Hahn-Banach Theorem/Complex Vector Space
Theorem
Let $X$ be a vector space over $\C$.
Let $p : X \to \R$ be a seminorm on $X$.
Let $X_0$ be a linear subspace of $X$.
Let $f_0 : X_0 \to \C$ be a linear functional such that:
- $\cmod {\map {f_0} x} \le \map p x$ for each $x \in X_0$.
Then there exists a linear functional $f$ defined on the whole space $X$ which extends $f_0$ and satisfies:
- $\cmod {\map f x} \le \map p x$ for each $x \in X$.
That is, there exists a linear functional $f : X \to \C$ such that:
- $\cmod {\map f x} \le \map p x$ for each $x \in X$
and:
- $\map f x = \map {f_0} x$ for each $x \in X_0$.
Corollary
Let $\struct {X, \norm \cdot}$ be a normed vector space over $\C$.
Let $X_0$ be a linear subspace of $X$.
Let $f_0 : X_0 \to \C$ be a bounded linear functional.
Then $f_0$ can be extended to a bounded linear functional $f : X \to \C$ with:
- $\norm f_{X^\ast} = \norm {f_0}_{\paren {X_0}^\ast}$
where $\norm \cdot_{X^\ast}$ and $\norm \cdot_{\paren {X_0}^\ast}$ are the norms of the normed dual spaces $X^\ast$ and $\paren {X_0}^\ast$.
Proof
Define $g_0 : X_0 \to \R$ by:
- $\map {g_0} x = \map \Re {\map {f_0} x}$
for each $x \in X_0$.
By Real Part of Linear Functional is Linear Functional, $g_0$ is an $\R$-linear functional.
Also, define $h_0 : X_0 \to \R$ by:
- $\map {h_0} x = \map \Im {\map {f_0} x}$
for each $x \in X$.
By Imaginary Part of Linear Functional is Linear Functional, $h_0$ is an $\R$-linear functional.
Now, for each $x \in X$, we have:
\(\ds \cmod {\map {g_0} x}\) | \(=\) | \(\ds \sqrt {\paren {\map {g_0} x}^2}\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \sqrt {\paren {\map {g_0} x}^2 + \paren {\map {h_0} x}^2}\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \cmod {\map {f_0} x}\) | ||||||||||||
\(\ds \) | \(\le\) | \(\ds \map p x\) |
Let $X_\R$ be the realification of $X$.
By Hahn-Banach Theorem: Real Vector Space on $X_\R$, there exists a $\R$-linear functional $g : X \to \R$ extending $g_0$ and satisfying:
- $\cmod {\map g x} \le \map p x$
for each $x \in X$.
Define $f : X \to \C$ by:
- $\map f x = \map g x - i \map g {i x}$
for each $x \in X$.
Then for $\lambda, \mu \in \R$ and $x, y \in X$ we have:
\(\ds \map f {\lambda x + \mu y}\) | \(=\) | \(\ds \map g {\lambda x + \mu y} - i \map g {i \paren {\lambda x + \mu y} }\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \lambda \map g x + \mu \map g x - i \lambda \map g {i x} - i \mu \map g {i x}\) | since $g$ is $\R$-linear | |||||||||||
\(\ds \) | \(=\) | \(\ds \lambda \map f x + \mu \map f y\) |
so $f$ is $\R$-linear.
To show that $f$ is $\C$-linear, we need to show that $\map f {i x} = i \map f x$ for each $x \in X$.
We have:
\(\ds \map f {i x}\) | \(=\) | \(\ds \map g {i x} - i \map g {-x}\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \map g {i x} + i \map g x\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds i \paren {\map g x - i \map g {i x} }\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds i \map f x\) |
We now want to show that $f$ extends $f_0$.
Let $F_0$ be the restriction of $f$ to $X_0$.
We want to show that $f_0 = F_0$.
We have:
- $\map \Re {\map f x} = \map g x$
for each $x \in X$, and so:
- $\map \Re {\map f x} = \map {g_0} x = \map \Re {\map {f_0} x}$
for each $x \in X_0$.
So:
- $\map \Re {\map {F_0} x} = \map \Re {\map {f_0} x}$
for each $x \in X_0$.
From Linear Functional on Complex Vector Space is Uniquely Determined by Real Part, it follows that:
- $\map {F_0} x = \map {f_0} x$
for each $x \in X_0$.
So $f$ indeed extends $f_0$.
Now take $x \in X$.
Pick $\lambda \in \C$ such that $\cmod \lambda = 1$ and:
- $\lambda \map f x = \cmod {\map f x}$
Then we have, since $f$ is $\C$-linear:
\(\ds \cmod {\map f x}\) | \(=\) | \(\ds \lambda \map f x\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \map f {\lambda x}\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \map g {\lambda x} - i \map g {i \lambda x}\) |
Since:
- $\cmod {\map f x} \in \R$
we have that $\map g {i \lambda x} = 0$.
Finally, we have:
\(\ds \cmod {\map f x}\) | \(=\) | \(\ds \map g {\lambda x}\) | ||||||||||||
\(\ds \) | \(\le\) | \(\ds \map p {\lambda x}\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \cmod \lambda \map p x\) | Seminorm Axiom $\text N 2$: Positive Homogeneity | |||||||||||
\(\ds \) | \(=\) | \(\ds \map p x\) |
So $f$ is a linear functional satisfying our requirements.
$\blacksquare$
Source of Name
This entry was named for Hans Hahn and Stefan Banach.