AFTD Auto-Formalizing Theoretical Domains

qutrit_four_product_steering_parent_from_weighted

theoremverified

Steering any four-product state by three binary qutrit effects yields a finite positive joint parent, conditional on the weighted hidden-parent construction.

Statement

theorem qutrit_four_product_steering_parent_from_weighted
    (n : ℕ) (A : Fin 3 → Matrix (Fin 3) (Fin 3) ℂ)
    (v : Fin 4 → Fin 3 → ℂ) (w : Fin 4 → Fin n → ℂ)
    (hA : ∀ j, (A j).PosSemidef ∧ (1-A j).PosSemidef)
    (hweighted : ∀ (Q : Fin 4 → Matrix (Fin n) (Fin n) ℂ)
      (r : Fin 4 → ℝ) (p : Fin 3 → Fin 4 → ℝ) (P : Matrix (Fin n) (Fin n) ℂ),
      (∀ t, (Q t).PosSemidef) → (∑ t, (r t) • Q t) = P →
      (∀ t, 0 ≤ r t) → (∀ j t, 0 ≤ p j t ∧ p j t ≤ r t) →
      ∃ G : (Fin 3 → Bool) → Matrix (Fin n) (Fin n) ℂ,
        (∀ s, (G s).PosSemidef) ∧ (∑ s, G s) = P ∧
        ∀ j : Fin 3, (∑ s, if s j then G s else 0) = ∑ t, (p j t) • Q t) :
    ∃ G : (Fin 3 → Bool) → Matrix (Fin n) (Fin n) ℂ,
      (∀ s, (G s).PosSemidef) ∧
      (∑ s, G s) = (fun x y => ∑ i : Fin 3, ∑ t : Fin 4,
        (v t i * w t x) * star (v t i * w t y)) ∧
      ∀ j : Fin 3, (∑ s, if s j then G s else 0) =
        (fun x y => ∑ i : Fin 3, ∑ k : Fin 3,
          (∑ t : Fin 4, (v t i * w t x) * star (v t k * w t y)) * A j k i)
Source
Local compatibility helper for OP-693.
Verified
12 Nov 2026
Axioms
Classical.choiceQuot.soundpropext

The proof, explained

To show that steering a four-product qutrit-qudit state yields a positive joint parent, the proof applies the assumed general weighted hidden-parent construction `hweighted` to the rank-one component matrices $Q_t = w_t w_t^*$ with weights $r_t = \langle v_t, v_t \rangle$ and steered coefficients $p_j(t) = \langle v_t, A_j v_t \rangle$. The non-negativity of the weights and the inequalities $0 \le p_j(t) \le r_t$ follow directly from the effect bounds $0 \le A_j \le I$. Finally, coordinate expansions and ring automation show that the total sum and conditional marginals of the resulting joint parent $G$ coincide entrywise with the standard marginal and steered state assemblages.

  1. Applying the weighted hidden-parent hypothesis to the rank-one component matrices $Q_t = w_t w_t^*$ reduces the problem to checking non-negativity of weights and probability bounds.

    We instantiate the assumed weighted hidden-parent hypothesis `hweighted` with components $Q_t = w_t w_t^*$, weights $r_t = \operatorname{Re}(\langle v_t, v_t \rangle)$, steered coefficients $p_j(t) = \operatorname{Re}(\langle v_t, A_j v_t \rangle)$, and total sum $P = \sum_t r_t Q_t$. Each matrix $w_t w_t^*$ is positive semidefinite by `Matrix.posSemidef_vecMulVec_self_star`, and the sum condition holds by definition (`rfl`).

    lines 21–27

    uses Matrix.posSemidef_vecMulVec_self_star

      apply hweighted (fun t => vecMulVec (w t) (star (w t)))
        (fun t => (star (v t) ⬝ᵥ v t).re)
        (fun j t => (star (v t) ⬝ᵥ ((A j) *ᵥ v t)).re)
        (∑ t, (star (v t) ⬝ᵥ v t).re • vecMulVec (w t) (star (w t)))
      · intro t
        exact Matrix.posSemidef_vecMulVec_self_star _
      · rfl
  2. For every component $t$, the weight $r_t = \operatorname{Re}(\langle v_t, v_t \rangle)$ is non-negative.

