Summary: Spectrum vs. Numerical Range

871 Views Asked by At

This thread is only Q&A!

Given a Hilbert space $\mathcal{H}$.

Consider operators: $$T:\mathcal{D}(T)\to\mathcal{H}$$

Denote for shorthand: $$\Omega\subseteq\mathbb{C}:\quad\langle\Omega\rangle:=\operatorname{conv}(\Omega)$$

Then one has: $$\langle\sigma(T)\rangle=\overline{\mathcal{W}(T)}$$

Does this hold true?

1

There are 1 best solutions below

0
On BEST ANSWER

Bounded Operators

For bounded operators: $$A\in\mathcal{B}(\mathcal{H}):\quad\langle\sigma(A)\rangle=\overline{\mathcal{W}(A)}\subseteq\mathbb{C}$$

For nilpotent operators: $$N^{K+1}=0:\quad\langle\sigma(N)\rangle=(0)\subseteq\overline{\mathcal{W}(N)}$$

As standard example: $$\left\langle\sigma\begin{pmatrix}0&1\\0&0\end{pmatrix}\right\rangle=(0)\subsetneq\frac12\mathbb{D}=\overline{\mathcal{W}\begin{pmatrix}0&1\\0&0\end{pmatrix}}$$

But inclusion always holds!

Unbounded Operators

For normal operators: $$N^*N=NN^*:\quad\langle\sigma(N)\rangle=\overline{\mathcal{W}(N)}\subseteq\mathbb{C}$$

For symmetric operators: $$S\subseteq\overline{S}\subsetneq S^*:\quad\langle\sigma(S)\rangle\nsubseteq\mathbb{R}\quad\overline{\mathcal{W}(S)}\subseteq\mathbb{R}$$

For unclosed operators: $$T\subsetneq\overline{T}=T^{**}:\quad\langle\sigma(T)\rangle=\mathbb{C}\supseteq\overline{\mathcal{W}(T)}$$

It may happen even:* $$S=\overline{S}\subsetneq S^*:\quad\langle\sigma(S)\rangle=\mathbb{C}\supsetneq\mathbb{R}\supseteq\overline{\mathcal{W}(S)}$$

And even worse also:* $$T=\overline{T}=T^{**}:\quad\langle\sigma(T)\rangle=\varnothing\subsetneq\overline{\mathcal{W}(T)}\subseteq\mathbb{C}$$

So inclusion mostly fails!

References

For examples see: Closed Operators

For the proof see: Normal Operators, Spectral Measures