next up previous
Next: Orthonormal Sets and Bases Up: Introduction Previous: Structure of a Hilbert

Linear Transformations

Let $ H, K$ be Hilbert spaces and let $ T: H \rightarrow K$ be a linear transformation.

Properties The following are equivalent:

  1. $ T$ is continuous.
  2. $ T$ is continuous at zero.
  3. $ T$ is continuous at a point
  4. $ T$ is bounded; i.e., % latex2html id marker 12163
$ \exists c > 0$ such that $ \Vert T(h)\Vert \leq c \Vert h\Vert$ for all $ h \in H$.

We define

$\displaystyle \Vert T\Vert = \sum \{ \Vert T(h)\Vert : \Vert h\Vert \leq 1 \} $

Monday, 8-29-05:

Definition 1.3.1   A linear functional on a vector space $ V$ over $ \mathbb{F}$ is a linear transformation

$\displaystyle T : V \rightarrow \mathbb{F}$

(where $ \mathbb{F}$ is regarded as a vector space over itself).

Example: If $ \mathcal{H}$ is a Hilbert space and $ f \in \mathcal{H}$, then the map $ T: \mathcal{H}\rightarrow \mathbb{F}$ given by $ T(h) = \left\langle{ h, f }\right\rangle $ is a bounded linear functional. By the CBS inequality, $ \Vert T \Vert \leq \Vert f \Vert$. In fact, $ \vert T(f/\Vert f\Vert)\vert = \vert\left\langle{ f,f }\right\rangle \vert/\Vert f\Vert = \Vert f\Vert$. Thus, $ \Vert T\Vert = \Vert f\Vert$.

Theorem 1.3.2   Riesz-Representation Theorem (for Hilbert spaces)
If $ T: \mathcal{H}\rightarrow \mathbb{F}$ is a bounded linear functional, then there is a unique $ g \in \mathcal{H}$ such that

$\displaystyle T(h) = \left\langle{ h,g }\right\rangle $

for all $ h \in \mathcal{H}$.

Proof. If $ T = 0$, clearly $ f = 0$ (uses the fact we have an inner product, and not a semi-inner product).
Assume $ T \neq 0$. Then $ \ker T \neq H$. Pick $ f \in (\ker T)^\perp$ such that $ T(f) = 1$. For $ h \in \mathcal{H}$, consider $ h - T(h) f$. Then,

$\displaystyle T(h - T(h) f) = T(h) - T(h) T(f) = 0 $

Thus,

$\displaystyle \left\langle{ h - T(h) f, f }\right\rangle = 0 $

Distributing this out,

$\displaystyle \left\langle{ h,f }\right\rangle = T(h) \Vert f\Vert^2 $

Thus,

$\displaystyle T(h) = \left\langle{ h, \frac{f}{\Vert f\Vert^2} }\right\rangle $

Let $ \displaystyle g = \frac{f}{\Vert f\Vert^2}$.

If $ k$ is another vector in $ H$ with $ T(h) = \left\langle{ h,k }\right\rangle $ for all $ h \in H$, then

$\displaystyle \left\langle{ h, g - k }\right\rangle = T(h) - T(h) = 0 $

for all $ h$, thus $ g - k \in H^\perp = \{\vec 0 \}$. $ \qedsymbol$


next up previous
Next: Orthonormal Sets and Bases Up: Introduction Previous: Structure of a Hilbert
Brian Bockelman 2005-12-12