AFTD Auto-Formalizing Theoretical Domains

three_orthonormal_rows_codimension_one

theoremverified

The complementary projection of three complex orthonormal rows in four dimensions is a rank-one projector.

Statement

theorem three_orthonormal_rows_codimension_one :
  ∀ U : Matrix (Fin 3) (Fin 4) ℂ, U*Uᴴ = 1 →
    ∃ w : Fin 4 → ℂ,
      1 - Uᴴ*U = vecMulVec w (star w)
Source
Linear-algebra helper for the rank-four source in OP-693.
Verified
12 Nov 2026
Axioms
Classical.choiceQuot.soundpropext
Built on 1
  • VThe ambient space $V = \mathbb{F}_p^n \times \mathbb{F}_p^n$ of pairs of coordinate vectors.
Used by 2

The proof, explained

To show that the complementary projection 1 - Uᴴ U of three orthonormal rows U ∈ ℂ³ˣ⁴ is rank one, we construct an explicit unit vector w ∈ ℂ⁴ in the orthogonal complement of the row space using the four signed 3 × 3 minors of U (the four-dimensional cross product). By Laplace expansion along repeated rows, U w = 0, and by the Cauchy-Binet formula, ‖w‖² = det(U Uᴴ) = det(1) = 1. Appending star w as the fourth row completes U to a unitary 4 × 4 matrix V satisfying V Vᴴ = 1. Because V is square, Vᴴ V = 1 as well; expanding the product Vᴴ V = Uᴴ U + w wᴴ = 1 yields 1 - Uᴴ U = w wᴴ.

  1. The vector w formed by the signed 3 × 3 minors of U satisfies U w = 0.

    We check componentwise that U *ᵥ w = 0. For each row i ∈ {0, 1, 2}, the dot product of row i with w represents the Laplace expansion of a 4 × 4 determinant with row i repeated, which vanishes identically. Expanding all 3 × 3 determinants by Sarrus' formula and simplifying via ring arithmetic verifies that each coordinate equals 0.

    lines 13–16

    uses Fin.sum_univ_four, Matrix.det_fin_three, Matrix.mulVec, Matrix.submatrix_apply, dotProduct, in767_codimproof_codim_vector

      ext i
      fin_cases i <;>
        simp [in767_codimproof_codim_vector, Matrix.mulVec, dotProduct, Fin.sum_univ_four,
          Matrix.det_fin_three, Matrix.submatrix_apply] <;> ring
  2. The squared Euclidean norm of w equals det(U Uᴴ).

    By the Cauchy-Binet formula in dimension 3 × 4, the sum of the squared absolute values of the maximal minors of U equals det(U Uᴴ). Expanding both sides into polynomials in the matrix entries and their complex conjugates, ring normalization establishes the identity.

    lines 20–23

    uses Fin.sum_univ_four, Fin.sum_univ_three, Matrix.conjTranspose_apply, Matrix.det_fin_three, Matrix.mul_apply, Matrix.submatrix_apply, dotProduct, in767_codimproof_codim_vector, star_add, star_mul, star_sub

      simp [in767_codimproof_codim_vector, dotProduct, Fin.sum_univ_four, Matrix.det_fin_three,
        Matrix.submatrix_apply, Matrix.mul_apply, Matrix.conjTranspose_apply,
        star_add, star_sub, star_mul, Fin.sum_univ_three]
      ring
  3. The rows of U and the vector w satisfy mutual orthonormality relations.

    Assuming U Uᴴ = 1, the rows of U are orthonormal (hrows). Since U w = 0, each row of U is orthogonal to w (hker), and conjugating shows star w is orthogonal to each row of U (hkerc). Because ‖w‖² = det(U Uᴴ) = det(1) = 1, w is a unit vector (hnorm). The sums over Fin 4 and complex conjugation operations are then expanded for subsequent matrix algebra.