    Because the identity matrix is positive semidefinite (`Matrix.PosSemidef.one`), the quadratic form $\langle v_t, I v_t \rangle$ is non-negative in the complex order. By `Complex.nonneg_iff.mp`, its real part is non-negative. Simplifying $I v_t = v_t$ yields $\operatorname{Re}(\langle v_t, v_t \rangle) \ge 0$.

    lines 28–31

    uses Matrix.PosSemidef.one, Complex.nonneg_iff.mp

      · intro t
        have h := (Matrix.PosSemidef.one : (1 : Matrix (Fin 3) (Fin 3) ℂ).PosSemidef).dotProduct_mulVec_nonneg (v t)
        have hr := (Complex.nonneg_iff.mp h).1
        simpa using hr
  3. For each effect $j$ and component $t$, the steered probability satisfies $0 \le p_j(t) \le r_t$.

    Since $A_j$ is positive semidefinite, the quadratic form $\langle v_t, A_j v_t \rangle$ is non-negative, giving $0 \le \operatorname{Re}(\langle v_t, A_j v_t \rangle) = p_j(t)$ via `Complex.nonneg_iff.mp`. Likewise, since $I - A_j$ is positive semidefinite, $\langle v_t, (I - A_j) v_t \rangle \ge 0$. Distributing matrix multiplication and the dot product using `Matrix.sub_mulVec`, `Matrix.one_mulVec`, `dotProduct_sub`, and `Complex.sub_re` shows that $\operatorname{Re}(\langle v_t, v_t \rangle) - \operatorname{Re}(\langle v_t, A_j v_t \rangle) \ge 0$. Linear arithmetic (`linarith`) concludes that $p_j(t) \le r_t$.

    lines 32–36

    uses Complex.nonneg_iff.mp, Matrix.sub_mulVec, Matrix.one_mulVec, dotProduct_sub, Complex.sub_re

      · intro j t
        have hlo := (Complex.nonneg_iff.mp ((hA j).1.dotProduct_mulVec_nonneg (v t))).1
        have hhi := (Complex.nonneg_iff.mp ((hA j).2.dotProduct_mulVec_nonneg (v t))).1
        simp only [Matrix.sub_mulVec, Matrix.one_mulVec, dotProduct_sub, Complex.sub_re] at hhi
        exact ⟨hlo, by linarith⟩
  4. The coordinate sum formula for steering a four-product state by effect $A$ matches the outer-product sum $\sum_t \langle v_t, A v_t \rangle w_t w_t^*$. automation

    Unfolding the definitions of `dotProduct` and `Matrix.mulVec`, expanding the finite sums over `Fin 3` and `Fin 4` using `Fin.sum_univ_three` and `Fin.sum_univ_four`, and distributing complex conjugation using `star_mul`, both sides expand to identical polynomials in the matrix and vector coordinates, which the `ring` tactic verifies.

    lines 45–47

    uses dotProduct, Matrix.mulVec, Fin.sum_univ_three, Fin.sum_univ_four, star_mul

      simp [dotProduct, Matrix.mulVec, Fin.sum_univ_three, Fin.sum_univ_four,
        star_mul]
      ring
  5. The coordinate sum formula for the marginal of a four-product state matches the outer-product sum $\sum_t \langle v_t, v_t \rangle w_t w_t^*$. automation

    Unfolding `dotProduct`, expanding the finite sums over `Fin 3` and `Fin 4` via `Fin.sum_univ_three` and `Fin.sum_univ_four`, and distributing complex conjugation via `star_mul`, both sides reduce to polynomial expressions that `ring` verifies to be identical.

    lines 53–54

    uses dotProduct, Fin.sum_univ_three, Fin.sum_univ_four, star_mul

      simp [dotProduct, Fin.sum_univ_three, Fin.sum_univ_four, star_mul]
      ring
  6. There exists a positive semidefinite parent candidate $G$ whose total and conditional sums match the weighted product sums for $m = 4$.

    We specialize the helper lemma `in767_productparent_product_assemblage_parent_of_weighted` to $m = 4$ components, obtaining a family of positive semidefinite matrices $G(s)$ satisfying the total sum property `hsum` and conditional marginal properties `hmarg`. We then refine the existential goal with $G$.

    lines 74–75

    uses in767_productparent_product_assemblage_parent_of_weighted

      obtain ⟨G, hG, hsum, hmarg⟩ := in767_productparent_product_assemblage_parent_of_weighted n 4 A v w hA hweighted
      refine ⟨G, hG, ?_, ?_⟩
  7. The total sum $\sum_s G(s)$ equals the marginal state of the four-product assemblage.

