Let $E$ be a complex Hilbert space, with inner product $\langle\cdot\;, \;\cdot\rangle$ and the norm $\|\cdot\|$. Let $T\in \mathcal{L}(E)$ be bounded linear operators from $E$ to $E$ and $M\in \mathcal{L}(E)^+$ (i.e. $M^*=M$ and $\langle Mx\;, \;x\rangle\geq 0$ for all $x\in E$.

Assume that there exists a sequence $(y_n)\subset E$ such that $\langle My_n\;, \;y_n\rangle=1$.

Does the sequence $(\langle MTy_n\;, \;y_n\rangle)_n$ converge or have a subsequence that converges?

In the case when $M=I$, clearly $(\langle MTy_n\;, \;y_n\rangle)_n$ is bounded, so it has a subsequence that converges.


If $M$ is invertible then $(y_n)_n$ must be a bounded sequence since $$1 = \langle My_n, y_n\rangle = \|M^{1/2}y_n\|^2 \geq \frac{\|y_n\|^2}{\|M^{-1/2}\|^2}. $$ Thus, $(\langle MTy_n, y_n\rangle)_n$ is a bounded sequence by the Cauchy-Schwarz inequality and so it has a convergent subsequence.

The Cauchy-Schwarz inequality also gives that $(\langle MTy_n, y_n\rangle)_n$ will have a convergent subsequence if $(y_n)_n$ has a bounded subsequence or if $T$ commutes with $M$.

If $M$ is not invertible, $(y_n)_n$ has no bounded subsequence and $T$ does not commute with $M$ then in general there will be no convergent subsequence of $(\langle MTy_n, y_n\rangle)_n$. For example, let $$ M = \left[\begin{array}{cccc} 1 \\ & \frac{1}{2!} \\ && \frac{1}{3!} \\ &&&\ddots \end{array}\right], \ \ y_n = \frac{\sqrt{n!}}{\sqrt 2}e_n + \frac{\sqrt{(n+1)!}}{\sqrt 2}e_{n+1}, \ \ \textrm{and} \ \ T = \left[\begin{array}{cccc} 0&1\\&0&1\\&&0&\ddots\\&&&\ddots \end{array}\right]. $$ Then $\langle My_n, y_n\rangle = 1$ and $$\langle MTy_n, y_n\rangle = \frac{\sqrt{(n+1)!}\sqrt{n!}}{2n!} = \frac{\sqrt{n+1}}{2} \rightarrow \infty$$ which also has no convergent subsequences.

  • $\begingroup$ Why M is not invertible in your example? Thank you $\endgroup$ – Student Nov 1 '17 at 20:10
  • $\begingroup$ @Student The "inverse" is not bounded. $\endgroup$ – Chris Ramsey Nov 1 '17 at 20:21
  • $\begingroup$ If T commutes with M, then the sequence $\{\langle MTy_n,y_n\rangle\}_{n}$ is bounded. $\endgroup$ – T. Le Nov 3 '17 at 1:28
  • $\begingroup$ If $T$ commute with $M^{1/2}$, then $(\langle MTy_n, y_n\rangle)_n$ is bounded. But if $T$ commute with $M$, why we have $(\langle MTy_n, y_n\rangle)_n$ is bounded? Thank you $\endgroup$ – Student Nov 4 '17 at 11:57
  • 1
    $\begingroup$ By functional calculus if $T$ commutes with $M$ then it also commutes with $M^{1/2}$. $\endgroup$ – Chris Ramsey Nov 4 '17 at 13:16

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.