Consider a linear operator $V$ between two Hilbert spaces $H_A$ and $H_B$, we say that $V:H_A\rightarrow H_B$ is an isometry if $$ V^*V=\mathbb{1}_A$$

if $V$ is an isometry, does it always hold that $VV^*=\mathbb{1}_B$? In other words, is the adjoint on an isometry an isometry? If not, does it in finite dimension? If not even in finite dimension, does anyone have a finite dimensional counterexample?


It will hold in finite dimensional spaces if $\dim(H_A) = \dim(H_B)$. It will not, however, hold more generally.

If we take $H_A = \Bbb C^2$ and $H_B = \Bbb C^3$, then the map $$ V(x,y) = (x,y,0) $$ is an isometry. Its adjoint $$ V^*(x,y,z) = (x,y) $$ is not an isometry.

For an infinite dimensional example with $H_A = H_B$, we can take the right shift operator over $\ell^2$.

  • $\begingroup$ Thank you, math doesn't forgive I guess, does it hold in finite dimension that the restriction of $VV^*$ to a space of the same dimension of $H_A$ is the identity? It holds for the example you provided $\endgroup$ – user438666 Feb 4 at 13:45
  • 2
    $\begingroup$ @user438666 Yes, and you can say a bit more: $VV^*$ will always be the orthogonal projection onto the image of $V$ in $H_B$. $\endgroup$ – Omnomnomnom Feb 4 at 14:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.