    Rewriting $\sum_s G(s)$ using `hsum` and comparing entrywise (`ext x y`), we simplify entry evaluation using `Matrix.sum_apply`, `Matrix.smul_apply`, `Matrix.vecMulVec_apply`, `Pi.star_apply`, and `smul_eq_mul`. Rewriting with `in767_productparent_marginal_four_products`, we compare the sums term by term with `Finset.sum_congr`. Because the identity matrix is positive semidefinite (`Matrix.PosSemidef.one`, `Matrix.one_mulVec`), $\langle v_t, v_t \rangle$ is non-negative (`Complex.nonneg_iff.mp`), so its imaginary part vanishes and it equals its real part (`Complex.ext`). Coercing real scalar multiplication to complex multiplication with `Complex.real_smul` equates the summands.

    lines 76–88

    uses Matrix.sum_apply, Matrix.smul_apply, Matrix.vecMulVec_apply, Pi.star_apply, smul_eq_mul, in767_productparent_marginal_four_products, Finset.sum_congr, Complex.nonneg_iff.mp, Matrix.PosSemidef.one, Matrix.one_mulVec, Complex.ext, Complex.real_smul

      · rw [hsum]
        ext x y
        simp only [Matrix.sum_apply, Matrix.smul_apply, Matrix.vecMulVec_apply, Pi.star_apply,
          smul_eq_mul]
        rw [in767_productparent_marginal_four_products]
        apply Finset.sum_congr rfl
        intro t ht
        have hn := (Complex.nonneg_iff.mp ((Matrix.PosSemidef.one :
          (1 : Matrix (Fin 3) (Fin 3) ℂ).PosSemidef).dotProduct_mulVec_nonneg (v t))).2
        simp only [Matrix.one_mulVec] at hn
        have hr : (((star (v t) ⬝ᵥ v t).re : ℝ) : ℂ) = star (v t) ⬝ᵥ v t := by
          apply Complex.ext <;> simp [hn]
        rw [Complex.real_smul, hr]
  8. For each measurement $j$, the conditional sum $\sum_{s : s(j)=\text{true}} G(s)$ equals the steered state assemblage.

    For each effect $j$, rewriting with the marginal property `hmarg j` and comparing entries at $(x, y)$, we simplify matrix operations and rewrite with `in767_productparent_steer_four_products`. Comparing terms using `Finset.sum_congr`, since $A_j$ is positive semidefinite, the quadratic form $\langle v_t, A_j v_t \rangle$ is non-negative (`Complex.nonneg_iff.mp`), so its imaginary part vanishes and it equals its real part (`Complex.ext`). Rewriting scalar multiplication with `Complex.real_smul` completes the proof.

    lines 89–101

    uses Matrix.sum_apply, Matrix.smul_apply, Matrix.vecMulVec_apply, Pi.star_apply, smul_eq_mul, in767_productparent_steer_four_products, Finset.sum_congr, Complex.nonneg_iff.mp, Complex.ext, Complex.real_smul

      · intro j
        rw [hmarg j]
        ext x y
        simp only [Matrix.sum_apply, Matrix.smul_apply, Matrix.vecMulVec_apply, Pi.star_apply,
          smul_eq_mul]
        rw [in767_productparent_steer_four_products]
        apply Finset.sum_congr rfl
        intro t ht
        have hn := (Complex.nonneg_iff.mp ((hA j).1.dotProduct_mulVec_nonneg (v t))).2
        have hr : (((star (v t) ⬝ᵥ (A j *ᵥ v t)).re : ℝ) : ℂ) =
            star (v t) ⬝ᵥ (A j *ᵥ v t) := by
          apply Complex.ext <;> simp [hn]
        rw [Complex.real_smul, hr]
Matrix.posSemidef_vecMulVec_self_star
The outer product of a vector with its conjugate transpose, $w w^*$, is positive semidefinite.
Matrix.PosSemidef.one
The identity matrix is positive semidefinite.
Complex.nonneg_iff.mp
A complex number is non-negative if and only if its imaginary part is zero and its real part is non-negative.
Matrix.sub_mulVec
Matrix-vector multiplication distributes over subtraction of matrices.
Matrix.one_mulVec
Multiplying a vector by the identity matrix yields the vector.
dotProduct_sub
The vector dot product distributes over subtraction in the right argument.
Complex.sub_re
The real part of a difference of complex numbers is the difference of their real parts.
dotProduct
The dot product of two vectors, defined by $\sum_i u_i v_i$.
Matrix.mulVec
Matrix-vector multiplication, defined by $(M v)_i = \sum_j M_{i, j} v_j$.
Fin.sum_univ_three
A sum over `Fin 3` equals the sum of its terms at 0, 1, and 2.
Fin.sum_univ_four
A sum over `Fin 4` equals the sum of its terms at 0, 1, 2, and 3.
star_mul
Complex conjugation reverses multiplication: $(r s)^* = s^* r^*$.
in767_productparent_product_assemblage_parent_of_weighted
For an $m$-component product mixture and effects $0 \le A_j \le I$, the weighted hidden-parent hypothesis provides a positive semidefinite joint parent reproducing the weighted product sums.
Matrix.sum_apply
The $(i, j)$-th entry of a sum of matrices is the sum of their $(i, j)$-th entries.
Matrix.smul_apply
The $(i, j)$-th entry of a scalar multiple of a matrix is the scalar multiple of the entry.
Matrix.vecMulVec_apply
The $(i, j)$-th entry of the outer product $w v^*$ is $w_i v_j^*$.
Pi.star_apply
The $i$-th component of the conjugate of a vector is the conjugate of its $i$-th component.
smul_eq_mul
Scalar multiplication in a ring agrees with ring multiplication.
in767_productparent_marginal_four_products
For 4 product components in dimension 3, the coordinate formula for the marginal state equals $\sum_t \langle v_t, v_t \rangle w_t w_t^*$.
Finset.sum_congr
Two finite sums are equal if their index sets are equal and their summands agree at every index.
Complex.ext
Two complex numbers are equal if their real parts and imaginary parts are respectively equal.
Complex.real_smul
Scalar multiplication of a complex number by a real scalar equals multiplication by the real scalar coerced to ℂ.
in767_productparent_steer_four_products
For 4 product components in dimension 3, the coordinate formula for the steered state under effect $A$ equals $\sum_t \langle v_t, A v_t \rangle w_t w_t^*$.

