AFTD Auto-Formalizing Theoretical Domains

qutrit_product_sum_congruence

theoremverified

A matrix congruence transforms a sum of rank-one operators by transforming each factor.

Statement

theorem qutrit_product_sum_congruence
    (F : Matrix (Fin 3) (Fin 3) ℂ) (v : Fin 4 → Fin 3 → ℂ) (c : Fin 4 → ℂ) :
    F * (∑ t, c t • vecMulVec (v t) (star (v t))) * Fᴴ =
      ∑ t, c t • vecMulVec (F *ᵥ v t) (star (F *ᵥ v t))
Source
Product-decomposition helper for OP-693.
Verified
12 Nov 2026
Axioms
Classical.choiceQuot.soundpropext

The proof, explained

The proof establishes the congruence identity by linearity and the algebraic interaction between matrix multiplication and rank-one outer products. Distributing the matrix multiplications over the summation and factoring out the scalar coefficients reduces the theorem to showing that each rank-one summand transforms as $F (v_t v_t^*) Fᴴ = (F v_t) (F v_t)^*$. This termwise identity then follows by associating matrix multiplication with the vectors forming the outer product and using the conjugate-transpose property $(F v_t)^* = v_t^* Fᴴ$.

  1. Distributing the congruence transformation across the sum and factoring out scalars reduces the proof to verifying the identity term by term: for every index $t \in \operatorname{Fin} 4$, $c(t) \cdot (F (v(t) v(t)^*) Fᴴ) = c(t) \cdot ((F v(t)) (F v(t))^*)$.

    By the linearity of matrix multiplication, the factors $F$ on the left and $Fᴴ$ on the right distribute over the summation (`Matrix.mul_sum`, `Matrix.sum_mul`) and commute past the scalar coefficients $c(t)$ (`Matrix.mul_smul`, `Matrix.smul_mul`). By congruence on the summation (`congr 1`) and functional extensionality (`funext t`), proving that the two sums are equal reduces to proving that their summands are equal for an arbitrary index $t$.

    lines 10–12

    uses Matrix.mul_sum, Matrix.sum_mul, Matrix.mul_smul, Matrix.smul_mul

      simp only [Matrix.mul_sum, Matrix.sum_mul, Matrix.mul_smul, Matrix.smul_mul]
      congr 1
      funext t
  2. For each index $t$, the transformed rank-one operator $F (v(t) v(t)^*) Fᴴ$ equals $(F v(t)) (F v(t))^*$, completing the proof.

    Multiplying the outer product $v(t) v(t)^*$ on the left by $F$ produces $(F v(t)) v(t)^*$ by `Matrix.mul_vecMulVec`, and multiplying on the right by $Fᴴ$ produces $(F v(t)) (v(t)^* Fᴴ)$ by `Matrix.vecMulVec_mul`. On the right-hand side, `Matrix.star_mulVec` rewrites the conjugate transpose $(F v(t))^*$ into the identical row-vector product $v(t)^* Fᴴ$, so both sides match.

    line 13

    uses Matrix.mul_vecMulVec, Matrix.vecMulVec_mul, Matrix.star_mulVec

      rw [Matrix.mul_vecMulVec, Matrix.vecMulVec_mul, Matrix.star_mulVec]
Matrix.mul_sum
Left matrix multiplication distributes over a finite sum: $M (\sum_i A_i) = \sum_i M A_i$.
Matrix.sum_mul
Right matrix multiplication distributes over a finite sum: $(\sum_i A_i) M = \sum_i A_i M$.
Matrix.mul_smul
Matrix multiplication on the left commutes with scalar multiplication: $M (a \cdot N) = a \cdot (M N)$.
Matrix.smul_mul
Matrix multiplication on the right commutes with scalar multiplication: $(a \cdot M) N = a \cdot (M N)$.
Matrix.mul_vecMulVec
Left-multiplying a rank-one outer product matrix by a matrix acts on the left vector: $M (x y^T) = (M x) y^T$.
Matrix.vecMulVec_mul
Right-multiplying a rank-one outer product matrix by a matrix acts on the right vector: $(x y^T) M = x (y M)^T$.
Matrix.star_mulVec
The conjugate transpose of a matrix-vector product is the product of the conjugated vector with the conjugate transpose of the matrix: $(M v)^* = v^* Mᴴ$.

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 falsetheorem qutrit_product_sum_congruence    (F : Matrix (Fin 3) (Fin 3) ℂ) (v : Fin 4 → Fin 3 → ℂ) (c : Fin 4 → ℂ) :    F * (∑ t, c t • vecMulVec (v t) (star (v t))) * Fᴴ =      ∑ t, c t • vecMulVec (F *ᵥ v t) (star (F *ᵥ v t)) := by  simp only [Matrix.mul_sum, Matrix.sum_mul, Matrix.mul_smul, Matrix.smul_mul]  congr 1  funext t  rw [Matrix.mul_vecMulVec, Matrix.vecMulVec_mul, Matrix.star_mulVec]
See it in the graphDiscuss this resultChallenge it