Documentation

Oseledets.Lyapunov.ExteriorNorm.Plucker

Singular-value submultiplicativity, Plücker coordinates and the eigenvalue ceiling #

Building on the exterior-power operator-norm engine and the compound matrix, this module proves the submultiplicativity of the product of the top-k singular values, the Rayleigh / sin-Θ off-diagonal estimates, and the Plücker bridge giving the top eigenpair and second-eigenvalue ceiling of the compound of a gapped symmetric map in compound-matrix coordinates.

Main results #

Submultiplicativity of the product of singular values #

Submultiplicativity of the product of the top-k singular values, assembled from the submultiplicativity engine and the singular-value bridge: ∏_{i<k} σᵢ(g ∘ f) ≤ (∏_{i<k} σᵢ(g)) · (∏_{i<k} σᵢ(f)).

The rank-1 exterior Rayleigh-deficit bound #

The band-projector increment reduces to a rank-1 dominant-eigenvector sin Θ estimate (sin_sq_le_rayleigh_deficit_div_gap in Oseledets.Lyapunov.OseledetsLimit). This section provides the deficit-side pieces feeding that core: the per-vector compound operator-norm step (Lemma 1), the Rayleigh quotient identity and top-eigenvalue ceiling μ₀ = ‖compound‖² (Lemma 2), and the assembled deficit bound μ₀ − ⟨C_n v', v'⟩ ≤ (1 − 1/κ²)·μ₀ (Lemma 3), with κ = ‖compound B‖·‖(compound B)⁻¹‖ the compound condition number.

Per-vector L2 operator-norm bound for toEuclideanLin: ‖toEuclideanLin N w‖ ≤ ‖N‖·‖w‖. Routed through the bundled continuous-linear-map toEuclideanCLM, whose operator norm is the L2 matrix norm ‖N‖ by Matrix.l2_opNorm_toEuclideanCLM.

The k-th compound of the identity matrix is the identity. Via the compound bridge conjExteriorMap_eq_toEuclideanLin_compound, since ⋀^k id = id (exteriorPower.map_id).

For invertible B, compound k B⁻¹ is a right inverse of compound k B (compoundMatrix_mul + compoundMatrix_one).

For invertible B, compound k B⁻¹ is a left inverse of compound k B.

theorem Oseledets.ExteriorNorm.compoundMatrix_eq_inv_mul {d : } (k : ) {B : Matrix (Fin d) (Fin d) } (hB : B.det 0) (M : Matrix (Fin d) (Fin d) ) :

The compound factorisation compound M = (compound B)⁻¹ · compound(B · M), for invertible B. Used in Lemma 3 to lower-bound ‖compound M‖ by ‖compound(B·M)‖.

The rank-1 lower bound μ̃₀ ≥ cM²/cBi². For invertible B, the squared compound operator norm of the perturbed cocycle step B · M (= the top eigenvalue μ̃₀ of Cₙ₊₁ = adjoint Gₙ₊₁ ∘ₗ Gₙ₊₁) is bounded below by cM²/cBi², where cM = ‖compound k M‖ and cBi = ‖compound k B⁻¹‖. Route: compound k M = compound k B⁻¹ · compound k (B·M) gives cM ≤ cBi·‖compound(B·M)‖, hence ‖compound(B·M)‖ ≥ cM/cBi; squaring yields the bound.

Lemma 1 — the rank-1 per-vector step. The squared norm of the compound of a product, applied to w, is dominated by ‖compound B‖² times the squared norm of the M-compound at w: ‖compound(B·M) w‖² ≤ ‖compound B‖²·‖compound M w‖². This relates the Rayleigh quotients of the compound Gram operators C_{n+1} (from B·M) and C_n (from M). Via toEuclideanLin_compoundMatrix_mul + the per-vector operator-norm step.

Lemma 2 (Rayleigh identity). The Rayleigh quotient of the compound Gram operator C_n = adjoint(compound M) ∘ₗ compound M at w equals ‖compound M w‖².

