Why bounded linear operators on Hilbert space are defined on the whole space

434 Views Asked by At

I've read different references about bounded operators but some are unclear with what they used and others justify but I'm unable to figure out where it comes.

They defined that a linear operator $T: H \to K$, with $H$ and $K$ two Hilbert Spaces, is bounded if $D(T)=H$ ($D(T)$ is the domain of $T$) and $$\sup \{\|Tx\|\ | \|x\|\le 1\} < \infty.$$

The point that bothers me is that $D(T) = H$, some justify it by the fact a bounded linear operator can be extended to $\overline{D(T)}$ and that $D(T)$ is dense in $H$ so we can always work with operator defined everywhere because all the others can be extended to $H$. But why $D(T)$ is dense in $H$ ? Is there the result which states that ? A reference says that most of the operators we use are densely defined, may be they assume that we only work with this kind of operators.

I can solve my problem by thinking that we only use densely or everywhere defined operator but the problem resurfaces when we talk about inverse of a bounded operators. If an operator $T: H \to H$ is 1-1 when can talk about $T^{-1}$ with $D(T^{-1}) = \text{im}(T)$ but I think there is not reason for $T^{-1}$ to be densely or everywhere defined. Then they present the following result : $$T \text{ has a bounded inverse iff } T \text{ is bijective}$$ But what does they mean by inverse ? A everywhere defined inverse ? I have difficulties to prove $\Rightarrow$ because I don't really know what's the definition of bounded inverse.

Here are the references I read :

  • Spectral Theory of Operators on Hilbert space, A. D. Andrew and W. L. Green, 2002
  • A primary introduction to spectral theory, lecture from Jianwei Yang in 2019.
  • Wikipedia and forums.