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 Matrixopen scoped BigOperators Matrix ComplexOrderset_option backward.isDefEq.respectTransparency falseset_option linter.unusedSimpArgs falseprivate lemma in767_productparent_product_assemblage_parent_of_weighted    (n m : ℕ) (A : Fin 3 → Matrix (Fin 3) (Fin 3) ℂ)    (v : Fin m → Fin 3 → ℂ) (w : Fin m → Fin n → ℂ)    (hA : ∀ j, (A j).PosSemidef ∧ (1-A j).PosSemidef)    (hweighted : ∀ (Q : Fin m → Matrix (Fin n) (Fin n) ℂ)      (r : Fin m → ℝ) (p : Fin 3 → Fin m → ℝ) (P : Matrix (Fin n) (Fin n) ℂ),      (∀ t, (Q t).PosSemidef) → (∑ t, (r t) • Q t) = P →      (∀ t, 0 ≤ r t) → (∀ j t, 0 ≤ p j t ∧ p j t ≤ r t) →      ∃ G : (Fin 3 → Bool) → Matrix (Fin n) (Fin n) ℂ,        (∀ s, (G s).PosSemidef) ∧ (∑ s, G s) = P ∧        ∀ j : Fin 3, (∑ s, if s j then G s else 0) = ∑ t, (p j t) • Q t) :    ∃ G : (Fin 3 → Bool) → Matrix (Fin n) (Fin n) ℂ,      (∀ s, (G s).PosSemidef) ∧      (∑ s, G s) = ∑ t, (star (v t) ⬝ᵥ v t).re • vecMulVec (w t) (star (w t)) ∧      ∀ j : Fin 3, (∑ s, if s j then G s else 0) =        ∑ t, (star (v t) ⬝ᵥ ((A j) *ᵥ v t)).re • vecMulVec (w t) (star (w t)) := by  apply hweighted (fun t => vecMulVec (w t) (star (w t)))    (fun t => (star (v t) ⬝ᵥ v t).re)    (fun j t => (star (v t) ⬝ᵥ ((A j) *ᵥ v t)).re)    (∑ t, (star (v t) ⬝ᵥ v t).re • vecMulVec (w t) (star (w t)))  · intro t    exact Matrix.posSemidef_vecMulVec_self_star _  · rfl  · intro t    have h := (Matrix.PosSemidef.one : (1 : Matrix (Fin 3) (Fin 3) ℂ).PosSemidef).dotProduct_mulVec_nonneg (v t)    have hr := (Complex.nonneg_iff.mp h).1    simpa using hr  · intro j t    have hlo := (Complex.nonneg_iff.mp ((hA j).1.dotProduct_mulVec_nonneg (v t))).1    have hhi := (Complex.nonneg_iff.mp ((hA j).2.dotProduct_mulVec_nonneg (v t))).1    simp only [Matrix.sub_mulVec, Matrix.one_mulVec, dotProduct_sub, Complex.sub_re] at hhi    exact ⟨hlo, by linarith⟩private lemma in767_productparent_steer_four_products    {n : ℕ} (A : Matrix (Fin 3) (Fin 3) ℂ)    (v : Fin 4 → Fin 3 → ℂ) (w : Fin 4 → Fin n → ℂ) (x y : Fin n) :    (∑ i : Fin 3, ∑ j : Fin 3,      (∑ t : Fin 4, (v t i * w t x) * star (v t j * w t y)) * A j i) =      ∑ t : Fin 4, (star (v t) ⬝ᵥ (A *ᵥ v t)) * (w t x * star (w t y)) := by  simp [dotProduct, Matrix.mulVec, Fin.sum_univ_three, Fin.sum_univ_four,    star_mul]  ringprivate lemma in767_productparent_marginal_four_products    {n : ℕ} (v : Fin 4 → Fin 3 → ℂ) (w : Fin 4 → Fin n → ℂ) (x y : Fin n) :    (∑ i : Fin 3, ∑ t : Fin 4, (v t i * w t x) * star (v t i * w t y)) =      ∑ t : Fin 4, (star (v t) ⬝ᵥ v t) * (w t x * star (w t y)) := by  simp [dotProduct, Fin.sum_univ_three, Fin.sum_univ_four, star_mul]  ringtheorem qutrit_four_product_steering_parent_from_weighted    (n : ℕ) (A : Fin 3 → Matrix (Fin 3) (Fin 3) ℂ)    (v : Fin 4 → Fin 3 → ℂ) (w : Fin 4 → Fin n → ℂ)    (hA : ∀ j, (A j).PosSemidef ∧ (1-A j).PosSemidef)    (hweighted : ∀ (Q : Fin 4 → Matrix (Fin n) (Fin n) ℂ)      (r : Fin 4 → ℝ) (p : Fin 3 → Fin 4 → ℝ) (P : Matrix (Fin n) (Fin n) ℂ),      (∀ t, (Q t).PosSemidef) → (∑ t, (r t) • Q t) = P →      (∀ t, 0 ≤ r t) → (∀ j t, 0 ≤ p j t ∧ p j t ≤ r t) →      ∃ G : (Fin 3 → Bool) → Matrix (Fin n) (Fin n) ℂ,        (∀ s, (G s).PosSemidef) ∧ (∑ s, G s) = P ∧        ∀ j : Fin 3, (∑ s, if s j then G s else 0) = ∑ t, (p j t) • Q t) :    ∃ G : (Fin 3 → Bool) → Matrix (Fin n) (Fin n) ℂ,      (∀ s, (G s).PosSemidef) ∧      (∑ s, G s) = (fun x y => ∑ i : Fin 3, ∑ t : Fin 4,        (v t i * w t x) * star (v t i * w t y)) ∧      ∀ j : Fin 3, (∑ s, if s j then G s else 0) =        (fun x y => ∑ i : Fin 3, ∑ k : Fin 3,          (∑ t : Fin 4, (v t i * w t x) * star (v t k * w t y)) * A j k i) := by  obtain ⟨G, hG, hsum, hmarg⟩ := in767_productparent_product_assemblage_parent_of_weighted n 4 A v w hA hweighted  refine ⟨G, hG, ?_, ?_⟩  · rw [hsum]    ext x y    simp only [Matrix.sum_apply, Matrix.smul_apply, Matrix.vecMulVec_apply, Pi.star_apply,      smul_eq_mul]    rw [in767_productparent_marginal_four_products]    apply Finset.sum_congr rfl    intro t ht    have hn := (Complex.nonneg_iff.mp ((Matrix.PosSemidef.one :      (1 : Matrix (Fin 3) (Fin 3) ℂ).PosSemidef).dotProduct_mulVec_nonneg (v t))).2    simp only [Matrix.one_mulVec] at hn    have hr : (((star (v t) ⬝ᵥ v t).re : ℝ) : ℂ) = star (v t) ⬝ᵥ v t := by      apply Complex.ext <;> simp [hn]    rw [Complex.real_smul, hr]  · intro j    rw [hmarg j]    ext x y    simp only [Matrix.sum_apply, Matrix.smul_apply, Matrix.vecMulVec_apply, Pi.star_apply,      smul_eq_mul]    rw [in767_productparent_steer_four_products]    apply Finset.sum_congr rfl    intro t ht    have hn := (Complex.nonneg_iff.mp ((hA j).1.dotProduct_mulVec_nonneg (v t))).2    have hr : (((star (v t) ⬝ᵥ (A j *ᵥ v t)).re : ℝ) : ℂ) =        star (v t) ⬝ᵥ (A j *ᵥ v t) := by      apply Complex.ext <;> simp [hn]    rw [Complex.real_smul, hr]
See it in the graphDiscuss this resultChallenge it