Partial Isometries: Introduction

1k Views Asked by At

Attention

This question has been modified drastically.

It is done so the answer below is still correct.

It is done so to allow more specialized threads.

Problem

How do I deal with partial isometries?

How come that partial isometries give rise to projections?

And what about the that weird characterization given on wiki?

Reference

For more details see wiki's: Partial Isometries

1

There are 1 best solutions below

9
On BEST ANSWER

If $\Omega^{\star}\Omega$ is a projection, then it is an orthogonal projection. It is always true that $\mathcal{N}(\Omega^{\star}\Omega)=\mathcal{N}(\Omega)$ and, hence, $\Omega^{\star}\Omega$ is the orthogonal projection onto $\mathcal{N}(\Omega)^{\perp}=\overline{\mathcal{R}(\Omega^{\star})}$. Therefore, $$\Omega^{\star}\Omega\Omega^{\star}=\Omega^{\star},\\ \Omega\Omega^{\star}\Omega=\Omega. $$ (The second follows by adjoint from the first.) Likewise, if $\Omega\Omega^{\star}$ is a projection, then the same arguments apply as before where $\Omega$ and $\Omega^{\star}$ are swapped, which is enough because the last two expressions give the same two expressions when $\Omega$ and $\Omega^{\star}$ are swapped.

Note: It is convenient to think of a partial isometry $\Omega$ as a unitary map from one closed subspace $V$ of $\mathcal{H}$ to another closed subspace $W$ of $\mathcal{H}$. Then $\Omega^{\star}$ inverts $\Omega$ on its range $W$ and $\Omega$ inverts $\Omega^{\star}$ on its range $V$. That's why $\Omega^{\star}\Omega$ is the projection onto $V$ and $\Omega\Omega^{\star}$ is the projection onto $W$.