Definition
Partial isometry
A bounded Hilbert-space operator that is isometric on the orthogonal complement of its kernel.
Definition
Let and be Hilbert spaces. A bounded operator is a partial isometry if its restriction to preserves norms. The closed subspace is the initial space, and is the final space; the final space is closed because this restriction is an isometry. Equivalently, using the adjoint, is the orthogonal projection onto the initial space. Then is the orthogonal projection onto the final space.
Polar decomposition
Every bounded operator has a polar decomposition
where is a partial isometry with initial space and final space . Requiring makes this partial isometry unique. Thus partial isometries are the correct polar factors even when is neither injective nor surjective Conway, Chapter II.
Operator-algebra formulation
An element of a -algebra is called a partial isometry when is a projection; this implies that is also a projection. In a concrete operator algebra the two projections record the initial and final spaces. Partial isometries therefore implement Murray–von Neumann equivalence between projections.
Examples and non-examples
Every orthogonal projection, unitary operator, and isometric embedding is a partial isometry. The unilateral shift is a partial isometry whose initial space is all of and whose final space has codimension one. The zero operator also qualifies, with zero initial and final spaces. By contrast, on a nonzero Hilbert space is not a partial isometry because it does not preserve norms on the orthogonal complement of its kernel.
References
- John B. Conway, A Course in Functional Analysis, 2nd ed., Graduate Texts in Mathematics 96, Springer, 1990. DOI record. Relevant: Chapter II on adjoints and polar decomposition.
- Richard V. Kadison and John R. Ringrose, Fundamentals of the Theory of Operator Algebras, Volume I: Elementary Theory, American Mathematical Society, 1997. DOI record. Relevant: §2.5 on partial isometries and projections.