Lemma 2 (top-eigenvalue ceiling). The Rayleigh quotient of the compound Gram operator is bounded by ‖compound M‖²·‖w‖²; equivalently the top eigenvalue μ₀ of C_n = adjoint(compound M) ∘ₗ compound M is ‖compound M‖² (the squared operator norm of the compound = top eigenvalue of AᵀA).

theorem Oseledets.ExteriorNorm.rayleigh_deficit_kernel {BM CB r CBi mu : } (hCBn : 0 CB) (hCBin : 0 CBi) (hmun : 0 mu) (hstep1 : BM CB * r) (hstep2 : mu CBi * BM) :
mu ^ 2 - r ^ 2 (1 - 1 / (CB * CBi) ^ 2) * mu ^ 2

Pure-real algebraic kernel of the deficit bound: from BM ≤ CB·r and mu ≤ CBi·BM (with all nonnegative) one gets mu² − r² ≤ (1 − 1/(CB·CBi)²)·mu².

Lemma 3 — the rank-1 exterior Rayleigh-deficit bound. For invertible B and a unit vector v' that achieves the operator norm of the compound compound (B·M) (so ‖compound(B·M) v'‖ = ‖compound(B·M)‖, i.e. v' is a top right-singular vector / dominant eigenvector of C_{n+1}), the Rayleigh deficit of the operator C_n = adjoint(compound M) ∘ₗ compound M at v' against its top value μ₀ = ‖compound M‖² obeys μ₀ − ⟨C_n v', v'⟩ ≤ (1 − 1/κ²)·μ₀ with κ = ‖compound B‖·‖(compound B)⁻¹‖.

This is the deficit-side input to sin_sq_le_rayleigh_deficit_div_gap (with ε := μ₀ − ⟨C_n v', v'⟩, μ₀ := ‖compound M‖²). The v'-achieves-the-op-norm hypothesis encodes that v' is the top eigenvector of C_{n+1}; its existence is the caller's responsibility.

The off-diagonal residual estimate and the perturbed Gram ceiling #

The refined Davis–Kahan sin-Θ estimate in off-diagonal/residual form (offdiag_sin_le_residual_div_gap in Oseledets.Lyapunov.OseledetsLimit) needs two cocycle-specific inputs:

Both are abstract operator facts (no compound/exterior structure); the cocycle specialisation in standard coordinates (where G = toEuclideanLin (compoundMatrix k ·)) follows by toEuclideanLin_compoundMatrix_mul (functoriality G' = H ∘ₗ G) and the per-vector operator-norm bound norm_toEuclideanLin_apply_le. These pieces feed the band-projector increment bound together with the back-transport norm_proj_sub_le_wedge.

theorem Oseledets.ExteriorNorm.residual_orthogonal {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace E] {C : E →ₗ[] E} {v₀ : E} (hv₀ : v₀ = 1) :
inner (C v₀ - inner (C v₀) v₀ v₀) v₀ = 0

The off-diagonal residual is orthogonal to v₀. For a unit v₀, the residual C v₀ − ⟪C v₀, v₀⟫ v₀ = (I − P) C v₀ is orthogonal to v₀.

Rayleigh of the Gram operator is the squared norm: ⟪(adjoint G ∘ₗ G) v, v⟫ = ‖G v‖² (abstract form; rayleigh_compound_eq_norm_sq is the compound-matrix specialisation).

The off-diagonal inner product reduction: ⟪(adjoint G' ∘ₗ G') v₀, z⟫ = ⟪G' v₀, G' z⟫ (plain adjoint move; for z ⊥ v₀ this is the off-diagonal block of Cₙ₊₁).

