Operator with Zero Numerical Range is Zero Operator
Theorem
Let $\struct {\HH, \innerprod \cdot \cdot}$ be a Hilbert space over $\C$.
Let $\norm {\, \cdot \,}$ be the inner product norm on $\struct {\HH, \innerprod \cdot \cdot}$.
Let $\struct {\map D T, T}$ be a densely-defined linear operator on $\HH$ such that:
- $\map W T = \set 0$
where $\map W T$ is the numerical range of $T$.
Then $T = 0$.
Corollary
Let $\DD$ be a dense linear subspace of $\HH$.
Let $\struct {\DD, T}$ and $\struct {\DD, S}$ be densely-defined linear operators on $\HH$ such that:
- $\innerprod {T x} x = \innerprod {S x} x$ for each $x \in \DD$.
Then $T = S$.
Proof
Let $x, y \in \map D T$.
Since $\map W T = \set 0$, we have:
- $\innerprod {\map T {x + y} } {x + y} = \innerprod {T x} x = \innerprod {T y} y = 0$
From Inner Product is Sesquilinear, we have:
- $\innerprod {T x} x + \innerprod {T x} y + \innerprod {T y} x + \innerprod {T y} y = 0$
That is:
- $\innerprod {T x} y = -\innerprod {T y} x$
Similarly, replacing $y$ with $i y$ and applying Inner Product is Sesquilinear, we have:
- $-i \innerprod {T x} y = -\innerprod {\map T {i y} } x$
Since $T$ is linear, we have:
- $\innerprod {T x} y = \innerprod {T y} x$
Hence:
- $\innerprod {T x} y = -\innerprod {T x} y$
This gives:
- $\innerprod {T x} y = 0$ for all $x, y \in \map D T$.
That is:
- $T x \in \map D T^\bot$ for all $x \in \map D T$
where $\map D T^\bot$ is the orthocomplement of $\map D T$.
From Linear Subspace Dense iff Zero Orthocomplement, we have $\map D T^\bot = \set 0$.
Hence:
- $T x = 0$ for all $x \in \map D T$.
$\blacksquare$
Sources
- 1991: Walter Rudin: Functional Analysis (2nd ed.) ... (previous) ... (next): $12.7$: Bounded Operators