Reducing subspaces of inner-product spaces #
A subspace reduces an endomorphism when both it and its orthogonal complement are invariant.
For a symmetric endomorphism, invariance of the subspace alone is enough. When an orthogonal
projection exists, reduction is equivalently commutation with that projection; in particular the
orthogonal projection onto a subspace invariant under a symmetric endomorphism commutes with it.
More generally, an idempotent whose range and kernel are orthogonal both for the inner product and
for the form of a symmetric endomorphism commutes with that endomorphism
(LinearMap.IsSymmetric.commute_of_isIdempotentElem).
For a bounded operator A on a Hilbert space and a complete subspace invariant under both A
and its adjoint, the adjoint of the restriction of A is the restriction of the adjoint
(ContinuousLinearMap.adjoint_restrict), so the restriction of a normal operator is normal
(IsStarNormal.restrict).
A subspace is reducing for an endomorphism when both it and its orthogonal complement are invariant under the endomorphism.
Instances For
A subspace is reducing exactly when it and its orthogonal complement are invariant.
A reducing subspace is invariant under the endomorphism.
The orthogonal complement of a reducing subspace is invariant under the endomorphism.
The orthogonal complement reduces an endomorphism exactly when the original subspace does.
A subspace reduces a symmetric endomorphism exactly when it is invariant.
A subspace admitting an orthogonal projection reduces an endomorphism exactly when its orthogonal projection commutes with that endomorphism.
The orthogonal projection onto a subspace invariant under a symmetric endomorphism commutes with that endomorphism.
An idempotent endomorphism p whose range is orthogonal to its kernel, both for the inner
product and for the form โชA ยท, ยทโซ of a symmetric endomorphism A, commutes with A: its range
and its kernel are invariant under A.
The zero subspace reduces every endomorphism.
The whole space reduces every endomorphism.
The supremum of a family of reducing subspaces is reducing.
The span of two reducing subspaces is reducing.
A subspace with an orthogonal projection reduces a bounded operator A exactly when A
commutes with the orthogonal projection onto it.
A subspace with an orthogonal projection reduces a bounded operator A on a Hilbert space
exactly when it is invariant under both A and its adjoint.
A subspace with an orthogonal projection reduces the adjoint of a bounded operator A on a
Hilbert space exactly when it reduces A.
On a finite-dimensional space, a subspace reduces an endomorphism A exactly when it is
invariant under both A and its adjoint.
On a finite-dimensional space, a subspace reduces the adjoint of an endomorphism A exactly
when it reduces A.
For an invariant subspace of a symmetric map, applying the complementary projection to
A X - X M is the same as forming the residual of the projected trial map. No isometry or
finite-dimensionality hypothesis is needed.
The corresponding DKPS projected-residual statement is in
ForTauCeti/Analysis/InnerProductSpace/Residual/AngleEmbedding.lean (Jon Crall and
GPT 5.6 High; Kitware, Apache-2.0). The proof uses invariant-projection commutation.
The adjoint of the restriction of a bounded operator A to a complete subspace invariant
under both A and its adjoint is the restriction of the adjoint.
The restriction of a normal operator to a complete subspace invariant under both the operator and its adjoint is normal.