theorem Oseledets.ExteriorNorm.offdiag_residual_norm_le {E : Type u_1} {F : Type u_2} [NormedAddCommGroup E] [InnerProductSpace E] [FiniteDimensional E] [NormedAddCommGroup F] [InnerProductSpace F] [FiniteDimensional F] {G : E →ₗ[] F} {H : F →ₗ[] F} {G' : E →ₗ[] F} (hcomp : G' = H ∘ₗ G) {v₀ : E} {τ₀ τ₁ nH : } (hτ₀ : 0 τ₀) (hτ₁ : 0 τ₁) (hnH : 0 nH) (hv₀ : v₀ = 1) (htop : G v₀ = τ₀) (hH : ∀ (y : F), H y nH * y) (hperp : ∀ (z : E), inner z v₀ = 0z 1G z τ₁) :
(LinearMap.adjoint G' ∘ₗ G') v₀ - inner ((LinearMap.adjoint G' ∘ₗ G') v₀) v₀ v₀ τ₀ * τ₁ * nH ^ 2

The off-diagonal residual norm estimate. For the perturbed Gram operator Cₙ₊₁ = adjoint G' ∘ₗ G' with G' = H ∘ₗ G (functoriality) and v₀ the top unit eigenvector of Cₙ = adjoint G ∘ₗ G, the off-diagonal residual Cₙ₊₁ v₀ − ⟪Cₙ₊₁ v₀, v₀⟫ v₀ has norm at most τ₀ · τ₁ · ‖H‖², where τ₀ = ‖G v₀‖ (the top singular value of G) and τ₁ is the second-singular-value ceiling on v₀^⊥ (hperp : ∀ z ⊥ v₀, ‖z‖ ≤ 1 → ‖G z‖ ≤ τ₁).

Proof: the residual res ⊥ v₀; ‖res‖² = ⟪res, res⟫ = ⟪Cₙ₊₁ v₀, res⟫ (since res ⊥ v₀) = ⟪H G v₀, H G res⟫ ≤ ‖H‖²‖G v₀‖‖G res‖ ≤ ‖H‖² τ₀ τ₁ ‖res‖ by Cauchy–Schwarz, the per-vector operator-norm bound on H, htop, and hperp applied to the unit normalisation of res. Dividing by ‖res‖ gives the bound.

theorem Oseledets.ExteriorNorm.perturbed_gram_ceiling {E : Type u_1} {F : Type u_2} [NormedAddCommGroup E] [InnerProductSpace E] [FiniteDimensional E] [NormedAddCommGroup F] [InnerProductSpace F] [FiniteDimensional F] {G : E →ₗ[] F} {H : F →ₗ[] F} {G' : E →ₗ[] F} (hcomp : G' = H ∘ₗ G) {v₀ : E} {μ₁ nH : } (hH : ∀ (y : F), H y nH * y) (hceil : ∀ (z : E), inner z v₀ = 0inner ((LinearMap.adjoint G ∘ₗ G) z) z μ₁ * z ^ 2) (z : E) :
inner z v₀ = 0inner ((LinearMap.adjoint G' ∘ₗ G') z) z μ₁ * nH ^ 2 * z ^ 2

The ν-ceiling for the perturbed Gram operator. From a Rayleigh ceiling ∀ z ⊥ v₀, ⟪Cₙ z, z⟫ ≤ μ₁ ‖z‖² on the unperturbed Gram operator Cₙ = adjoint G ∘ₗ G, the perturbed operator Cₙ₊₁ = adjoint G' ∘ₗ G' with G' = H ∘ₗ G obeys the amplified ceiling ∀ z ⊥ v₀, ⟪Cₙ₊₁ z, z⟫ ≤ (μ₁ ‖H‖²) ‖z‖². Proof: ⟪Cₙ₊₁ z, z⟫ = ‖H G z‖² ≤ ‖H‖² ‖G z‖² = ‖H‖² ⟪Cₙ z, z⟫ ≤ ‖H‖² μ₁ ‖z‖². This supplies the ν := μ₁ ‖H‖² ceiling consumed by offdiag_sin_le_residual_div_gap.

