Spectrum of Bounded Linear Operator is Non-Empty
![]() | This article needs to be tidied. Please fix formatting and $\LaTeX$ errors and inconsistencies. It may also need to be brought up to our standard house style. To discuss this page in more detail, feel free to use the talk page. When this work has been completed, you may remove this instance of {{Tidy}} from the code. |
![]() | This article needs to be linked to other articles. You can help $\mathsf{Pr} \infty \mathsf{fWiki}$ by adding these links. To discuss this page in more detail, feel free to use the talk page. When this work has been completed, you may remove this instance of {{MissingLinks}} from the code. |
Theorem
Let $B$ be a Banach space over $\C$.
Let $\map {\mathfrak L} {B, B}$ be the set of bounded linear operators from $B$ to itself.
Let $T \in \map {\mathfrak L} {B, B}$.
Then the spectrum $\map \sigma T$ of $T$ is non-empty.
Proof
Let $f: \C \to \map {\mathfrak L} {B, B}$ be the resolvent mapping defined as $\map f z = \paren {T - z I}^{-1}$.
Aiming for a contradiction, suppose the spectrum of $T$ is empty, so that $\map f z$ is well-defined for all $z \in \C$.
We first show that $\norm {\map f z}_*$ is uniformly bounded by some constant $C$.
Observe that:
- $(1): \quad \norm {\map f z}_* = \norm {\paren {T - z I}^{-1} }_* = \dfrac 1 {\size z} \norm {\paren {I - \dfrac T z}^{-1} }_*$
For $\size z \ge 2 \norm T_*$, Operator Norm is Norm implies that:
- $\norm {\dfrac T z}_* \le \dfrac {\norm T_*} {2 \norm T_*} = \dfrac 1 2$
Hence by $(1)$ and Invertibility of Identity Minus Operator, we get:
\(\ds \norm {\map f z}_*\) | \(=\) | \(\ds \frac 1 {\size z} \norm {\sum_{j \mathop = 0}^\infty \paren {\frac T z}^j}_*\) | ||||||||||||
\(\ds \) | \(\le\) | \(\ds \frac 1 {\size z} \sum_{j \mathop = 0}^\infty \frac {\norm T_*^j} {\size z^j}\) | by Triangle Inequality and Operator Norm on Banach Space is Submultiplicative on each term | |||||||||||
\(\ds \) | \(\le\) | \(\ds \frac 1 {2 \norm T_*} \sum_{j \mathop = 0}^\infty \frac {\norm T_*^j} {\paren {2 \norm T_*}^j}\) | as $\size z \ge 2 \norm T_*$ | |||||||||||
\(\ds \) | \(=\) | \(\ds \frac 1 {2 \norm T_*} \sum_{j \mathop = 0}^\infty 1/2^j\) | ||||||||||||
\(\ds \) | \(<\) | \(\ds \infty\) |
Therefore, the norm of $\map f z$ is bounded for $\size z \ge 2 \norm T_*$ by some constant $C_1$.
Next, consider the disk $\size z \le 2 \norm T_*$ in the complex plane.
It is compact.
From Resolvent Mapping is Continuous $\map f z$ is continuous on the disk.
From Norm is Continuous, we get from Continuous Function on Compact Space is Bounded that $\norm f_*$ is bounded on the this disk by some constant $C_2$.
Thus, $\norm {\map f z}_*$ is bounded for all $z \in \C$ by $C = \max \set {C_1, C_2}$.
Finally, pick any $x \in B$ and $\ell \in B^*$, the dual of $B$.
Define the function $g: \C \to \C$ as:
- $\forall z \in \C: \map g z = \map \ell {\map f z x}$
Since $f$ has empty spectrum, Resolvent Mapping is Analytic and Strongly Analytic iff Weakly Analytic together imply that $g$ is an entire function.
Thus we have:
\(\ds \size {\map g z}\) | \(=\) | \(\ds \size {\map \ell {\paren {T - z I}^{-1} x} }\) | ||||||||||||
\(\ds \) | \(\le\) | \(\ds \norm \ell_{B^*} \norm {\paren {T - z I}^{-1} }_* \norm x_B\) | since $\ell$ and $\paren {T - z I}^{-1}$ are bounded by assumption | |||||||||||
\(\ds \) | \(\le\) | \(\ds \norm \ell_{B^*} \norm x_B C\) | by the above | |||||||||||
\(\ds \) | \(<\) | \(\ds \infty\) |
So $g$ is a bounded entire function.
It is therefore equal to some constant $K$ by Liouville's Theorem.
But the inequality above:
- $\size {\map g z} \le \norm \ell_{B^*} \norm {\paren {T - z I}^{-1} }_* \norm x_B$
together with Resolvent Mapping Converges to 0 at Infinity, implies that:
- $\ds \size K = \lim_{z \mathop \to \infty} \size {\map g z} \le 0$
So $g$ is the constant function $0$.
We have therefore shown that:
- $\forall x \in B, \ell \in B^*: \map \ell {\map f z x} = 0$
This implies from Condition for Bounded Linear Operator to be Zero that $\map f z = 0$.
In particular:
- $\map f 0 = T^{-1} = 0$
But this is a contradiction, since our assumption that the spectrum of $T$ is empty implies that $T$ has a two-sided bounded inverse.
$\blacksquare$