qutrit_rank_four_gram_schur_one
For a rank-four Gram factor with three orthonormal upper rows, a rank-one complement gives a rank-one Schur residual.
Statement
theorem qutrit_rank_four_gram_schur_one (U V : Matrix (Fin 3) (Fin 4) ℂ) (w : Fin 4 → ℂ) (hcomp : 1 - Uᴴ * U = vecMulVec w (star w)) : let B := U * Vᴴ let u := V *ᵥ w V * Vᴴ = Bᴴ * B + vecMulVec u (star u)
- Source
- Gram-factor helper for the rank-four source in OP-693.
- Verified
- 12 Nov 2026
- Axioms
Classical.choiceQuot.soundpropext
The proof, explained
The proof decomposes the Gram matrix V V* by inserting the resolution of the identity I = (I - U* U) + U* U between V and V*. Distributing the multiplication splits V V* into a projection term V (U* U) V* and a residual term V (I - U* U) V*. By the rank-one hypothesis I - U* U = w w*, the residual simplifies to the rank-one outer product u u* where u = V w, while the projection term factors into B* B where B = U V*.
It suffices to show that V V* = (U V*)* (U V*) + (V w)(V w)*
We expand the definitions of the shorthand variables B := U V* and u := V w in the goal.
dsimp onlyV (w w*) V* = (V w)(V w)*
We establish an auxiliary identity showing that conjugating the rank-one outer product w w* by V yields the outer product of V w with itself. Multiplying w w* on the left by V produces (V w) w* using Matrix.mul_vecMulVec, and multiplying on the right by V* produces (V w)(w* V*) using Matrix.vecMulVec_mul. Finally, Matrix.star_mulVec equates the row vector w* V* with the conjugate vector (V w)*.
uses
Matrix.mul_vecMulVec,Matrix.vecMulVec_mul,Matrix.star_mulVechave hw : V * vecMulVec w (star w) * Vᴴ = vecMulVec (V *ᵥ w) (star (V *ᵥ w)) := by rw [Matrix.mul_vecMulVec, Matrix.vecMulVec_mul, Matrix.star_mulVec]
V V* = V ((I - U* U) + U* U) V*
We insert the decomposition of the identity matrix I = (I - U* U) + U* U between V and V*. The terms -U* U and U* U cancel algebraically, so simplification confirms this is equal to V V*.
calc V * Vᴴ = V * ((1 - Uᴴ * U) + Uᴴ * U) * Vᴴ := by simp
V ((I - U* U) + U* U) V* = V (U* U) V* + V (I - U* U) V*
We expand the product by distributing V from the left using Matrix.mul_add and V* from the right using Matrix.add_mul. Commutativity of addition (add_comm) swaps the two terms so that the main projection term comes first.
uses
Matrix.mul_add,Matrix.add_mul,add_comm_ = V * (Uᴴ * U) * Vᴴ + V * (1 - Uᴴ * U) * Vᴴ := by rw [Matrix.mul_add, Matrix.add_mul, add_comm]
V (U* U) V* + V (I - U* U) V* = (U V*)* (U V*) + (V w)(V w)*
In the second term, we substitute the hypothesis I - U* U = w w* (hcomp) and use the auxiliary identity hw to rewrite V (w w*) V* as (V w)(V w)*. In the first term, we express V U* as (U V*)* using Matrix.conjTranspose_mul and Matrix.conjTranspose_conjTranspose. Associativity of matrix multiplication (Matrix.mul_assoc) regroups the factors into (U V*)* (U V*), completing the equality with B* B + u u*.
uses
Matrix.conjTranspose_mul,Matrix.conjTranspose_conjTranspose,Matrix.mul_assoc_ = (U * Vᴴ)ᴴ * (U * Vᴴ) + vecMulVec (V *ᵥ w) (star (V *ᵥ w)) := by rw [hcomp, hw, Matrix.conjTranspose_mul, Matrix.conjTranspose_conjTranspose] simp only [Matrix.mul_assoc]
Matrix.mul_vecMulVec- Multiplying a matrix on the left by an outer product x y* yields the outer product (M x) y*.
Matrix.vecMulVec_mul- Multiplying an outer product x y* on the right by a matrix M yields the outer product x (y M).
Matrix.star_mulVec- The conjugate of a matrix-vector product M v equals the vector-matrix product of the conjugate of v with the conjugate transpose of M, i.e., (M v)* = v* M*.
Matrix.mul_add- Matrix multiplication distributes over matrix addition from the left: L (M + N) = L M + L N.
Matrix.add_mul- Matrix multiplication distributes over matrix addition from the right: (L + M) N = L N + M N.
add_comm- Addition is commutative: a + b = b + a.
Matrix.conjTranspose_mul- The conjugate transpose of a matrix product is the product of conjugate transposes in reverse order: (M N)* = N* M*.
Matrix.conjTranspose_conjTranspose- The conjugate transpose operation is an involution: (M*)* = M.
Matrix.mul_assoc- Matrix multiplication is associative: (L M) N = L (M N).
- Line 11 (`dsimp only`) is a Lean-specific syntactic step to expand the local let-bindings for B and u in the goal, which on paper is purely definitional.
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 falsetheorem qutrit_rank_four_gram_schur_one (U V : Matrix (Fin 3) (Fin 4) ℂ) (w : Fin 4 → ℂ) (hcomp : 1 - Uᴴ * U = vecMulVec w (star w)) : let B := U * Vᴴ let u := V *ᵥ w V * Vᴴ = Bᴴ * B + vecMulVec u (star u) := by dsimp only have hw : V * vecMulVec w (star w) * Vᴴ = vecMulVec (V *ᵥ w) (star (V *ᵥ w)) := by rw [Matrix.mul_vecMulVec, Matrix.vecMulVec_mul, Matrix.star_mulVec] calc V * Vᴴ = V * ((1 - Uᴴ * U) + Uᴴ * U) * Vᴴ := by simp _ = V * (Uᴴ * U) * Vᴴ + V * (1 - Uᴴ * U) * Vᴴ := by rw [Matrix.mul_add, Matrix.add_mul, add_comm] _ = (U * Vᴴ)ᴴ * (U * Vᴴ) + vecMulVec (V *ᵥ w) (star (V *ᵥ w)) := by rw [hcomp, hw, Matrix.conjTranspose_mul, Matrix.conjTranspose_conjTranspose] simp only [Matrix.mul_assoc]