The cocycle specialisation in compound-matrix coordinates #

Specialising offdiag_residual_norm_le / perturbed_gram_ceiling to the cocycle Gram operators Cₙ = adjoint Gₙ ∘ₗ Gₙ, Gₙ = toEuclideanLin (compoundMatrix k Mₙ), with the one-step left factor B = A(Tⁿx) (so Mₙ₊₁ = B · Mₙ and Gₙ₊₁ = (compound B) ∘ Gₙ by toEuclideanLin_compoundMatrix_mul). The SVD ceiling hperp of the abstract lemma is discharged from a μ₁-ceiling on Cₙ via rayleigh_compound_eq_norm_sq: ‖Gₙ z‖² = ⟪Cₙ z, z⟫ ≤ μ₁ ‖z‖² ≤ μ₁ for ‖z‖ ≤ 1, hence ‖Gₙ z‖ ≤ √μ₁ =: τ₁.

The off-diagonal residual estimate for the compound Gram operators. With Gₙ = toEuclideanLin (compoundMatrix k M), Cₙ = adjoint Gₙ ∘ₗ Gₙ, the one-step left factor B, and Cₙ₊₁ = adjoint Gₙ₊₁ ∘ₗ Gₙ₊₁ for Gₙ₊₁ = toEuclideanLin (compoundMatrix k (B * M)): if v₀ is a unit vector achieving the compound operator norm ‖Gₙ v₀‖ = ‖compoundMatrix k M‖ = τ₀ (the top right-singular vector of Gₙ, i.e. the top eigenvector of Cₙ) with a μ₁-Rayleigh ceiling on v₀^⊥, then the off-diagonal residual obeys ‖Cₙ₊₁ v₀ − ⟪Cₙ₊₁ v₀, v₀⟫ v₀‖ ≤ ‖compoundMatrix k M‖ · √μ₁ · ‖compoundMatrix k B‖². (τ₀ = ‖compoundMatrix k M‖, τ₁ = √μ₁, ‖H‖ = ‖compoundMatrix k B‖.)

The ν-ceiling for the perturbed compound Gram operator. From a μ₁-Rayleigh ceiling on Cₙ = adjoint Gₙ ∘ₗ Gₙ over v₀^⊥, the perturbed compound Gram operator Cₙ₊₁ = adjoint Gₙ₊₁ ∘ₗ Gₙ₊₁ (with Gₙ₊₁ = toEuclideanLin (compoundMatrix k (B * M))) obeys the amplified ceiling ∀ z ⊥ v₀, ⟪Cₙ₊₁ z, z⟫ ≤ (μ₁ ‖compoundMatrix k B‖²) ‖z‖². This is the ν := μ₁ ‖H‖² ceiling consumed by offdiag_sin_le_residual_div_gap.

The Plücker bridge #

For a symmetric PD map f with orthonormal eigenbasis u and eigenvalues lam, the compound ⋀^k f, conjugated through the eigenbasis wedge trivialization onbTriv u, is a diagonal Euclidean operator: it scales basisFun i by the subset product ∏_{a ∈ Sᵢ} lam a. The top set {0,…,k-1} (maximal by prod_le_prod_top for antitone weights) gives the top eigenvector v₀ = basisFun i₀ with eigenvalue μ₀, and every other weight is ≤ μ₁ (the second-eigenvalue ceiling). The bridge is completed by the det-Gram identity for the Plücker (wedge) inner product.

The Plücker (wedge) inner product is the cross-Gram determinant. For two families v, w : Fin k → E, the L2 inner product of their Hodge-trivialized wedges equals the determinant of the cross-Gram matrix ⟪v j, w i⟫. With orthonormal frames this is the wedge-sine identity ⟪w_E, w_E'⟫ = det(UᵀV) feeding the Frobenius back-transport norm_proj_sub_le_wedge.

noncomputable def Oseledets.ExteriorNorm.colE {d k : } (U : Matrix (Fin d) (Fin k) ) (j : Fin k) :

The j-th column of a d×k matrix, viewed as a vector in EuclideanSpace ℝ (Fin d). The columns of the band-projector frames U_top (bandProjector_indicator_eq_frame) are the orthonormal top-block eigenvectors; their wedge is the Plücker top eigenvector.

Equations
Instances For
    theorem Oseledets.ExteriorNorm.inner_colE {d k : } (U V : Matrix (Fin d) (Fin k) ) (i j : Fin k) :
    inner (colE U i) (colE V j) = (U.transpose * V) i j

    The L2 inner product of two matrix columns (as Euclidean vectors) is the cross-Gram entry (Uᵀ V) i j = ∑ₐ Uₐᵢ Vₐⱼ.

