MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let T be a (unbounded) selfadjoint operator in $B(H)$, the bounded operator acting on Hilbert space $H$.

Def: A compression of T is an operator $pTp$, where $p$ is a projection in $B(H)$.

I am thinking the relation between the resolution of identity (spectrum decomposition) of $pTp$ and that of $T$. I guess there should be something like $e_{pTp}(-\infty, t)$ $\le $ $e_T(-\infty, t)$ $\le$ $e_{pTp}(-\infty, t) + (I-p)$.. But I do not know how to prove it. I think someone here might help me out.

share|cite|improve this question

You should define your notations. But let me make a guess: $e_T(-\infty,t)$ is the unitary projector over the subspace $H(T,t)$, invariant under $T$, on which $T\le tI$. And an inequality between projectors means that the range of the first one is included in that of the second. Right ?

If yes to that, then the answer to your question is No. Because there is no reason why $T$ and $pTp$ have an invariant subspace in common. Take the finite dimensional case in dimension $2$, and $$T=\begin{pmatrix} -1 & a \\\\ a & 1 \end{pmatrix},$$ with $a\ne0$. Take $p$ the projection onto the first axis. Finally, take $t=-\frac12$. Then $e_{pTp}(-\infty,t)=p$, while $e_T(-\infty,t)$ is the projection onto a line which is not an axis (the eigenvalues of $T$ are $\pm\sqrt{1+a^2}$). Thus both projectors are not comparable.

share|cite|improve this answer
Denis, Thanks for your answer. You guess is correct. I am sorry I should clarify my notations. If we write $T=\int_R \lambda d(e_{\lambda}) $, then $e_T(-\infty, t)$ means the projections corresponding to unique spectral decomposition of the operator $T$. Similarly, for $e_{pTp}(-\infty, t)$. And $e_{pTp}(-\infty, t)$ $\le$ $e_T(-\infty, t)$ means the range of the first projection is included in the range of the second projection. – Paul Z Nov 22 '10 at 15:22
But $e_{pTp}(-\infty, t)$ $\le$ $e_T(-\infty, t)$ does not imply that the range of $e_{pTp}(-\infty, t)$ is invariant under $T$. – Paul Z Nov 22 '10 at 15:25
This is exactly what I mean. In my example, $e_{pTp}(-\infty,t)$ is the eigenspace (a line) of $pTp$ associated with the eigenvalue $-\sqrt{1+a^2}$, whereas $e_{T}(-\infty,t)$ is the eigenspace (a line) of $T$ associated with the eigenvalue $-1$. These lines turn out to be distinct, so there is no inequality between the projectors. – Denis Serre Nov 22 '10 at 15:38

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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