Sign up ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

If $H$ is a Hilbert space and $T$ is in $\mathcal{L}(H)$, the numerical range of $T$ is defined by $$W(T) := \left\{(Tx; x) \mid x \in H,\ \|x\| = 1 \right\}.$$ We have to prove that

  1. The point and residual spectrum are subsets of $W(T)$.
  2. The continuous spectrum is a subset of closure of $W(T)$.

Please help me out, Thank you.

share|cite|improve this question
"We have to prove that" - is this an exercise from a course? I ask, because it might be the case that one is meant to prove these using specific results that have already been shown in that particular course – user16299 Jun 30 '12 at 22:39
By the way, in your title, it should be the "numerical range" – user16299 Jun 30 '12 at 22:39
We can prove it in any way we want. However, there was a hint to the second part: If s does not belong to closure of W(T). Prove that norm(sId-T) is greater or equal to dist(s;W(T))norm(x). Conclude that (sId - T) : H to R(sId - T) is continuously invertible. – johnathan Jun 30 '12 at 22:46
Thanks for the clarification. BTW, if this is homework, then I believe posters are encouraged to tag it with the "homework" tag. – user16299 Jul 1 '12 at 5:12
Hello, actually, this is not homework per se, since our teacher gave us a list of recommended questions from this topic, but it is voluntary to solve them. I couldn't get the solutions to some of the questions. – johnathan Jul 1 '12 at 6:56

1 Answer 1

up vote 2 down vote accepted

$\lambda$ is in the point spectrum.

By definition it exists a $x\in H$ such that $\Vert x \Vert = 1$ and $Tx = \lambda x$. We have $$ (Tx,x) = (\lambda x, x) = \lambda $$ so $\lambda$ belongs to $W(T)$.

$\lambda$ is in the residual spectrum.

Then it exists $x\in H$ such that $\Vert x \Vert = 1$ and $x$ is orthogonal to the range of $T - \lambda$.
For each $y\in H$ we have $((T - \lambda)y, x) = 0$, in particular $$ 0 = ((T - \lambda)x, x) = (Tx, x) - \lambda $$ Also in this case $\lambda$ belongs to $W(T)$.

$\lambda$ is in the continuous spectrum.

$(T- \lambda)^{-1}$ is not bounded, therefore it exists a sequence $\{x_n\}_{n\in \mathbb N}$ of elements of $H$, with $\Vert x_n\Vert = 1$ and $\Vert (T - \lambda)^{-1} x_n\Vert \to \infty$.
Let's consider the sequence $$ y_n := \frac {(T - \lambda)^{-1} x_n} {\Vert (T - \lambda)^{-1} x_n \Vert} $$ we have $\Vert y_n \Vert = 1$ and $$ (T - \lambda)y_n = \frac {x_n} {\Vert (T - \lambda)^{-1} x_n \Vert} \to 0 $$ If $\lambda \notin \overline{W(T)}$ then it exists a $M > 0$ such that $\vert \lambda - (Tx, x)\vert > M$ for each $x\in H$, $\Vert x \Vert = 1$. As conseguence $$ \Vert (T-\lambda)x \Vert = \Vert (T-\lambda)x \Vert \Vert x \Vert \geq \vert ((T - \lambda)x, x)\vert = \vert \lambda - (Tx, x)\vert > M $$ for each $x\in H$, $\Vert x \Vert = 1$. But that contradicts the existence of the sequence $y_n$.

share|cite|improve this answer
Thank you. It is indeed clear now. – johnathan Jul 3 '12 at 10:15

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.