    The Plücker frame ↔ wedge determinant bridge (matrix form). For two d×k column-frames U, V, the determinant of the cross-Gram Uᵀ V equals the L2 inner product of the Hodge-trivialized wedges of their columns. This is the hdet : det(UᵀV) = ⟪vt, v₀⟫ plumbing fact consumed by Oseledets.norm_bandProjector_succ_sub_le, with v₀ = wedge of U-columns (the Plücker top eigenvector of Cₙ) and vt = wedge of V-columns (the perturbed top eigenvector of Cₙ₊₁). Since the band-projector frames U_top have orthonormal eigenvector columns (bandProjector_indicator_eq_frame) which are the same eigenbasis the Plücker eigenpair (plucker_eigenpair_ceiling_standard, applied with u = eigenvectorBasis) is built from, the two wedges are exactly these Hodge-trivialized column wedges, and this identity supplies hdet.

    Abstract diagonal Euclidean operators: eigenpair and second-eigenvalue ceiling #

    The Plücker eigenpair and second-eigenvalue ceiling #

    theorem Oseledets.ExteriorNorm.plucker_eigenpair_ceiling {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace E] [FiniteDimensional E] {n : } (f : E →ₗ[] E) (u : OrthonormalBasis (Fin n) E) (lam : ) (hanti : Antitone lam) (hpos : ∀ (i : ), 0 lam i) (hf : ∀ (i : Fin n), f (u i) = lam i u i) {k : } (hk1 : 1 k) (hkn : k n) (hgap : lam k < lam (k - 1)) :
    ∃ (i₀ : Fin (Module.finrank (⋀[]^k E))), (conjExteriorMap k (onbTriv u k) (onbTriv u k) f).IsSymmetric (conjExteriorMap k (onbTriv u k) (onbTriv u k) f) ((EuclideanSpace.basisFun (Fin (Module.finrank (⋀[]^k E))) ) i₀) = (∏ iFinset.range k, lam i) (EuclideanSpace.basisFun (Fin (Module.finrank (⋀[]^k E))) ) i₀ (∏ iFinset.range (k - 1), lam i) * lam k < iFinset.range k, lam i ∀ (w : EuclideanSpace (Fin (Module.finrank (⋀[]^k E)))), inner w ((EuclideanSpace.basisFun (Fin (Module.finrank (⋀[]^k E))) ) i₀) = 0inner ((conjExteriorMap k (onbTriv u k) (onbTriv u k) f) w) w (∏ iFinset.range (k - 1), lam i) * lam k * w ^ 2

    The Plücker bridge for a symmetric map. Let f be symmetric with orthonormal eigenbasis u : OrthonormalBasis (Fin n) and antitone nonnegative eigenvalues lam : ℕ → ℝ (f (u i) = lam i • u i). At a genuine gap lam k < lam (k-1) (with 1 ≤ k ≤ n), the conjugated compound C = ⋀^k f (through the eigenbasis wedge trivialization onbTriv u) is a symmetric operator with:

    • top eigenpair: C v₀ = μ₀ • v₀, where v₀ = basisFun i₀ is the Plücker image of the top-k eigenframe and μ₀ = ∏_{i<k} lam i;
    • second-eigenvalue ceiling: ∀ w ⊥ v₀, ⟪C w, w⟫ ≤ μ₁ ‖w‖² with μ₁ = (∏_{i<k-1} lam i)·lam k;
    • the gap: μ₁ < μ₀.