    lines 30–46

    uses Complex.star_def, Fin.sum_univ_four, Matrix.conjTranspose_apply, Matrix.det_one, Matrix.mul_apply, Matrix.mulVec, congrArg, congrFun, dotProduct, in767_codimproof_codim_vector, in767_codimproof_codim_vector_kernel, in767_codimproof_codim_vector_norm, star_mul, star_sum

      let w := in767_codimproof_codim_vector U
      have hrows (i j : Fin 3) :
          (∑ k : Fin 4, U i k * star (U j k)) = (1 : Matrix (Fin 3) (Fin 3) ℂ) i j := by
        have h := congrArg (fun M : Matrix (Fin 3) (Fin 3) ℂ => M i j) hU
        simpa [Matrix.mul_apply, Matrix.conjTranspose_apply] using h
      have hker (i : Fin 3) : (∑ k : Fin 4, U i k * w k)=0 := by
        have h := congrFun (in767_codimproof_codim_vector_kernel U) i
        simpa [w,Matrix.mulVec,dotProduct] using h
      have hkerc (i : Fin 3) : (∑ k : Fin 4, star (w k) * star (U i k))=0 := by
        have h := congrArg star (hker i)
        simpa [star_sum, star_mul] using h
      have hnorm : (∑ k : Fin 4, star (w k)*w k)=1 := by
        have h := in767_codimproof_codim_vector_norm U
        rw [hU, Matrix.det_one] at h
        simpa [w, dotProduct] using h
      simp only [Fin.sum_univ_four] at hrows hker hkerc hnorm
      simp only [w, Complex.star_def] at hrows hker hkerc hnorm
  4. The completed 4 × 4 matrix V satisfies V Vᴴ = 1.

    We verify componentwise that each of the 16 entries of V Vᴴ equals the corresponding entry of the 4 × 4 identity matrix. When both indices are in {0, 1, 2}, the entry is δ_ij by the orthonormality of the rows of U; when one index is 3, the entry vanishes because row 3 is orthogonal to the rows of U; and entry (3, 3) equals 1 because w has unit norm.

    lines 47–50

    uses Fin.sum_univ_four, Matrix.conjTranspose_apply, Matrix.mul_apply, Matrix.one_apply, in767_codimproof_completed_rows

      ext i j
      fin_cases i <;> fin_cases j <;>
        simp [in767_codimproof_completed_rows, w, Matrix.mul_apply, Matrix.conjTranspose_apply, Fin.sum_univ_four,
          Matrix.one_apply, hrows, hker, hkerc, hnorm]
  5. The completed matrix V also satisfies Vᴴ V = 1.

    Since V is a square 4 × 4 matrix over ℂ satisfying V Vᴴ = 1, right invertibility implies two-sided invertibility (mul_eq_one_comm.mp), so Vᴴ V = 1.

    lines 54–57

    uses in767_codimproof_completed_rows_isometry, mul_eq_one_comm.mp

      let w := in767_codimproof_codim_vector U
      let V := in767_codimproof_completed_rows U
      have hV : V*Vᴴ=1 := in767_codimproof_completed_rows_isometry U hU
      have hV' : Vᴴ*V=1 := mul_eq_one_comm.mp hV
  6. The vector w witnesses that 1 - Uᴴ U = w wᴴ.

    Providing w as the witness, we evaluate entry (i, j) of Vᴴ V = 1. The definition of V decomposes (Vᴴ V)_{ij} into (Uᴴ U)_{ij} + w_i * star (w_j). Since this equals δ_ij, subtracting (Uᴴ U)_{ij} gives (1 - Uᴴ U)_{ij} = w_i * star (w_j) = (vecMulVec w (star w))_{ij}.

    lines 58–66

    uses Complex.star_def, Fin.sum_univ_four, Fin.sum_univ_three, Matrix.conjTranspose_apply, Matrix.mul_apply, Matrix.sub_apply, Matrix.vecMulVec_apply, Pi.star_apply, congrArg, in767_codimproof_completed_rows

      refine ⟨w, ?_⟩
      ext i j
      have hh := congrArg (fun M : Matrix (Fin 4) (Fin 4) ℂ => M i j) hV'
      simp only [Matrix.mul_apply, Matrix.conjTranspose_apply, Fin.sum_univ_four] at hh
      simp only [Matrix.sub_apply, Matrix.mul_apply, Matrix.conjTranspose_apply,
        Matrix.vecMulVec_apply, Pi.star_apply, Fin.sum_univ_three]
      simp [V, in767_codimproof_completed_rows, w] at hh
      simp only [w, Complex.star_def] at hh ⊢
      linear_combination -hh
  7. For any 3 × 4 matrix U with orthonormal rows, 1 - Uᴴ U is a rank-one projector.

