Two-vector quadratic-form identities in real or complex inner-product spaces #
Five elementary lemmas useful for block-operator perturbation estimates and planar motions:
LinearMap.IsSymmetric.re_inner_map_symm: the real mixed coefficients agree for symmetric maps;LinearMap.re_inner_smul_add_smul_map: expand the real quadratic form of a two-vector real-linear combination;TauCeti.abs_re_inner_le_of_norm_le:‖w‖ ≤ ε * ‖v‖bounds|re ⟪u, w⟫|by‖u‖ * ‖v‖ * ε, so an operator bound‖H v‖ ≤ ε * ‖v‖bounds the real part of every coefficient⟪u, H v⟫;TauCeti.norm_smul_add_smul_sq: the Pythagorean norm formula for a two-vector real-linear combination when the real part of the mixed inner product vanishes;TauCeti.norm_smul_add_smul_eq_one_of_sq_add_sq: its unit-circle case,‖γ y + σ z‖ = 1for orthonormaly, zandγ² + σ² = 1.
No finite-dimensionality, spectral-gap, or eigenvector assumptions are needed. The results
are used by the per-eigenvector double-angle estimates in DoubleAngle.Vector.
Adapted from the Apache-2.0 DKPS formalization (Kitware, Inc.).
Real parts of the two mixed entries of a symmetric operator agree; the reason no complex phase alignment is needed anywhere in this file.
Quadratic form of an operator at a real linear combination of two vectors, expanded into the four scalar entries. Pure sesquilinear algebra.
If ‖w‖ ≤ ε * ‖v‖, then |re ⟪u, w⟫| ≤ ‖u‖ * ‖v‖ * ε. For w = H v this says that an
operator bound ‖H v‖ ≤ ε * ‖v‖ bounds the real part of every coefficient ⟪u, H v⟫.
Squared norm of a real linear combination of two vectors orthogonal with respect to the real part of the inner product.
For unit vectors y, z orthogonal with respect to the real part of the inner product, the
real linear combination γ y + σ z with γ² + σ² = 1 is a unit vector; this is
norm_smul_add_smul_sq read on the unit circle.