    This lands in exactly the shape consumed by sin_sq_le_rayleigh_deficit_div_gap (hC, hv₀, hev, hgap, hμ₁).

    theorem Oseledets.ExteriorNorm.plucker_eigenpair_ceiling' {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace E] [FiniteDimensional E] {n : } (f : E →ₗ[] E) (u : OrthonormalBasis (Fin n) E) (lam : ) (hanti : Antitone lam) (hpos : ∀ (i : ), 0 lam i) (hf : ∀ (i : Fin n), f (u i) = lam i u i) {k : } (hk1 : 1 k) (hkn : k n) (hgap : lam k < lam (k - 1)) :
    ∃ (i₀ : Fin (Module.finrank (⋀[]^k E))), (EuclideanSpace.basisFun (Fin (Module.finrank (⋀[]^k E))) ) i₀ = (onbTriv u k) ((exteriorPower.ιMulti k) fun (j : Fin k) => u j, ) (conjExteriorMap k (onbTriv u k) (onbTriv u k) f) ((EuclideanSpace.basisFun (Fin (Module.finrank (⋀[]^k E))) ) i₀) = (∏ iFinset.range k, lam i) (EuclideanSpace.basisFun (Fin (Module.finrank (⋀[]^k E))) ) i₀ (∏ iFinset.range (k - 1), lam i) * lam k < iFinset.range k, lam i ∀ (w : EuclideanSpace (Fin (Module.finrank (⋀[]^k E)))), inner w ((EuclideanSpace.basisFun (Fin (Module.finrank (⋀[]^k E))) ) i₀) = 0inner ((conjExteriorMap k (onbTriv u k) (onbTriv u k) f) w) w (∏ iFinset.range (k - 1), lam i) * lam k * w ^ 2

    Witness-exposing Plücker bridge (eigenbasis coords). Same as plucker_eigenpair_ceiling, but with the top eigenvector index i₀ produced explicitly as wIndexEquiv u k topSet (where topSet is the top-k prefix subset), and with the extra identity pinning the standard basis vector basisFun i₀ to the explicit Hodge-trivialized wedge onbTriv u k (e_{u₀} ∧ ⋯ ∧ e_{u_{k-1}}) of the top-k eigenframe. This is the variant plucker_eigenpair_ceiling_standard' transports to standard coordinates to expose the band-projector frame wedge.

    The reconciliation bridge: transporting the Plücker eigenpair into standard coordinates #

    plucker_eigenpair_ceiling produces the top eigenpair and second-eigenvalue ceiling of the conjugated compound conjExteriorMap k (onbTriv u) (onbTriv u) f in the eigenbasis wedge trivialization (u = an orthonormal eigenbasis of the symmetric f). The Rayleigh-deficit input rayleigh_deficit_le lives in the standard trivialization, where the conjugated compound is toEuclideanLin (compoundMatrix k ·) (the compound matrix). These are the same abstract operator ⋀^k f viewed through two isometric o.n.-basis wedge trivializations, hence unitarily equivalent by the orthogonal change-of-coordinates onbChange. Since an isometry preserves the inner product, the Rayleigh quotient is trivialization-independent; this lets sin_sq_le_rayleigh_deficit_div_gap be applied in eigenbasis coordinates with the deficit supplied from standard coordinates.

    theorem Oseledets.ExteriorNorm.eigenpair_ceiling_transport {E : Type u_2} [NormedAddCommGroup E] [InnerProductSpace E] [FiniteDimensional E] {ιE : Type u_3} {ιE' : Type u_4} [Fintype ιE] [LinearOrder ιE] [Fintype ιE'] [LinearOrder ιE'] (b : OrthonormalBasis ιE E) (b' : OrthonormalBasis ιE' E) (k : ) (f : E →ₗ[] E) (i₀ : Fin (Module.finrank (⋀[]^k E))) (μ₀ μ₁ : ) (hev : (conjExteriorMap k (onbTriv b' k) (onbTriv b' k) f) ((EuclideanSpace.basisFun (Fin (Module.finrank (⋀[]^k E))) ) i₀) = μ₀ (EuclideanSpace.basisFun (Fin (Module.finrank (⋀[]^k E))) ) i₀) (hceil : ∀ (w : EuclideanSpace (Fin (Module.finrank (⋀[]^k E)))), inner w ((EuclideanSpace.basisFun (Fin (Module.finrank (⋀[]^k E))) ) i₀) = 0inner ((conjExteriorMap k (onbTriv b' k) (onbTriv b' k) f) w) w μ₁ * w ^ 2) :