    We introduce U and the orthonormality hypothesis U Uᴴ = 1, and directly invoke the projection lemma in767_codimproof_codimension_one_projection.

    lines 72–73

    uses in767_codimproof_codimension_one_projection

      intro U hU
      exact in767_codimproof_codimension_one_projection U hU
Complex.star_def
The ring automorphism of complex conjugation on ℂ.
Fin.sum_univ_three
The sum of a function over the three-element set Fin 3 is f 0 + f 1 + f 2.
Fin.sum_univ_four
The sum of a function over the four-element set Fin 4 is f 0 + f 1 + f 2 + f 3.
Matrix.conjTranspose_apply
The (j, i) entry of the conjugate transpose Mᴴ is the complex conjugate of M i j.
Matrix.det_fin_three
The determinant of a 3 × 3 matrix given by Sarrus' rule (six permutation terms).
Matrix.det_one
The determinant of the identity matrix is 1.
Matrix.mul_apply
The (i, k) entry of the matrix product M * N is the sum ∑ j, M i j * N j k.
Matrix.mulVec
Matrix-vector multiplication: (M *ᵥ v) i is the dot product of row i with v.
Matrix.one_apply
The (i, j) entry of the identity matrix is 1 when i = j and 0 otherwise.
Matrix.sub_apply
The (i, j) entry of the matrix difference A - B is A i j - B i j.
Matrix.submatrix_apply
The (i, j) entry of a submatrix is the entry of the original matrix at the reindexed coordinates.
Matrix.vecMulVec_apply
The (i, j) entry of the outer product vecMulVec w v is w i * v j.
Pi.star_apply
The conjugate of a vector evaluated at index i is the conjugate of its i-th component.
congrArg
Equal inputs to a function produce equal outputs: a₁ = a₂ implies f a₁ = f a₂.
congrFun
Equal functions produce equal outputs on any input: f = g implies f a = g a.
dotProduct
The algebraic dot product of two vectors, defined as the sum of componentwise products.
in767_codimproof_codim_vector
Constructs the 4-vector of signed 3 × 3 minors of a 3 × 4 matrix U.
in767_codimproof_codim_vector_kernel
The minor vector w of U satisfies U *ᵥ w = 0.
in767_codimproof_codim_vector_norm
The squared Euclidean norm of the minor vector w equals det(U * Uᴴ).
in767_codimproof_codimension_one_projection
For any 3 × 4 matrix U with orthonormal rows, 1 - Uᴴ * U is the rank-one projector vecMulVec w (star w).
in767_codimproof_completed_rows
Constructs the 4 × 4 matrix obtained by appending star w as the fourth row to U.
in767_codimproof_completed_rows_isometry
If U * Uᴴ = 1, then the completed 4 × 4 matrix V satisfies V * Vᴴ = 1.
mul_eq_one_comm.mp
For square matrices over a commutative ring, right invertibility implies left invertibility: M * N = 1 implies N * M = 1.
star_add
Complex conjugation distributes over addition: star (r + s) = star r + star s.
star_mul
Complex conjugation reverses multiplication: star (r * s) = star s * star r.
star_sub
Complex conjugation distributes over subtraction: star (r - s) = star r - star s.
star_sum
