂ) ≠ 0 := by exact_mod_cast (ne_of_gt (Real.sqrt_pos.2 (by norm_num : (0:ℝ)<3437))) have hn5 : (Real.sqrt 5 : ℂ) ≠ 0 := by exact_mod_cast (ne_of_gt (Real.sqrt_pos.2 (by norm_num : (0:ℝ)<5))) have hs0 : (Real.sqrt 6884 : ℂ) / (Real.sqrt 3437 : ℂ) * (3437/6884) * ((Real.sqrt 6884 : ℂ) / (Real.sqrt 3437 : ℂ)) = 1 := by field_simp [hn3437] linear_combination 3437*h6884-6884*h3437 have hs2 : (Real.sqrt 3442 : ℂ) / (Real.sqrt 5 : ℂ) * (5/3442) * ((Real.sqrt 3442 : ℂ) / (Real.sqrt 5 : ℂ)) = 1 := by field_simp [hn5] linear_combination 5*h3442-3442*h5 ext i j fin_cases i <;> fin_cases j <;> simp only [ppt_qutrit_steered_normalizer, Matrix.mul_apply, Fin.sum_univ_three, Matrix.diagonal_apply, Matrix.cons_val_zero, Matrix.cons_val_one, Matrix.cons_val_two] <;> norm_num <;> first | exact hs0 | simpa [mul_comm,mul_left_comm,mul_assoc] using hs0 | exact hs2private lemma in767_effectvalidproof_normalizer_hermitian : ppt_qutrit_steered_normalizerᴴ = ppt_qutrit_steered_normalizer := by ext i j by_cases h : i = j · subst j by_cases hi : i = 2 <;> simp [ppt_qutrit_steered_normalizer, Matrix.conjTranspose_apply, Matrix.diagonal_apply, hi] · simp [ppt_qutrit_steered_normalizer, Matrix.conjTranspose_apply, Matrix.diagonal_apply, h, Ne.symm h]private lemma in767_effectvalidproof_positive_steering_from_four_vectors (v : Fin 4 → Fin 3 × Fin 3 → ℂ) (lam : Fin 4 → ℝ) (hlam : ∀ t, 0 ≤ lam t) (A : Matrix (Fin 3) (Fin 3) ℂ) (hA : A.PosSemidef) : Matrix.PosSemidef (show Matrix (Fin 3) (Fin 3) ℂ from fun x y => ∑ i : Fin 3, ∑ j : Fin 3, (∑ t : Fin 4, (lam t : ℂ) * v t (i,x) * star (v t (j,y))) * A j i) := by let B : Fin 4 → Matrix (Fin 3) (Fin 3) ℂ := fun t x i => v t (i,x) have hterm (t : Fin 4) : (B t * A.transpose * (B t)ᴴ).PosSemidef := (Matrix.posSemidef_transpose_iff.mpr hA).mul_mul_conjTranspose_same (B t) have hsum : (∑ t, lam t • (B t * A.transpose * (B t)ᴴ)).PosSemidef := by have H (T : Finset (Fin 4)) : (∑ t ∈ T, lam t • (B t * A.transpose * (B t)ᴴ)).PosSemidef := by induction T using Finset.induction_on with | empty => simpa using (Matrix.PosSemidef.zero : (0 : Matrix (Fin 3) (Fin 3) ℂ).PosSemidef) | @insert t T ht ih => rw [Finset.sum_insert ht] exact ((hterm t).smul (hlam t)).add ih exact H Finset.univ have heq : (fun x y : Fin 3 => ∑ i : Fin 3, ∑ j : Fin 3, (∑ t : Fin 4, (lam t : ℂ) * v t (i,x) * star (v t (j,y))) * A j i) = ∑ t, lam t • (B t * A.transpose * (B t)ᴴ) := by ext x y simp [B, Matrix.sum_apply, Matrix.smul_apply, Complex.real_smul, Matrix.mul_apply, Matrix.transpose_apply, Matrix.conjTranspose_apply, Fin.sum_univ_three, Fin.sum_univ_four] ring rw [heq] exact hsumprivate lemma in767_effectvalidproof_actual_steered_effects_valid (j : Fin 3) : (ppt_qutrit_steered_effect j).PosSemidef ∧ (1-ppt_qutrit_steered_effect j).PosSemidef := by let lam : Fin 4 → ℝ := fun t => if t = 0 then 3257/6884 else if t = 3 then 27/6884 else 450/1721 let ρ : Matrix (Fin 3 × Fin 3) (Fin 3 × Fin 3) ℂ := fun ax cy => ∑ t : Fin 4, (lam t : ℂ) * ppt_qutrit_bell_vector t ax * star (ppt_qutrit_bell_vector t cy) have hlam : ∀ t, 0 ≤ lam t := by intro t; dsimp [lam]; split_ifs <;> norm_num have hρ : ρ = ppt_qutrit_bell_state := by ext ax cy simp [ρ, ppt_qutrit_bell_state, lam, Fin.sum_univ_four, Matrix.sum_apply, Matrix.smul_apply, Matrix.vecMulVec_apply, Pi.star_apply, Complex.real_smul, mul_assoc] let S (B : Matrix (Fin 3) (Fin 3) ℂ) : Matrix (Fin 3) (Fin 3) ℂ := fun x y => ∑ i : Fin 3, ∑ k : Fin 3, ppt_qutrit_bell_state (i,x) (k,y) * B k i let M : Matrix (Fin 3) (Fin 3) ℂ := fun x y => ∑ i : Fin 3, ppt_qutrit_bell_state (i,x) (i,y) let A : Matrix (Fin 3) (Fin 3) ℂ := vecMulVec (ppt_qutrit_bell_setting j) (star (ppt_qutrit_bell_setting j)) let N := ppt_qutrit_steered_normalizer have hA : A.PosSemidef ∧ (1-A).PosSemidef := in767_effectvalidproof_alice_settings_valid j have hS (B : Matrix (Fin 3) (Fin 3) ℂ) (hB : B.PosSemidef) : (S B).PosSemidef := by have hh := in767_effectvalidproof_positive_steering_from_four_vectors ppt_qutrit_bell_vector lam hlam B hB change Matrix.PosSemidef (show Matrix (Fin 3) (Fin 3) ℂ from fun x y => ∑ i : Fin 3, ∑ k : Fin 3, ρ (i,x) (k,y) * B k i) at hh rw [hρ] at hh exact hh have hlin : S (1-A)=M-S A := by ext x y simp [S, M, Matrix.sub_apply, Matrix.one_apply, Fin.sum_univ_three] ring have hpos : (N * S A * N).PosSemidef := by have hh := (hS A hA.1).mul_mul_conjTranspose_same N rw [in767_effectvalidproof_normalizer_hermitian] at hh exact hh have hneg : (N * S (1-A) * N).PosSemidef := by have hh := (hS (1-A) hA.2).mul_mul_conjTranspose_same N rw [in767_effectvalidproof_normalizer_hermitian] at hh exact hh have heq : N*S (1-A)*N = 1-N*S A*N := by rw [hlin, Matrix.mul_sub, Matrix.sub_mul, in767_effectvalidproof_source_normalizer_marginal] rw [heq] at hneg exact ⟨hpos,hneg⟩theorem ppt_qutrit_steered_effect_is_valid : ∀ j : Fin 3, (ppt_qutrit_steered_effect j).PosSemidef ∧ (1 - ppt_qutrit_steered_effect j).PosSemidef := by intro j exact in767_effectvalidproof_actual_steered_effects_valid j