    Transport of a top-eigenpair + second-eigenvalue ceiling across the change of o.n.-wedge trivialization. Given the top eigenpair (hev) and the μ₁-ceiling on the orthogonal complement (hceil) of the conjugated compound in the b'-trivialization (b' = the eigenbasis), the same data transports — via the orthogonal W = onbChange b b' k — to the b-trivialization (b = the standard basis): the eigenvector is v₀ = W⁻¹ (basisFun i₀), the eigenvalue/gap are unchanged, and the Rayleigh ceiling holds verbatim on v₀ᗮ. This is the abstract (matrix-free) reconciliation core that feeds sin_sq_le_rayleigh_deficit_div_gap once conjExteriorMap (onbTriv b) f is identified with the standard compound.

    The Plücker eigenpair in standard (compound-matrix) coordinates #

    The matrix-level packaging plucker_eigenpair_ceiling_standard transports plucker_eigenpair_ceiling through the orthogonal change-of-trivialization onbChange (via eigenpair_ceiling_transport) into the standard wedge trivialization onbTriv (EuclideanSpace.basisFun (Fin d) ℝ), where the compound bridge conjExteriorMap_eq_toEuclideanLin_compound identifies the conjugated compound of toEuclideanLin Q with toEuclideanLin (compoundMatrix k Q) — exactly the operator consumed by the off-diagonal residual lemmas norm_offdiag_residual_compound_le / perturbed_compound_gram_ceiling.

    A single declaration combining plucker ∘ transport ∘ matrix-identification times out even at maxHeartbeats 1600000. The fix is to split the heavy matrix-identification step into an isolated scoped lemma (conjExteriorMap_basisFun_toEuclideanLin_eq_compound below — a thin alias of the compound bridge, kept separate so its ⋀^k finrank elaboration cost is contained) and to keep the transport/assembly in its own scoped declaration.

    (A) — the isolated matrix-identification step. Through the standard orthonormal-wedge trivialization (onbTriv (EuclideanSpace.basisFun (Fin d) ℝ)), the conjugated compound of toEuclideanLin M is toEuclideanLin (compoundMatrix k M). This is a thin re-export of conjExteriorMap_eq_toEuclideanLin_compound, isolated in its own scoped declaration so that the (heavy) ⋀^k finrank-indexed elaboration is paid here exactly once, keeping the assembled plucker_eigenpair_ceiling_standard under budget.

    theorem Oseledets.ExteriorNorm.plucker_eigenpair_ceiling_standard {d n : } (Q : Matrix (Fin d) (Fin d) ) (u : OrthonormalBasis (Fin n) (EuclideanSpace (Fin d))) (lam : ) (hanti : Antitone lam) (hpos : ∀ (i : ), 0 lam i) (hf : ∀ (i : Fin n), (Matrix.toEuclideanLin Q) (u i) = lam i u i) {k : } (hk1 : 1 k) (hkn : k n) (hgap : lam k < lam (k - 1)) :
    ∃ (v₀ : EuclideanSpace (Fin (Module.finrank (⋀[]^k (EuclideanSpace (Fin d)))))), v₀ = 1 (Matrix.toEuclideanLin (compoundMatrix k Q)) v₀ = (∏ iFinset.range k, lam i) v₀ (∏ iFinset.range (k - 1), lam i) * lam k < iFinset.range k, lam i ∀ (w : EuclideanSpace (Fin (Module.finrank (⋀[]^k (EuclideanSpace (Fin d)))))), inner w v₀ = 0inner ((Matrix.toEuclideanLin (compoundMatrix k Q)) w) w (∏ iFinset.range (k - 1), lam i) * lam k * w ^ 2

    (B) — plucker_eigenpair_ceiling_standard. The Plücker eigenpair + second-eigenvalue ceiling in standard compound-matrix coordinates. For a symmetric PSD f = toEuclideanLin Q with orthonormal eigenbasis u and antitone nonnegative eigenvalues lam, at a genuine gap lam k < lam (k-1), the operator toEuclideanLin (compoundMatrix k Q) (= ⋀^k Q in the standard trivialization) has:

