Explanation of reversion

edit

I reverted the edits; the characterization given was incorrect:

Suppose W is a partial isometry. Then W W* is a self-adjoint projection, so that in particular

 

Also assume as was claimed in the previous edit, that any partial isometry satisfies:

 

Then multiply both sides by W:

 

The square of a partial isometry is a projection only for very special partial isometries (it fails for instance for the unilateral shift). CSTAR 05:14, 9 Mar 2005 (UTC)