Complex conjugation distributes over finite sums: star (∑ x ∈ s, f x) = ∑ x ∈ s, star (f 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 Matrixopen scoped BigOperators Matrix ComplexOrderset_option backward.isDefEq.respectTransparency falseset_option maxRecDepth 8192set_option maxHeartbeats 4000000set_option linter.unusedSimpArgs falseprivate noncomputable def in767_codimproof_codim_vector (U : Matrix (Fin 3) (Fin 4) ℂ) : Fin 4 → ℂ :=  ![(U.submatrix id ![1,2,3]).det, -(U.submatrix id ![0,2,3]).det,    (U.submatrix id ![0,1,3]).det, -(U.submatrix id ![0,1,2]).det]private lemma in767_codimproof_codim_vector_kernel (U : Matrix (Fin 3) (Fin 4) ℂ) :    U *ᵥ in767_codimproof_codim_vector U = 0 := by  ext i  fin_cases i <;>    simp [in767_codimproof_codim_vector, Matrix.mulVec, dotProduct, Fin.sum_univ_four,      Matrix.det_fin_three, Matrix.submatrix_apply] <;> ringprivate lemma in767_codimproof_codim_vector_norm (U : Matrix (Fin 3) (Fin 4) ℂ) :    star (in767_codimproof_codim_vector U) ⬝ᵥ in767_codimproof_codim_vector U = (U*Uᴴ).det := by  simp [in767_codimproof_codim_vector, dotProduct, Fin.sum_univ_four, Matrix.det_fin_three,    Matrix.submatrix_apply, Matrix.mul_apply, Matrix.conjTranspose_apply,    star_add, star_sub, star_mul, Fin.sum_univ_three]  ringprivate noncomputable def in767_codimproof_completed_rows (U : Matrix (Fin 3) (Fin 4) ℂ) : Matrix (Fin 4) (Fin 4) ℂ :=  ![U 0, U 1, U 2, star (in767_codimproof_codim_vector U)]private lemma in767_codimproof_completed_rows_isometry (U : Matrix (Fin 3) (Fin 4) ℂ) (hU : U*Uᴴ=1) :    in767_codimproof_completed_rows U * (in767_codimproof_completed_rows U)ᴴ = 1 := by  let w := in767_codimproof_codim_vector U  have hrows (i j : Fin 3) :      (∑ k : Fin 4, U i k * star (U j k)) = (1 : Matrix (Fin 3) (Fin 3) ℂ) i j := by    have h := congrArg (fun M : Matrix (Fin 3) (Fin 3) ℂ => M i j) hU    simpa [Matrix.mul_apply, Matrix.conjTranspose_apply] using h  have hker (i : Fin 3) : (∑ k : Fin 4, U i k * w k)=0 := by    have h := congrFun (in767_codimproof_codim_vector_kernel U) i    simpa [w,Matrix.mulVec,dotProduct] using h  have hkerc (i : Fin 3) : (∑ k : Fin 4, star (w k) * star (U i k))=0 := by    have h := congrArg star (hker i)    simpa [star_sum, star_mul] using h  have hnorm : (∑ k : Fin 4, star (w k)*w k)=1 := by    have h := in767_codimproof_codim_vector_norm U    rw [hU, Matrix.det_one] at h    simpa [w, dotProduct] using h  simp only [Fin.sum_univ_four] at hrows hker hkerc hnorm  simp only [w, Complex.star_def] at hrows hker hkerc hnorm  ext i j  fin_cases i <;> fin_cases j <;>    simp [in767_codimproof_completed_rows, w, Matrix.mul_apply, Matrix.conjTranspose_apply, Fin.sum_univ_four,      Matrix.one_apply, hrows, hker, hkerc, hnorm]private lemma in767_codimproof_codimension_one_projection (U : Matrix (Fin 3) (Fin 4) ℂ) (hU : U*Uᴴ=1) :    ∃ w : Fin 4 → ℂ, 1-Uᴴ*U=vecMulVec w (star w) := by  let w := in767_codimproof_codim_vector U  let V := in767_codimproof_completed_rows U  have hV : V*Vᴴ=1 := in767_codimproof_completed_rows_isometry U hU  have hV' : Vᴴ*V=1 := mul_eq_one_comm.mp hV  refine ⟨w, ?_⟩  ext i j  have hh := congrArg (fun M : Matrix (Fin 4) (Fin 4) ℂ => M i j) hV'  simp only [Matrix.mul_apply, Matrix.conjTranspose_apply, Fin.sum_univ_four] at hh  simp only [Matrix.sub_apply, Matrix.mul_apply, Matrix.conjTranspose_apply,    Matrix.vecMulVec_apply, Pi.star_apply, Fin.sum_univ_three]  simp [V, in767_codimproof_completed_rows, w] at hh  simp only [w, Complex.star_def] at hh ⊢  linear_combination -hhtheorem three_orthonormal_rows_codimension_one :  ∀ U : Matrix (Fin 3) (Fin 4) ℂ, U*Uᴴ = 1 →    ∃ w : Fin 4 → ℂ,      1 - Uᴴ*U = vecMulVec w (star w) := by  intro U hU  exact in767_codimproof_codimension_one_projection U hU
See it in the graphDiscuss this resultChallenge it