    • top eigenpair: a unit vector v₀ with toEuclideanLin (compoundMatrix k Q) v₀ = μ₀ • v₀, μ₀ = ∏_{i<k} lam i;
    • the gap: μ₁ < μ₀ with μ₁ = (∏_{i<k-1} lam i)·lam k;
    • second-eigenvalue ceiling: ∀ w ⊥ v₀, ⟪(toEuclideanLin (compoundMatrix k Q)) w, w⟫ ≤ μ₁‖w‖².

    Assembled from plucker_eigenpair_ceiling (eigenbasis-wedge coords) → eigenpair_ceiling_transport (onbChange to standard basisFun coords) → conjExteriorMap_basisFun_toEuclideanLin_eq_compound (matrix identification, isolated in (A)). This is the top spectral data of Cₙ = ⋀^k Qₙ that norm_offdiag_residual_compound_le / perturbed_compound_gram_ceiling consume.

    theorem Oseledets.ExteriorNorm.plucker_eigenpair_ceiling_standard' {d n : } (Q : Matrix (Fin d) (Fin d) ) (u : OrthonormalBasis (Fin n) (EuclideanSpace (Fin d))) (lam : ) (hanti : Antitone lam) (hpos : ∀ (i : ), 0 lam i) (hf : ∀ (i : Fin n), (Matrix.toEuclideanLin Q) (u i) = lam i u i) {k : } (hk1 : 1 k) (hkn : k n) (hgap : lam k < lam (k - 1)) :
    (onbTriv (EuclideanSpace.basisFun (Fin d) ) k) ((exteriorPower.ιMulti k) fun (j : Fin k) => u j, ) = 1 (Matrix.toEuclideanLin (compoundMatrix k Q)) ((onbTriv (EuclideanSpace.basisFun (Fin d) ) k) ((exteriorPower.ιMulti k) fun (j : Fin k) => u j, )) = (∏ iFinset.range k, lam i) (onbTriv (EuclideanSpace.basisFun (Fin d) ) k) ((exteriorPower.ιMulti k) fun (j : Fin k) => u j, ) (∏ iFinset.range (k - 1), lam i) * lam k < iFinset.range k, lam i ∀ (w : EuclideanSpace (Fin (Module.finrank (⋀[]^k (EuclideanSpace (Fin d)))))), inner w ((onbTriv (EuclideanSpace.basisFun (Fin d) ) k) ((exteriorPower.ιMulti k) fun (j : Fin k) => u j, )) = 0inner ((Matrix.toEuclideanLin (compoundMatrix k Q)) w) w (∏ iFinset.range (k - 1), lam i) * lam k * w ^ 2

    (B') — witness-exposing plucker_eigenpair_ceiling_standard. Same spectral data as plucker_eigenpair_ceiling_standard, but with the top eigenvector produced explicitly as the standard-trivialization wedge w₀ = onbTriv basisFun k (e_{u₀} ∧ ⋯ ∧ e_{u_{k-1}}) of the top-k eigenframe of u — exactly the Plücker top eigenvector that the band-projector frame wedge equals. This is the variant whose witness can be plugged into det_transpose_mul_eq_inner_onbTriv to discharge the hdet hypothesis of Oseledets.norm_bandProjector_succ_sub_le.

    (C) — the Plücker frame ↔ wedge determinant bridge through the standard trivialization. The hdet plumbing fact for Oseledets.norm_bandProjector_succ_sub_le, expressed through the same trivialization onbTriv basisFun in which plucker_eigenpair_ceiling_standard' produces its top eigenvectors: det(UᵀV) = ⟪onbTriv basisFun (⋀ V-cols), onbTriv basisFun (⋀ U-cols)⟫. Together with plucker_eigenpair_ceiling_standard' (whose v₀/vt ARE these column wedges), this discharges the hdet hypothesis with v₀ = U-column wedge, vt = V-column wedge.